Copyright©2014GregoryCristal

CoverDesignbyAndreasBrooks

Allrightsreserved.

ISBN:1484867572ISBN-13:978-1484867570

CONTENTSSectionName

Purpose-Whywriteabookaboutthirdpartyadserving?

PartA-SettingtheStage

Theflowofthisbook

Whatisthirdpartyadserving?

GettingSomePerspective

Users

Publisher

Agencies

TheAdvertiser

DefinitionsandMethodology

1Chapter1-Section1-BrowsersandDevices

TheroleofthebrowserinDigitaladvertising

Figure1.1

Figure1.2

Trackingacrossmultiplebrowsers

firstpartyVs.thirdPartyCookies

Figure1.3

CookieDeletion

DistinctionbetweenDevicesandComputers

AdblockingTools

WhyisthereadvertisingontheInternet?

Chapter1-Section2-TheAnatomyofSitesandAds

Whattheusersees

Figure1.4

HTML

Figure1.5

JavaScript

Images

TextLinks

Figure1.6

Flash

TheadventofHTML5

Chapter1-Section3-AdServing

Whatdoestheadserverdo?

Figure1.7

TableofDefinitions(Table)

Figure1.8

TheroleoftheCDN

WaitingaroundforAdDelivery-theuserexperience

Pre-addisplay

Duringaddisplay

Onadclose

2Chapter2-TheIndustryPlayers

TheAgencyholdinggroups

WhoarethebigAgencies?

RunningaMediaAgency

CreativeAgencies

Figure2.1

Publishers

Chapter2-Section1-AdvertisersandMarketers

WhoistheAdvertiser?

TheMarketer

Agencypitch

RelationshipwiththeMediaAgency

RelationshipwiththeCreativeAgency

Figure2.2.1ControllingthepagesoftheAdvertiserSite-theCMS

LandingPage

BasketPage

ConfirmationPage

Figure2.2.2

CRMPlatform

OnsiteBehavioraldata

Trackingtheusersonsite-SiteSideAnalytics

ClicktoLand

Passingthetokensintheclick-throughURL

Referral

Chapter2-Section2-AgencyPlanningandBuying

TheroleofthePlanner

Howtoplan?

Figure2.3

ResearchTools

Figure2.4

DemographicDigitalMediaConsumptionbyPublisherWebsite

UniqueAudience-SiteOverlap

ResearchthroughBenchmarking

In-Agencyplanningsolutions

Figure2.5

TechnologyforBuying

AgencyTradingDesks

Figure2.6

TheAgency-sideMediaPlan

SelectinganAdServer

Whattechnology-basedServicesdoMediaAgenciesoffer?

MediaAgencydataanalysts

Chapter2-Section3-CreativeShop

TheCreativeShop

RelationshipwiththeAdvertiser

RelationshipwithPublishers

RelationshipwiththeMediaAgency

MarketingbudgetforCreativedevelopment

RolesandResponsibilities

Figure2.7

GettingtheDesignright

DevelopmentandProduction

Flashauthoring

Figure2.7.1

Actionscript

Figure2.7.2

Components,buildingandpublishingads

Adstoragesystems(AssetManagementSystems)

Localization

StandardQA

Figure2.7.3

UploadingCreative

What’sinaname?

CreativeInspiration

Chapter2-Section4-Publishers

WhatarePublishers?

Figure2.8.1

HowdoPublishersselltheiradspace?

CPM

CPC

CPA

PublisherNetworks

Figure2.8.2

NicheNetworks

BlindNetworks

Publisheradservers

Figure2.8.3

Inventory

PublisherSpecifications

PublisherQA

PublisherAdOps

Figure2.8.4

Adprioritization

Frequencycapping

OverandUnderDelivery

BrowserTargeting

Figure2.8.5

HowdoesPublisherreportingdifferfromthirdpartyreports?

Discrepancies

SectionB-CampaignSetupbyChannel

Figure2.8.6

HowdoestheendtoendCampaignManagementworkflowwork?

Billing,ElectronicIOandAdservingrates

3Chapter3-AdservingOperations

Trafficking

ThethirdpartyadserveruserInterface

UserAccessforcampaignsetup/Permissions

Usingmultiplethirdpartyad-servers(double-tracking)

Figure3.0.1

Figure3.0.2

Thegoldenruleofadserving

How does the trafficker check the ads areworking? (Setup verificationusinganHTTPtracer)

Chapter3-Section2-Displayadservingoperations

Dayinthelifeofthetrafficker

FoundationLeveltraffickingforstandarddisplay

1.Creatingaplaceintheadservertosetupthecampaign

Thethirdpartyadserverhierarchy

Figure3.1SettingsatAgencyandAdvertiserlevels-SiloCapabilities

Figure3.2

SettingsatAgencyandAdvertiserlevels-SharingCapabilities

2.StructuringthecampaignsetuptoreplicatethePublishersonthemediaplan

MediaPlan

Figure3.2.1

TraffickingSheetsforoutsourcetrafficking

Sites

3.Identifytheformatsoftheadsandcreatingtheplacements

AdFormats

Banner

Leaderboard

Skyscraper

MPU

RoadblockandSkins

PlacementsandPackages

CostPackages

4.Assignthecreativeandsettingrotationcontrols

NamingConventions

Ad&Creative

AutomatedQA

AdPreview

Addingtheclick-throughURL:clickTag

302Redirects

Addingmultipleclick-throughURLs

Adclassifications/CustomCategoriesandsearching

AdDuplication

Figure3.2.2

Figure3.2.3

ReadyingtheAdsforAssignment

Backupgif

PossiblePlacementAssignments

Figure3.2.4

Adassignment

Multipleadstoasingleplacement

Asingleadtomultipleplacements

Asingleadtoasingleplacement

Figure3.2.5

AdjustingtheAdControls

FrequencyCapping

AdRotation

Sequencing

Weighting

AdPrioritization

TimeBased

PostCapGif

Thetwolevelsofcontrol

Figure3.2.6

5.TraffickingtheTagstothePublishers

AdTagsandTagFormats

Figure3.2.7

Flighting

ThoroughlyCheckinganAdislive

6.HowdoesatraffickermanagecampaignswithhundredsofAds?

BulktraffickingthroughExceltemplates

BulktraffickingwithExcelastheUI

AutomatedtraffickingsetupviaadserverAPI

Figure3.2.7.1

TheJavaScriptcarriagecapability

SurveyvendorCodeandthirdpartyscripts-inads

Figure3.2.8

Trackingbutnotserving

IsitpossiblethataPublishercannotsupportthirdpartyadserving?

Figure3.2.9

AnoteonClickCommands

7.AdReplication

AdDeliveryandReplication

Server-SideCaching

Whatexactlyisatag?

8.MakingabiggerimpactwithDisplayCampaigns

Chapter3-Section3-RichMedia

WhatisRichMedia?

WhyuseRichMedia?

CoreFormats

Expandableads

Figure3.3.1a

Floatingads

Figure3.3.1b

Videoads

Figure3.3.1c

PushdownandSidekick

Figure3.3.1d

DataCapture

HomePageTakeovers,WallpaperandPageSkins

Figure3.3.1e

Figure3.3.1f

Polling

StateBasedads

LocalConnectionadsorSyncAds

Figure3.3.1g

Printads

StandardizationandPublishercertification

Fromthetop:BuildingRichMediaads

1.Trainingthecreativedeveloper

2.TwoschoolsofRichMediaspecs:PublisherandAdserver

3.ParentandChildfiles-thepoliteload

4.ExpansionsandFloating-thedeclarationofstate

5.Interactiontracking:DefaultandCustomEvents

6.Interactionsasadditionalclicks

7.BuildingCustomRichMediaads

QaingandTestingRichMediaads

TraffickingRichMediaads

PublisherTroubleshooting

RichMediaforMobile&RichMediaforInstream

Chapter3-Section4-DCO

StandardXMLDisplay

Figure3.4.1

Figure3.4.2

DCODesignPhase

Creatingthedynamictemplate

Figure3.4.3

DynamicCreativePlanning

Figure3.4.4

DCO-AdCreationPhase

SingleVersioningorManualCreation

ExcelMassVersioning

MassVersioningfromaFeed

Figure3.4.5

UsingawebservicetorestrictloadonAdvertiserhostedimages

Figure3.4.6

DCO-AssignmenttoPlacementPhase

Placementvs.audiences

Figure3.4.7TargetingdifferentVersionstodifferentAudiences

Figure3.4.8

ConnectingDCOwithAudienceBuying

DCO-OptimizationPhase

Auto-OptimizeVs.ManualAdControls

Figure3.4.9

Auto-optimizationasa“BlackBoxTechnology”

Multi-varianttesting

Chapter3-Section5-InstreamVideo

WhatisInstream?

Figure3.5.1

PreRollMidRollPostRoll

Figure3.5.2

SiteServedVideo

Figure3.5.3

WheredoVideoAdNetworksaddvalue?

VAST-thevideotagstandard

CompanionAds

Figure3.5.4

AdSkippingandTrueView

VPAID-StandardizedInteractiveVideo

Figure3.5.5

VPAIDAuthoring

Figure3.5.6

DCOforInstream

Figure3.5.7

BeyondVASTandVPAID

PublisherCertification

Figure3.5.8

TraffickingInstreamads

Chapter3-Section6-Mobile

TheConnectivityConsideration

InternetEnabledDeviceDivergence

PublisherCertificationforMobile

Figure3.6.1

Challengeswithinthebrowser

ResponsiveDesign

WhateverhappenedtoFlash?

ThefallbackMechanism

TraffickingadsforMobileBrowsers

QAformobilebrowser-basedads

Figure3.6.2

Challengeswithintheapp

Figure3.6.3

TraffickingadsforIn-appadcampaigns

WheredoMobileAdNetworksaddvalue?

Chapter3-Section7-PaidSearchandNaturalSearch

SEMandPPC

CreatingCampaignsandKeywordsintheEngine

Figure3.7

Figure3.7.1

ConnectingaBidManagementSystemtotheEngine

Figure3.7.2

Whytrackpaidsearchthroughathirdpartyadserver?

HowtotrackPaidSearchthroughathirdpartyadserver?

ManualSearchTracking

Figure3.7.3

AutomatedSearchTracking

Figure3.7.4

AdServertrackingintegratedwithBidManagement

Figure3.7.5

ExtendingtrackedPaidSearchtothelongtail

Figure3.7.6

ExtendingtrackedPaidSearchtoMobile

SearchandotherChannels

WhatisSEO?

Figure3.7.7

ReferralURLtotriggeraclick

Figure3.7.8

AdServingcostsforSearchChapter3-Section8-AffiliatesandAffiliateNetworks

AffiliateSitesandNetworks

HowtheAffiliateEarns

AuditingthequalityoftheLongTail

Figure3.8.1

TrackingAffiliateActivityinthirdpartyadserving

Figure3.8.2

Theabsenceofimpressiontracking

Affiliatesthatgrow

Chapter3-Section9-Email,in-gameandSocial

PermissionMarketing

Figure3.9.1

SendingEmailsonMass

EmailMarketingadvertisingmetrics

ThirdpartyadservertrackingforEmail

In-gameanddigitaloutdoortracking

SocialMediatracking

Figure3.9.2

Theversatilityoftrackingpixels4Chapter4-Section1-ConversionTagsandAttribution

Whatareconversions?

Figure4.1

Wheretoplacetheconversiontag?

Figure4.1.1

Wheretohaveconversiontags(table)

CPA

Conversiontagtypes-SalesVs.CounterTags

Figure4.1.2

Extendeddata/ExtendedParameters

Figure4.1.3

Figure4.1.4

Howdoyoutestconversions?

A. ConversionTagFunctionality-Piggybacking

ContainerTags

CertifiedContainerTagproviders

Whichconversiontagfiresfirst?

SelectivePinging

Figure4.1.5

PiggybackingwithExtendeddata

Figure4.1.6B. ConversionTagFunctionality-ConversionWindows

WindowSettings

Post-impressionandPost-clickwindows

LongandShortConversionWindows

ShadowTagging

ReportingonConversions

RepeatConversionWindowsC. ConversionTagFunctionality-AttributionModeling

Figure4.1.7

AttributionSettings

EngagementMapping

Figure4.1.8

CustomAttributionModeling

Challengesinimplementingconversiontags

TagManagementSystems

Figure4.1.9

UniversalTaggingSecurity

Chapter4-Section2-Audiencesandtargeting

AudiencesandTargetingatamedia/inventorylevel

AudiencesandTargetingatacreativelevel

Thirdpartyadservertargetingrules

LayeredTargeting

Figure4.2.1

Geo-targeting

Behavioraltargeting

Figure4.2.2

Contextualtargeting(page-scraping)

Figure4.2.3

Searchkeywordtargeting

TargetingbasedonVerificationData

AdditionalDatatargeting

A/BTesting

Figure4.2.4

Trulift

Figure4.2.5

AudienceAssignment

Channelswhichbendtherules

SectionC-AdServerAnalytics

5Chapter5-ReportingTechnology

Whatisadserverreporting?

AggregatedData

Figure5.0.1

Figure5.0.2

ReportingAPI

ThirdpartyAdServingvs.SiteSideAnalytics

Figure5.0.3

WhenistheDataProcessed?

Realtimereporting

ProcessingClock

Whereisthedataprocessed?

Howisthedatafiltered?

Whereisthedatadisplayed?

Chapter5-Section1-BasicAggregatedReporting

QAReports

Globalreporting

Figure5.1.1

Deliveryreports

VerificationReports

PerformanceReports(ConversionandAttributionReports)

CostandROIreports

404errorreporting

Expiringcostpackages

ChannelSpecificReports

TheStandardMetrics

Impressions

Anewimpression?

Figure5.1.2

ClicksandClick-throughRate(CTR)

Conversions&Conversionrate

Chapter5-Section2-AdVerification&Viewability

EffectiveandIneffectiveImpressions

Reactive,PreventativeandPre-EmptiveVerification

Figure5.2.1

PublisherandplacementQualityindicators

ContextualVerificationQualityIndicators

ContentClassification

BrandSafety

Figure5.2.2

DemographicorAudienceVerificationThepanel-basedAudienceVerificationMethod

Figure5.2.3

Thesocialprofile-basedAudienceVerificationMethod

Figure5.2.4

DiscoverabilityVerificationQualityIndicators

GeoReporting

DeviceVerification

Viewableimpressions

Figure5.2.5

DoubleImplementationsandFraudulentServingtechniques

Figure5.2.6

PageLevelVerificationQualityIndicators

DomainURLReporting

OtherusesforPageLevelinformation

WhiteListsandBlackLists

Figure5.2.7Seethrough,ViewthroughandRecordabilityRates

Figure5.2.8

Preventative-Adblocking

Pre-emptive-VerificationinProgrammatic

TheVerificationlandscape

Chapter5-Section3-AdvancedAggregatedReporting

CrossChannelanalysis

DayPartAnalysis

CrossChannelReports

Figure5.3.1

AdvancedConversionAnalysis

TimetoConvert

Figure5.3.2

EngagementMappingFactorsreport

ReachandFrequencyReporting

UniqueReach

AverageFrequency

Figure5.1.3

EffectiveFrequency

Figure5.1.4

ReachandFrequencyReport

Figure5.1.5

AdjustedUnique

Figure5.1.6

AdvancedSiteandCreativeAnalysis

UniqueSiteOverlap

Figure5.1.7

Adformatreporting

GlobalBenchmarks

AdvertiserSpecificBenchmarks

Figure5.1.8

Chapter5-Section4-ChannelSpecificReporting

DisplayChannelReporting

RichMediaChannelReporting

DCOChannelReporting

Multivarianttesting

DangersinVersionOverwriting

Versionreporting

Figure5.4.1

InstreamChannelReportingandMetrics

VideoStartedanduniquevideostarts

Adduration

Incrementadplay-through(25,50,75,100%)

Average%videocompletion

VPAIDcustominteractions

MobileChannelReporting

CookieDuplication

CookieAbsence

DeviceFingerPrinting

Cookiesinapplications

DeviceandCarrierreporting

PaidSearchChannelReporting

SearchOverview

KeywordPathwayAnalysis

NaturalSearchReporting

DoublecountingandSingleCounting

Figure5.4.2

AffiliateChannelReporting

Chapter5-Section5-CookieLevelData

ReportingattheCookieLevel

Figure5.5.1

Howisthedatasupplied?

Look-backDataDump

CookieLevelFeed

Reportdelivery,sizeestimationandstorageconsiderations

MetricsinCookieLevelData

Trackingimpulseconversions

PathwaytoConversionReporting

Figure5.5.2

PointSolutionsforAdvancedCookieLevelAttribution

ReportingonConversiontagActions

ExtendeddataattheCookieLevel

Figure5.5.3

Figure5.5.4

ImpressionandClickbasedtokendataattheCookieLevel

Whatismissingfromadserverreporting?

6Chapter6-Optimization

Questioningreportingdata

1.Wheredidthereportcomefrom?

2.Whatkindofreportisit?

3.Dothenumberslookright?Doesthedataaddup?

4.Whattimewasthereportpulled,whattimezoneistheaccountinandarethesethecorrectdateranges?

5.Arethesethecorrectmetrics,conversiontagsandhierarchyentities?6.Dothenamesoftheentitiesinthereportmatchthoseinthecampaignmanagementinterface?

7.Whatisthemethodologybehindthesemetrics?

8.Istheshowngranularityprovidingenoughortoomuchinformation?

9. Is there enough data in the report or were the traffic volumes largeenoughtogaugedecentstatisticalaccuracy?

10.Isthereachancethatthedatainthereportbeadjustedagainlateron?

11.Arealltherowsandcolumnsdisplaying?

12.Wasthecampaignsetupinwaythatmightaffectagovernedoutcome?

13.Didtheadserverencounteranykindofcountingfault?

WhatisOptimization?

OptimizingtoCampaignObjectives

OptimizingtowardsDeliveryGoals(idealforBrandAdvertisers)

UniqueReach

TargetAudience

Verification(qualitysafety)goalsandimpact

OptimizationstowardsPerformanceGoals(idealforDirectResponseAdvertisers)

EngagementandInteraction

TotalClicks

TotalConversionsorE-Conversions

ROI(&extendeddata)goals

OptimizationSample

RecordingchangesinOptimizationstrategy

7Chapter7-RetargetingandPrivacy

RetargetingTags

Figure7.1

CookiePools

RetargetingWindows

RetargetingforCreativeOptimization

Figure7.2

DynamicRetargeting

Figure7.3

CarouselAds

Figure7.4

PrivacyandBehavioraltargeting

TheAdChoicesProgram

Figure7.5

DoNotTrack

RetargetingforMediaOptimization

8Chapter8-Programmatic

AdvertisingExchanges

SupplySidePlatforms(SSPs)forthePublishers

DemandSidePlatforms(DSPs)fortheAdvertiser

Figure8.1

ConnectingtheadservertotheDSP

Figure8.2

QualityInventorySources

TheSecondHighestBidAuction

RealTimeBidding(RTB)

CostReportingintheadserver

DSPStrategies

DSPRetargetingwithDCO

Figure8.3

DSPKeywordTargeting

NegativeRetargeting

DSPdeliverycontrols

BuildingPrivateExchanges

Figure8.4

StrategiesbasedonPublisherAudienceDescriptors

CookieLevelDataProviders

DataManagementPlatforms(DMP)

Figure8.5

PageLevelDataProviders

Channel-SpecificComplicationsinRTB

DSP-adserveroftheFuture?

9Chapter9-TheBigPicture

Figure9.1-DisplayandProgrammatic

Figure9.2-LayeringinPaidSearch

Figure9.3-LayeringinAffiliates

Figure9.4-LayeringinUsers,AdvertiserProperties,AdsandTrafficking

Figure9.5-LayeringinRetargeting

Figure9.6-ThirdPartyAdServing-theDiagram

ClosingChapter

FurtherReadingandStudy

Acknowledgements

Glossary

PURPOSE

Whywriteabookaboutthirdpartyadserving?Working in online media or digital advertising, there are certain levels of technicalexpertisewhichexistacrossanyconnectedorganisationsinthespace.Fromadistanceitwouldbeeasytoassumethateveryonehasthesameleveloftrainingbutwhenitcomestousingthetechnicaltoolswhichenabletoworkthatwedo,thespaceiscomplexandtheideaoflearningeverythingisdaunting.SomeofusthatworkinOnlinebelievewedonotneed to learneverything, that the task is toogreatand thenecessity, trivial.The truth isthat the knowledge in any online advertising organisation is unfairly balanced betweenthose that take the time to learnasmuchas theycanand those that relyon the learned.Learning the role of theMediaAgencyplanner,TheAdvertisermarketer,TheCreativeAgencydeveloperorthetraffickeronlyliftsthelidonacornerofthecomplextechnicalworkflowofwhichweareallmandatoryparts.

Iwroteabookaboutadservingforonereason:therearenotenougheducationalresourcesaboutDigitalMarketing,OnlineAdvertisingandthirdpartyadservingtosatisfyanyofitsstakeholders.

Whena topic thiscomplexcomesalongandtheworkrequiresanacademicbackgroundwhich does not exist, the number of experts in the field shrinks and the circulation ofverifiable knowledge becomes thin. The knock on effect is a culture of assumptions, acultureofthefeartoasktoomanytechnicalquestionsofthefewthathavetheknowledgeandanenvironmentwhichisriskytoconductbusinesswithin.

This book provides in academia, a record of the technology as it stands today. I havegrownupwithtechnologythatisgrowingatthesamepace,andIamsharingwhatIknow.Ihavetriedmyverybesttonotmakeanyassumptionsabouthowthistechnologyworksbut despite asking many sources, the sources are too few to be completely verifiableacrosstheboardandforthisreasonIapologiseforanyincorrectassumptionsImayhavemadeabouthowthistechnologyworks.Myaimistokeeptheinformationuptodatebyoffering revised editions and maintaining an internet presence; I am open to technicalquestionsyoumayhaveregardingthismaterial.

Icurrentlyworkforawell-knownthirdpartyadserverbutIhavetriedmybestnottobebiastowardsitstechnology.Indeedtherearesomefeaturesmentionedinthisbookthatthetechnology provider I work for does not offer today but that exist elsewhere in theindustry.

Ireallyhopeyougetenjoymentandvaluefromthisbookandthatithelpsyourpersonalandprofessionalunderstandingofthesetechnologies.Iputalotofeffortintobringingthepartstogetherinadigestibleway(whichgiventhematrix-natureofthesubjectmatter,isnoteasy). Iwelcomewithopenarmsanyfeedbackyoumayhave regarding thecontentandanyopinionsImayhaveexpressedwithin(mycontributingeditorshavehelpedmetokeepthesetoaminimum).

Thanksforbuyingandhappyreading!

PartASettingtheStage

PartASettingtheStage

• Theflowofthisbook

• Whatisthirdpartyadserving?

• Gettingsomeperspective

• Users

• Publisher

• Agencies

• TheAdvertiser

• DefinitionsandMethodology

Theflowofthisbook

The interconnectednatureofOnlineAdvertising technologies remindsmeofneurons inthe brain. Their connected nature could be drawn two dimensionally and the resultingimagewould be a weave of connections. But the truth is that almost all nodes of thisimagewouldcommunicatewitheachother in someway, spanningdistances thatwouldmakeavisualconnectionin threedimensionalspace, toopainfulontheeyes toobserveand too detailed for the brain to absorb. I was asked by colleagues and my networkcontacts as I was planning this book as to how I would lay it out given this almostquantum nature of the technology landscape. I decided to focus on the campaignmanagementworkflowandbuildconceptsupfromthere.

Inthissensethebookstartswithsomebackgroundforthereader(PartA),relatingeachofthe concepts back to third party ad serving, such as the base technologies on a user’smachine and in “the Cloud” (Chapter 1) and the industry players or stakeholders thatassembleandpayeachothertobringanonlineadcampaigntolife(Chapter2).InPartBwe lookatcampaignmanagementbrokendownby thevariouschannelsavailable today(Chapter3)beforelookingattheconceptofconversiontracking(attributingasaleonlineas a result of an ad campaign). PartC then focusses on understanding the analytic andreportingcapabilitiesofanadserver(Chapter5)beforelookingatstrategiestoactonthatdatasuchasOptimization(Chapter6)Retargeting(Chapter7)andProgrammaticbuying(Chapter8).

Thisbookalsocontainsawholeheapofwordsandphrasesthathavebeenmoreor lessadopted across the industry. Due to the infancy of the digital advertising industry themeaningsofsomeofthesewordsarenotwhollyuniversalorunderstoodthereforethereisa convenient glossary located at the end of this book for your reference.Wordswhichappearinquotationmarkscanbereferencedintheglossaryattheback.

There are also some technology and company names that will appear in italics. Thesenames are used as examples of technologies thatmay not be available in everymarketaroundtheworldbutaregrowingorwelladoptedtechnologiesusedbybigAgenciesandAdvertisers in Europe and the US. To provide themost balanced understanding of the

technologylandscapeIhavetriedtoofferthenamesofasmanytechnologyprovidersaspossible although a large number of technology companies such asGoogle now offermany of the discussed features as part of their coreDoubleClick marketing suite. Thenames featuredare suggestedexamples. It is recommended that ifyouchoose toutilisethese technologies for digital marketing activity that you test their products first,preferably against a control groupwhere a technology is not in use and also against acompetitortoolwhichclaimstoofferthesameresults.Thison-goingexperimentationwillensurethatyouareusingthebesttoolsavailableinthemarkettoday.

As the flow of the various chapters suggests, the campaign management workflow iscyclical;thesetupofalltheadscanbechangedatanypointduringacampaignbasedontrends in the analytics data which is accessible just by logging into the ad servinginterface. In thisway the ad serving interface ismade up of two core areas: campaignmanagementandreporting.

Perhapsmostimportantlylet’saddresstheelephantintheroom:

Whatisthirdpartyadserving?

Athirdpartyad server is abusiness tobusinessmachineor system thatdoesa specificservice.Thismachinelivesinvariouslocationsaroundtheworldtocompletethetasksofthe service; it lives in “the Cloud”. As with other services that live in the Cloud it ispossibletoaccessthemonlyviaawebbrowser.Thirdpartyadservingcore“customers”areMediaAgenciesandAdvertisers.Thesecustomerspaytohaveaccesstobeabletologinto the third party ad serving interface.Once inside the interface, the tasks are brokendownintotwoseparateareas:campaignmanagementandreporting.

Customerschoosetouseathirdpartyadserverforthesemainreasons:

1. AccreditedIndependentthirdparty

2. ConsistentCountingMethodology

3. ReliabilitytoserveandServeruptime

4. FormatSupportandPublishercertification

5. SinglelocationReporting

6. De-DuplicatedReporting

7. CampaignSpecificNeedsandFeatures

Customersofthirdpartyadserverspayinitiallyfortheadspaceandthenpayatinyfeeontopforeachadverttobenefitfromtheabove.Thethirdpartyadservingspace,asasectionoftheadvertisingtechnologyindustryhasmatured,thethirdpartyadserversavailabletopotentialcustomerstodayisrathersmall.Infactthereareonlyreallytwotrulyglobalthirdpartyadserversleftchoosefrom:DoubleClickandDG/Mediamind.Theothershaveseendiminishing growth or have never seen adoption beyond a handful of countries and aglobaladserverisimportantsothatglobalAdvertisersareseeingtheabovebenefitsinallmarkets. These more local third party ad servers includeAdform,Mediaplex, AdRiver,Atlas, Flashtalking and Weborama. Thankfully there is uniformity among thesetechnologies-theyallworkinbasicallythesamewaysoalltheconceptsinthisbookare

valid across the board although one or two features may not be available in some adservers.

GettingSomePerspective

Nowwiththeglobalthirdpartyadserverhummingawayinthecloud,itstaskspermeateout across theweb affecting various groups of people. Let’s now look in detail atwhothesepeopleare:

Users

Everyoneisauser,weall“use”theinternet.Anyoneaccessingthe internetorbrowsingthe web is considered a user. This concept seems simple but if you work in onlineadvertisingyouareauserbutyoualsofallintoanotherofthecategoriesthatfollow.Whentalking to other media professionals about technology it always helps to declare yourposition: “In this problem that I am seeingwith this ad; I am taking on the role of theuser”.Userswill interactwith the thirdpartyadserverwhen loadingprettymucheverywebpageandthetotalnumberofusersintheworldisaroundacoupleofbillion.

Publishers

Ownawebsitewithadsonit?ThenyouareaPublisher(yourwebsiteisyourpublication).Ifyouownorrepresent lotsofwebsitesyouareaPublishernetwork.Publishers tendtoseetheworldfromtheirownperspective;theydonotlookbeyondtheirownnetworksanduse every tactic possible to drive “media spend” (advertising dollars) into their ownpocket. Advertisers and Agencies have tough jobs to do to spread that spend aroundeffectivelyandthisiswhereadservingcomesintoitsown.Publishersinteractingwiththethirdpartyadserverbyreceivingcodefromitshowthethirdpartyadsonitspagesandwillhavetoworkwithitsreportingsystems.HowmanyPublishersarethereintheworld:upwardsofhundredsofthousandsofbigsitesbutinactualfacthundredsofmillionsduetothenumberofwebsitesontheinternet.

Agencies

Agenciesfitintotwobuckets.FirstthereistheCreativeAgency;whoundertaketheworkofcreatingandbuildingthead(anddonotneedtotroublethemselvestoomuchwiththeworldofthirdpartyadserving).AndsecondlythereistheMediaAgency-amiddlemanintheadvertisingcampaignworkflow,managingandoptimizingtomakesureadvertisingdollars are spent effectively. TheMedia Agency has a tough job swimming through aminefieldoftechnologiestoachievethis.Anysinglepieceofkitthatcantrackalltheadservingdata,automateplanningandbuying,costingandreportingarethemagicbulletforasuccessfulagency.HowdoAgenciesuse thirdpartyadservers?TheCreativeAgencybuildsadsfortheadserveranduploadsthemontothesystem.TheMediaAgencywilldothecampaignmanagementandreporting.

TheAdvertiser

TheAdvertiser isanycompanywithproductsorservices tosell thathasenoughmoneyfor a marketing budget for advertising. References in this book to the Advertiser arereferences to the marketing team working for that Advertiser. To afford to utiliseenterpriseleveltechnologylikeanadservertherehavetobeafairfewmediadollarsinthebank for themarketer touse,yetwellover ten thousandAdvertisersgloballyutilise

these crucial technologies. Few Advertisers choose to stand alone without a MediaAgencyandgiven thecomplexnatureof thedigital environment it is no surprise.HowdoestheAdvertiseruseathirdpartyadserver?ForCampaignmanagementandreporting.

Selecting the rightad server isa trickybusiness, thereare somany features toconsiderand yet the differences between them are very small. This is all thanks to industrystandardisationwhich has been crucial to allowingAgencies andAdvertisers to switchfromoneAdservertoanother.Itallbeginswithdefinitionsandmethodology.

DefinitionsandMethodology

The Internet Advertising Bureau (IAB); arguably themostwidely recognised regularitybodygloballyfordigitalmarketingactivity,hasbeeninstrumentalintheshapingofdigitaladvertising. Standardisation of words and phrases is probably the most important andfundamentalconceptindigitaladvertising.AndalthoughtheIABhasdonealotofworktoreducemiscommunication, there are still occurrences of the same concepts being givendifferent names. For example DoubleClick have a “Floodlight tag”, Atlas have a“UniversalActionTag”,Mediamindhasa“Container tag” (allare thesame thing).Themethodology that underpinswhat ismeant by a “click” or an “impression” still causesheadaches industry wide despite clear guidelines being set down. In online tracking italwayscomesbacktomethodology.

Inmyfirstjobintheindustrymylinemanagersatmedowntoreadthroughthoroughlythemethodologiesofthebigresearchcompanies;NeilsenandComScore.Theserespectedplatformswere revealed tome as building their datasets on a sample of people from apanel.Thismeantthatthereportedfiguresinthesesystemswerenotaccuratewhenlownumbers were listed as they represented only one or two people and were not anabsolutelyaccuraterepresentationwhenscaledup(despite thegrowthof thesepanels inrecentyears,thereisstillsometruthtothesestatements).Collectinginformationbasedonapanelofusersasopposedtomeasuringthebehaviourofallusershelpstodemonstratethe importance of methodology; two datasets could differ wildly if one comes from asampleandtheothercomesfromafullaudiencemeasurement.Themethodologyprovidesanindispensablecontext.

The lesson is that the devil is in the detail and technology companies are not alwayswholly transparent in theirmethodologies so it is vital to askquestions about anydata.Sometimes two systems measuring the same things at the same time will not reportexactlythesamefigures.TheIABhasdoneitsbesttoclosethegap(suchasallowingforamaximumofatenpercentdiscrepancybetweentechnologiesmeasuringthesamething)butatthecorethesetechnologiesaredesigneddifferentlyandmightusedifferentmethodsto “ping” servers at slightly different times, resulting in mismatched figures. Thisuniversal acknowledgment (like the unstable world of measurement in quantummechanics)doesnotmean thatall systemsare flawed,but thatusingasinglesystemtotrack all advertising activity reduces confusion and discrepancy because all things aremeasuredunderthesamemethodology.Thisrequirementledtotheadventofthetoolswecallthirdpartyadservers(onecountingtoolusedtomeasureallactivity).Withoutgettingtoocomplex, it canalsobeargued thatusingmultiple tools tomeasure the same thingsand then taking an average is themost accurateway to capture data. In an idealworldmanyAgencieswouldchoosetousemultiplethirdpartyadserversforasinglecampaign

butthecostsoftendonotjustifytherewardsgainedfromamoreaccuratemeasurement.

Nowthatwehavesetthestageandgivenaveryhighlevelviewofwhousesthirdpartyadserversandwhytheyareused,wewillnowmovetolookattheneighbouringtechnologieson theweb page and on the user’smachine. These have an impact on how ad servingworksandwhatthelimitationsareindatacollection.

Chapter1Localvs.RemoteTechnologies

Chapter1-Section1BrowsersandDevices• TheroleofthebrowserinDigitaladvertising

• Figure1.1

• Figure1.2

• Trackingacrossmultiplebrowsers

• firstpartyVs.thirdPartyCookies

• Figure1.3

• CookieDeletion

• DistinctionbetweenDevicesandComputers

• AdblockingTools

• WhyisthereadvertisingontheInternet?

Third party ad servers do notwork isolation.Many technologieswork hand in hand toform websites and web channels that create a user’s internet experience (such webchannels include “Search Marketing” and “Rich Media”). The ways these channelsoperate and grow affect the usability of the internet such as throughweb page loadingspeedandthepaceofinnovationsinwebtechnologydevelopment.Thissectionlooksatthetoolsandtechnologiesthatwecomeintocontacteverydayasinternetusers,(whetherwe are aware of them or not) and looks and how they each relate to the world ofAdvertising and tracking. The bottom-up approach to considering the capabilities ofunderlying tracking technologies (learning how core code like HTML and JavaScriptrelatetothecapabilitiesofthebrowser)willprovideasolidframeworkofunderstandingwhichcanbeappliedtocomplexproblemsolvinginadcampaignsetupandoptimization(e.g.ifyouunderstandhowJavaScriptworksyoucanunderstandhowtouseittoadaptadtechnology to fit unique problems and situations which will arise every day). Perhapsmoreimportantly,thisintroductionprovidesastablegroundingfromwhichtounderstandmoreaboutthewebchannelsthemselves.

Despite being such a ubiquitous technology the concept of the “Internet” is commonlymisunderstoodasjustmeaningtheWorldWideWeb(thebitofinternetthatloadsintothebrowser window). The Internet is in fact a far bigger entity and “the internet” in thecontext of online advertisingmeans; the visualweb, its applications and access points.Where informationcanbe sharedacross theworld automatically, this is the internetwespeak of. It is important to be clear on this so that all industry stakeholders have acommon understanding. When work with new clients, Advertisers and Publishers, thesyntaxanddefinitionofsuchprinciplesisbestbeddedinearlyonbearinginmindthatthemediaindustryattractsfolkfromallwalksoflife.

Soasnot togetourheadstoofar intothecloud(pardonthepun)let’s just thinksimplyaboutthebrowsertostartoffwith:

TheroleofthebrowserinDigitaladvertising

DespiteitsparamountimportancetotheworldofDigitalAdvertising,thebrowser-owningtechnology companies almost never engage in an active dialoguewith any of the otherDigital Stakeholders (Publishers, Advertisers and Agencies).The work of updating abrowserandreleasinganewversion(howeverwild the impactmaybeon thedesignofWebsites and Ads) has virtually no consideration or input from the vast Advertisingcommunitythatitsupports.Thisstatementisthereforetrue:

Browserswerenotbuilttosupportthecreativerangeofdesignsavailableforonlineadsandwebsites.Trulycreativewebsitesandadshavebeenpredominantlybornoutoftheskillandvisionofartistsandcreativedesigners,whotakethefoundationofanewbrowser releaseand learnhow tobuild thebestuserexperienceon topof it.Thebrowserreleasewillclearlyprovidenewavenuestotakecodeanddesignbutthebrowserdoesnotarrivewithaflurryofsupportedadandwebsitedesigns.

Thekeytakeawayhereisnottoeverexpectsupportforanycampaignfromthebrowserthemselves no matter how large the campaign may be. In my experience finding atechnicalcontactworkingbrowser-side(evenifyouworksomewherelikeMicrosoftandyouareabsolutelydesperatetoseeifyourcampaigncanutilisesomekindofexception)isagoosechasenotworthtakingupandit’ssafetoassumeyou’llneverbeabletopointoutthe Mozilla team in a crowd and buy these guys a drink. The browser remains anunknowableforceworkingbyitsownrules,whichthewholeindustryhaslearnedtoadapttoovertime.

Figure 1.1: Users access the web through an internet enabled device (Computer orMobile)andthennavigatewebpagesfromwithinthebrowser.

Useropinionisdividedonthebestbrowserstousetosurfthewebandthebattleforthetopbrowserissettocontinueformanyyearstocome.Firefox,InternetExplorer,ChromeandSafaridominatethebrowserspace.Butwhatrelevancedoesthebrowserhavetothirdpartyadserving?(Anddoesitmakeanydifferencewhichbrowsertheuseruses?).

ThebrowserisveryimportanttoonlineAdvertisingbecausetheuserexperiencewiththevastmajorityoftheadvertsontheinternetishadinsidethebrowserenvironment.Ideallytheuserwouldbesittingatthesamecomputerwhileusingtheweballofthetime,astheysurf thewebandencounter relevant, innovative ads. In fact to store and track themostaccuratedatainthethirdpartyadserver,asingleuserwouldneedtobrowsethewebonasingledevice(usingthesamebrowserallthetime).Inthissinglebrowsertheuserwouldrarely delete their browser cookies and this browser would have acted as the singleconduitfor thefullrecordof thatuser’sfullywebjourney,engagingwithadsandusingtheinternet.Theuser’swebactivitiesareoccurringinjustoneplace(onebrowserandonedevice).

We are now going to start referencing “cookies”. In simplified terms, a cookie can bethoughtofasadocument.OnthisdocumentiswrittenauniquenumberwhichidentifiestheusercalledacookieID.Now,ifeachuserhadonlyonecookieIDandthecookiethatitbelongedtowasneverdeleted,everyuserwouldhavejustoneuniquenumber.IfyouweretocountallthetotaluniquecookieIDsinexistence,youwouldhaveanaccuratereadingofthetotalnumberofusersontheInternet.Asyoumayhavegathered,thingsarenotquiteassimpleasthis.

First if all, cookiesarenot storedonacomputerordevice ina foldercalled“cookies”.Eachbrowserhasitsown“cookie-vault”whichdoesnotletotherbrowsersin.Soifonedidwanttocompareandcontrastusersbetweenbrowsers,theprocessrequiredtodoitissomewhatcomplex.

Figure1.2:Eachbrowserhasthefacilitytostorecookies.AlthoughacookieismeanttorepresentasingleUser, theuseofmultiplebrowserscreatesmultiplecookies-aknownindustryproblem.

Trackingacrossmultiplebrowsers

Thesecondandperhapsmoreobviousproblemisthatusersveryrarelyuseasingleaccesspointorsinglebrowsertousetheinternet.Insteaduserscomeonfromdifferentdevicesandbrowsers,sometimesusingmorethanonedeviceatthesametime.Thisisapainforonline advertising for the simple reason that multiple cookies look like multipleindividualsinthedatathatcomesoutoftheadserverafterithascompletedtrackingandprocessingitsgargantuandatasets.

Users who usemultiple browsers frequently are harder to track as activity is recordedagainsttwoseparatecookieID’swhicharenevermatchedandtechnicallyspeakingtheadserverseestwoseparateindividuals.

ThisisadangerforthemarketerandtheAgencybecausethetruenumberofusersseeingadverts is skewed. If there was just a single browser in existence then online admeasurementwouldbefarmorereliable.Inthereportingchapterlateroninthisbookwewillseethatadserverreportingstilldoesnotofferasolutiontorootingouttheuniquenessofallcookieswhenfacedwiththeproblemofusersthatusemultiplebrowsers.Bearingthis inmindwhen conducting analysis of the ad serving datamay help in providing aclearerpictureofthetotalnumberofusersinteractingwithadsinthecampaign.Let’slookatcookiesinmoredetail.

Firstpartyvs.thirdPartyCookies

There are two main types of cookie. The “first party cookie” is placed on the user’s

computeror ‘dropped’by thewebsite theuser isvisiting.Bydropping,wemean that aremotemachineorserversendsatinyfiletouser’scomputerordevice,wherebythefileisdelivered and remains until it is deleted or removed. This file is a cookie and usuallycontains thecookieID, the timestampofwhen thecookiewasdroppedand thefilenameforthefilewillreferencethedomainitcamefrom.Ifawebsitedropsacookiefileontoyour computer from its own domain, it is a first party cookie (Theweb address of thewebsite is part of the websites “domain” so in http://autos.msn.com “msn.com” is thedomainand“autos.”isthesubdomain).Ifawebsitedropsacookieonyourmachinefromanotherdomain(suchasthedomainbelongingtoathirdpartyadserver)thisiscalleda“thirdpartycookie”.

Figure 1.3:Where a user uses a single browser, browsing to web pages triggers thedownloading of content and ads. The native identifiers for the user belonging to thecontentandthenadsarefirstandthenthirdpartycookies.

Cookies are a website’s way of remembering some information about you as a userbecauseoncetheyhavebeendroppedtheyremainonthemachineuntiltheyareremoved.Sincethecookieisresidingontheuser’smachineit issaidto‘live’onthe“clientside”and thecookie remainsasa signpostorunique referencepoint, for awebsite to lookupthatuniquenumberagainstitsowndatabase(wheretheremaybemoreinformationabouteachcookiestoredonthedatabasesouritsownmachines;thisiscalledthe“serverside”)

Thislookupormatchingprocessallowsawebsiteto‘remember’thingslikesomesettingsorchanges theusermayhavemade toawebsite.Mostcommonly thefirstpartycookiecanbeused to remember the information theuserput into formsonawebsite, (so thatwhentheyvisitthatsiteagain,theysavetimebynothavingtore-inputthisinformation).This is very useful when you, as a user, have to login to a website each time and itremembersyourusernameforinstance.

Thirdpartycookiessuchasthosebelongingtothethirdpartyadserverhavearecordof

theuser’sactivitymatchedalongsideengagementwithanadcampaign.Theadservercanreportonwhethertheusersawanad,clickedonanad,orwentontobuyaproductafterseeingorclickinganadonline.Tobeclear,itisnottheactualuser’snamethattheserverisremembering,buttheuniquereferencenumberstoredinthatuser’scookie.Theserverstoresnopersonallyidentifiableinformationandsoneverknowsthatyouareyou;itjustknowsthatyourcookieallowstheservertorememberbehaviouralinformationaboutthatcookieID.ThosereaderswholiveandbreatheExcelanddatabasescanbegintopictureamastertablesittinginthedeepbackendoftheadserver,whereeveryadengagementcanbetiedbacktoacookieID(seeChapter5-CookieLevelData)

CookieDeletion

A new consideration if you were just beginning to get comfortable with cookies andcookieID’s:

1. Cookiesdonotliveforever.

2. Themoretimepassesthehigherthelikelihoodofauserdeletingacookie.

CookieDeletionaffectsthirdpartyadservingbecauseeachnewcookieistreatedasanewuser.Users that delete their cookies frequently skew the data available in the ad serversince there is no way to tie a deleted cookie from a user machine to the new cookiedropped by the server to replace it. As we will see later this affects a lot of differentreportsincludingUniqueReachandFrequencyandPerformancereports(seeChapter5).

There is lots of research posted around the internet regarding cookie deletion. Sometechnologiesclaimtobeabletoforecastandpredictcookiedeletionindifferentcounties.Someadserversofferalternativereportingmethodsthatcanaccountforcookiedeletion(SeeChapter5-Reporting-AdjustedUniquereporting).

It is inthebestinterestofAdvertisers,AgenciesandPublisherstopreventcookiesfrombeingdeleted,thebenefitsoftheirusearehotlydebatedintheforumofrightsprotectionorganisations in the field of personal privacy and it is here that theDigitalAdvertisingstakeholdersdefendtheuseofcookieswhereinformationthatisheldaboutusersdoesnotlead to a direct identification of that user. In this legal minefield the debate continuesglobally as to the acceptance and regulation of cookie-use,where the user is given theoption to opt-in or out of the automatic action of cookie dropping (see Chapter 7 -Retargeting).

DistinctionbetweenDevicesandComputers

Thewebbrowseronauser’sphoneortabletoffersadifferentexperiencethantheoneonadesktop computer. Examples include the screen size and mouse-free navigation.Considerationsshouldbemadeforthesedifferenceswhenplanninganadcampaign,fromthebuildingoftheadstothedesignofthewebsitethattheusershouldenduponwhenclicking on the ad. In some cases, Advertisers will pay a Creative Agency to producedifferenttypesofadsanddesigndifferentwebsitesforusersusingdifferentbrowsers.

Thisvarietyincreasesthecostofrunninganadcampaign,andinsomeinstancesisusedtopreservethesameuserexperienceacrossdevicesandinothersoffersacreativediversity.Adservingtechnologyallowsforbrowser-specifictargeting(seeChapter2-Publishers-PublisherAdServers)toenabledifferentadstobedisplayedondifferentdevices.

Technicallyspeaking“mobile”browsersarejustdifferentversionsofthesamepcandmacbasedbrowsersthatuserstheworldoverarenowusedtousing.Aswithdesktopdevicebrowsers, the differences between the versions can affect the way that ads and pagesappearorarerendered.Thispresentsarealproblemforconsistentaddelivery,especiallyformorecomplex formats likeRichMedia (fordefinitionseeChapter3 -RichMedia).TheAdvertiserorAgencymayplanwiththeintentofofferingthesameuserexperiencetoall users across all devices but to do someans testing ads on all such devices before acampaignbegins.ThesameistrueintheworldofPublishingwhereallwebpagesmustbecheckedinallnewdevicesandbrowserstoo.SomeAgencieshaveoptedtofillvastroomswith all of the available consumer internet-enabled devices for the purposes of suchtesting. Others will try to emulate the multitude of available browsers using softwarefoundonline.

HTML5mayfinallybealightattheendofalongawaitedtunnel,asitisaleapforwardinstandardisation for both page and ad rendering. HTML5 has been devised to try anderadicate the need for device by device testing, which has extended the testing timerequiredbeforeacampaignlaunches,creatingindustry-wideinefficiency.

HTML5 isoneof themost important things tohappen toadvertising in recentmemorysincethetimelinesofanadcampaignwillaffectallparties;andtheimprovementintheflexibility in the codingof anadcanmean thedifferencebetweena frustrating internetexperienceandafully-fledgedengagementbetweenuserandad.HTML5isnotreallyacomponent of the browser itself but amark-up language, somethingwewill look at inmore detail in the next section. Beyond HTML support, browsers bring plug-intechnologies to consideration and these are controlled by the user.One plug-in type inparticularhashistoricallycausedastirinthebirthandgrowthofOnlineAdvertisingandtheseareAdblockingplugins.

Adblockingtoolsinthebrowser

Someusers installsoftwareorbrowserextensions(alsocalled“plugins”) to“block”adsfromappearingontheirmachine.ThesetoolsdisruptthedeliveryoftheimageoftheadorinfactanyinformationfromthePublisher-sideadserver(SeeChapter2-Publishers) tothe user’s browser. Ad blocking was more functional in the days of dial-up whenbandwidthwaspreciousandloadinganadmeantanextratenminutesofpageloadtime.Todaybandwidthissoquickthatadblockingwouldserveverylittlepurposeinspeedinguppageloadtimes(adsusuallymakeupatinypercentageofthe“weight”ofthewholepageasthefilesaredownloadedfromtheserver).Insteadtheseadblockingtoolsareusedtoprovideanad-freeinternetexperience.AllstakeholdersoperatinginOnlineAdvertisingneedtobeabletorespondtoquestionsaboutAdBlockingbutsincethemechanismissosimple it is the wider impact of these tools that requires focus as an Ad-Free internetexperience is a threat to the whole medium of the web. This in turn concludes oursummaryoftherelevantfeaturesofthebrowserinOnlineAdvertising.

WhyisthereadvertisingontheInternet?

With such far reaching potential, the web comes with its own politics and the politicsframe the development and demands for further supporting technologies. The web ispredominantlyused,by itsusers to sourceanddigestwrittenand (more recently)video

content.Contenttakestimetoproduce.Ioncesetaboutbuildingacontentheavywebsite:ittookhours,daysandweeksoftimetocompile.Apartfromlearningobtainingnewskillsand having the warming experience of contributing to a giant database of usefulinformation about theworld, Iwas only compelled to continue adding such content byreceivingmoneyinreturn.

The twomostcompellingoptions forawebsiteowner toproducequalitycontentare tochargeasubscription fee for theservice (plaguing theuserwithacostoutof theirownwallet)or to seem tooffer the content for free to theuser, at the cost to theAdvertiser(whoarereallyjustpayingfortheattentionspanofa“targeted”user).

Adblockingisnotanundetectableweapononconsumercombat;thefailedattemptbytheservertodeliveradsisdetectedeachtimeitoccurs.Asthisfailurerategrows,thereisaninverse decline in the size of the audience for the website/Publisher to offer to theirAdvertiser base. A decline in the audience size means a decline in revenue, whichultimatelyleadstoadeclineinboththeamount(andmoreimportantly)thequalityofthecontentbecauselessmoneyisavailabletohelpproduceit.

Contentiskingonline.Thebestcontentdrawsthemostattentionandalthoughthegrowthofusergeneratedcontent(UGC)hasbeenrapidandfarreaching;thequalityuser“traffic”remains with big Publishers where it always has done (and more niche “long tail”audiences hop to more obscure websites with lower levers of traffic). ThankfullytechnologieslikeAdvertisingExchangesaremakingsomegroundinbalancingthebooks,butthisisatopicforChapter8.

Let’s move on from the browser to the content of the web; the websites and the adsthemselvesandthebasetechnologiesonwhichtheyareconstructed.

Chapter1-Section2TheAnatomyofSitesandAds

• Whattheusersees

Figure1.4

• HTML

• Figure1.5

• JavaScript

• Images

• TextLinks

• Figure1.6

• Flash

• TheadventofHTML5

Whattheusersees

Takeamagnifyingglassandpeer into thesourcecodeofanywebpage(or justuse thebrowserbuttonstoloadupthepagesource)andinsideyouwillfindaratherdifficult toreadsetof instructionsandcode.Theeducatedeyecanpickoutcommands in thecodethathelpidentifyhowthewebpagehasbeenconstructed.Thisincludesinformationaboutwhere the ads come from as well. Both HTML and JavaScript are the most commonlanguagestofindinside thissourcecode.Theyloadthecontent in thebrowserwindow,renderingasawebsiteconstructedbyasetofcommonelementsorentities.Text,Imagesand Flash make up the vast majority of the web today (soon to be overtaken by anexplosioninvideo).Withoutthesecomponentsthewebcouldnotsupportadvertisingasitexists today. Understandingmore about them personifies their importance, highlightingtechnicalencountersonallsidesoftheAdvertisingecosystem.

Figure 1.4: Navigating to the Web Page via the main browser, the user encountersContentandAds.Thesearemadeupofwhatcanbeclassedasentities:technologiesandcodethatformtheveryfoundationsofthefamiliarweb.

HTML

“HTML”(Hyper-TextMark-upLanguage) is thecore‘code’of thewebbasedbrowsingexperience.It literallyis thecodethatdescribestothebrowserwhatthecontentofeachweb page should be and the browser renders this code to give the effect of a page ofinformation.HTMListhelanguageof“tags”andwhenwereferto“tags”inthirdpartyadserving,we are talking about a few lines ofHTML code. This code can be embeddeddirectlyontoawebpagebyawebdevelopmentteamworkingonthewebsiteandwhenapageisloadedupinabrowserbyauserthecode‘executes’,andanadormorecontentisdisplayed.(ToseeanHTMLadtag,skiptoChapter3-DisplayAdservingOperations-AdTags).

HTMLishowalmostalloftheadsservedtodayarerequestedfromtheadserver.IftheenvironmentinwhichanadneedstobeserveddoesnotsupportHTML,thenothercodecanbeusedtomakethatrequest(suchasinacomputergamingenvironment).Infactinmost instances, the HTML is used to house a more complex request which occurs inJavaScript.JavaScriptdoesthefetching,carryingandcalculatingofthecontent.HTMLissimplyawayofcarryingtherequiredadcodethatthePublishercanquiteeasilyslipontotheirwebsites.

Figure1.5:BehindthescenesastandardwebpageismadeupofaseriesofHTMLtags.AnAdtagisnodifferent,whichcansitonanywebpagetodisplayanadtotheuser.Inthisinstance(bottomright)theHTMLtagcontainsareferencetoJavaScript.

JavaScript

JavaScriptiscalledaclient-sidelanguage.Itrunsor“executes”intheuser’sbrowserandallowsforawholehostofimportantwebpagetrickssuchaspopupwindowsandbrowsermanipulation(likeacommandtoprintsomethingonthepage).JavaScriptcangatherandsendauser’sinformationandtoaremoteserversothattheserveriscapableofrespondingbysendingoutanewpieceof information (likemakingapreviouslyhiddenpartof thewebsite,suddenlyappear).

JavaScriptisthemostcommonmethodofaddeliverythatthirdpartyadserversuse.TheflexibilityoftheJavaScriptcodeallowstheadtobepulleddirectlyintoaPublisher’swebpagefromthe thirdpartyadserverandmakedecisionsaboutwhatad todisplay inrealtime.JavaScriptcandointerestingthingslikeworkoutwhattimeit is,orfindoutwhatbrowsertheuserisusing:informationlikethiscanbepassedbackintotheadserverwhenthe ad loads allowing for some impressive customisation inside the ad itself. It is notcommon for an Advertiser to want to get this technical when running a campaign buteverynowandthenJavaScriptdoescometotherescuewhenanadcampaignjustwantstobealittleclevererormoreinnovative.

To understand how JavaScript is essential in the technical process of ad request anddeliverywewilllookinmoredetailattherelationshipbetweenAdServingandJavaScriptinChapter1 -BasicsonServerhardware.For the timebeing it is sufficient to say thatJavaScriptreturnsoneoffourremoteentitiesfromtheadserverandpullsthemintothewebpageasthepageloads.ThesefourentitiesareImagefiles,Flashfiles,textlinksandvideo files. Video files are a little more complex to explain so this is explained afterproviding a more thorough technical introduction - details for video can be found inChapter3-InstreamVideo).

Images

Imagestransformedtheearlywebfromacompletelytext-basedenvironmenttosomething

thatcouldbedigestedasanentertainmentformbythemassesandadvertisingwasoneofthe first places to utilise the ability of early browsers to load images. TheGIF type ofimagewasmorefavouredby theadvertisingspacebecauseGIFscancontainmore thanone image at a time and so a very rudimentary formof frameby frame animationwaspossible.ThetroublewithGIFsisthattheyareveryrestrictingintermsofcreativity.TheyarenotinteractiveandimageslookpoornexttoFlash.Today,GIFsarestillusedinonlineadvertisingasa“fall-backoption”incasetheuser’sbrowsercannotsupportFlash(afterall, a grainy image of a company logo is better than completely wasted and white adspace).The traditional*.gif fileextension isalsono longermandatoryandbetter imagecompressionmethodssuchasinthe*.jpgformator*.jpegformatsarecommonlyutilised.The legacy of the ad as an image remainswith ad serving technology todaywhere allstandarddisplayadsarecomprisedofan imageportionof theHTMLcodeaswellasaclickor text linkportion.The two sectionswhichmakeup thisHTMLad tagallowalldisplay ads to report back to the ad server; counting the loading of the image as an“impression”metricandtheuseractionofthe“click”asaclickmetric(moredetailonthisinChapter5-Reporting).

Text-Links

SinceHTMLcodeallowsanypieceoftext,objectorimagetobeclickablebytheuserviatheuseofthe“Anchortag”,asimpletextlinkcanbeissuedbytheAdServerasthemostbasic type of Ad. These text-link ads are extremely common all over theweb and aremostly utilised by ads that exist in Search engines. By issuing an image and text linktogether,theimageisclickablebutthetwopartsofthecodecanbeseparatedoutsothattheimagecansimplybeprogrammedtoappearwithouttheabilitytoclick,orthetextlinkcanbeimplementedwithoutanyimagepresent.

Figure1.6:Asampleofaverybasicnon-JavaScriptadtag.Thenumberedrowsareforreferencepurposes:Row2describestheopeningoftheanchortag:theURLdescribedistheclickthroughURL.Thecodeinrows3and4 issurroundedby theanchor tagwhichactsastheclickredirectportionoftheadtag.Row4cancontaina1x1otherwiseknownasatrackingpixel.

Thetextinfigure1.6(row3)isclickableinsidethewebpagethattheusercansee.Thisisthe“text-link”.Text-linksarethemostbasicbuildingblockofanonlinead;theiruseandbehaviouraredescribedinaseriesofwidelyrecognisedphrases.Asmanyofthesesoundsimilaritisimportanttoclarifythiswithparticipatingpartieswhensettingupandrunninganadcampaign.

• The action of a user clicking and being directed to an Advertiser site is called a“clickthrough“.

• Theprocessofausercompletingtheclickthroughsuccessfullyiscalleda“click-to-land“.

• Thedestinationaddressiscalleda“clickthroughURL”

• TheportionoftheadtagwhichdeclarestheclickthroughURLaspartofanyadortextlinkiscalleda“clickredirect“.

• Aclickredirectmayalsobereferredtoasa“clicktracker”

• Theequivalentformeasuringthe loadingof thead(evenasa text link) iscalleda“trackingpixel“.(CoveredoffinmoredetailinChapter3-DisplayAdserving).

Sofarwehavedescribedthetwoelementsofanadtagasbeingtheimageportionandtheclickportion.Howeveritisnotcommonfora“staticimage”(i.e.astillpicture)tobeusedasthevisualad;rathertheimagedisplaysasaFlashanimation.

Flash

Having a familiaritywith Flash is essential toworking in theDisplay andRichMediachannels inDigital Advertising. “Flash” are the animated and interactive panels dottedaround the internet that can house video, get bigger, or show multiple images in animpactfuldelivery.

Flash is nascent to theweb browsing experience and Flash files or ads themselves arecreated inAdobe’s Flash authoring environment (more on this in Chapter 2 - CreativeAgencies). Flash is so widely used and adopted across the web because of its ease tocreate animations and its flexibility to be creative with the code and the developmentenvironment.InChapter2wewilllookinmuchmoregranulardetailtounderstandmoreabout Flash and its capabilities. Flash is so popular that it seems that it will be in useacrosstheinternetforeverbutnewtechnologiesareemergingthatwillreplaceit.HTML5isthefavourite.

TheAdventofHTML5

HTML5 is created to amalgamate all the capabilities of both Flash and HTML in anexperiencewhichprovidesaconsistentuserexperienceontheweb.FlashwaspopularforsolongbecauseofasimilarconsistentusabilitybuttheflawinFlashistheneedforusersto have installed the Flash player plugin.Withmore devices and browsers available tousers, supporting the plugin has become difficult forAdobe and since using the Flashauthoringtoolcomesatacosttothecreativedeveloper,butthedevelopmentofHTML5isfree,HTMLwinsasthestandardforthefuture.WithHTML5thereisnoneedforFlash,asvideoandanimationbecomepartofHTML.ThevectoranimationsandfunctionsthatFlash built fame for facilitating are now supported in HTML5 and the community ofsupporters for thisnewstandardare strongerwhere theupdatesandmaintenanceof theauthoring environment of HTML5 will become increasingly faster and stronger thanFlash.

Sofarwecoveredoffinternettechnologiestogivecontexttoonlineadvertising.Thenextsectionmovesfocusfromthelocalusercomputertotheservers.

Chapter1-Section3AdServing• Whatdoestheadserverdo?

o Figure1.7

o TableofDefinitions

o Figure1.8

• TheroleoftheCDN

• WaitingaroundforAdDelivery-theuserexperience

o Pre-addisplay

o Duringaddisplay

o Onadclose

This sectionwill touch lightly on the relationship between the user’s computer and thetechnologies sending packets of information to that user’s machine. This is useful toexplainwhat is reallygoingonbeyond the familiarglareofyourownPCandbrowser.ThewordsandphrasescontainedinthissectionarecommonlythrownaroundintechnicalconversationswhereanAdvertisingtechnologyisinvolved.

Whatdoestheadserverdo?

Many other available texts describe how user’s machines and servers exchangeinformationtoloadthevariouspagesoftheweb.Thefocusforthissectionistoexplaintheexchangesthatoccurbetweenmachinestodistribute,loadandcountadsrightacrosstheweb.AsaremindertheThirdPartyAdServerhastwocorefunctions:

1. Thefunctiontoserveordelivertheadstotheusersmachine

2. Tocountandtrackthedeliveryandperformanceoftheseads

Figure 1.7:Data in and out on-site at the ad server. This highly generalised diagramshowshow the user request into the ad server gateway gets split for the two functions.First theadservercountsusingacountingmoduleand inparallelmakesadecision toservethecorrectassetbacktotheuser.

Figure1.7showsthefastfunctionoftheadservertocountanddeliveranadbuttheboldarrowsontheleftonlyshowdatacomingintoandgoingoutoftheadserver.Theinternetinfrastructurethatwillcarrythisdatatothispointundergoesalongerjourney.Beforewe

undertakethisjourneyweneedtomakesureweareusingtherightlanguagetodescribethevariousportsofcallandprocessesthatthedataundergoes.

DefinitionLocalMachine Theuser’scomputerordevicetoaccessthewebRemoteMachine Theserverthattheuser’scomputerisexchangingdatawith

HostServer Anothernameforremotemachine,theserveris“hosting”theusersrequestforcontentorads

Packet

Dataistransmittedbetweenmachinesinpackets.Utilizingthebest server route between the local and remote machine.Sometimespacketsget lostandthiscanhaveaminorimpactonthecountingofadsserved.

Router Toconnect to theISP theuser’smachinemustutilizea localswitchorrouter.Again,alldatacomesthroughhere.

ISP InternetServiceProvider; theuser’s accesspoint to theweb,alldatapacketscomeandgothroughtheISP

CDN

ContentDeliveryNetwork;alargenetworkofgeographicallylocated content servers allowing users to download contentfrom a regional source. Ad servers themselves do notdistribute replicas of the files all over theworld. TheyworkwithCDNstoachievethis.

DataCenter

DataCentersare thephysicalbuildings thathouseadserversand servers of theCDN.As the user’s browsermakes a calldirectly to the ad server to request an ad, the ad servermustdelivertheinformationbackandcounttherequestassoonaspossible.Havingstrategicallylocatedbuildingsinglobalhubsandhighlypopulatedcitiesensuresafasterdelivery.

ServerFarm Anon-technicaltermtodescribeadatacenterorgroupofdatacenters

AdRequestTheinitialloadingoftheadtagonaPublisherpagesendsanHTTP request for to display the correct ad to the ad server.ThisiscalledanAdRequest.

JavaScriptResponse Theadserverrespondstotheadrequestwitharesponse;thisisresponsepredominantlyoccursinJavaScript.

ImpressionTrackerAlladscontainImpressionTrackercodewhichsendsdataasan“adrequest”tothecountingmoduleor“adcounter”attheadserver

Impression WhenanadcounterreceivesandrespondstoanHTTPrequestforatrackingassetfromaclient(IAB.netGuidelines)

ClickTracker Whentheuserinitiatesaclick,thiscoderoutesdatatothe“adcounter”attheadservertocounta“click”

AdCall AnothernameforAdRequestAdPlay Anindustrywidedisputedmeaning,notbackedbytheIAB.

AdClose Whena JavaScript servedad is closedor thebrowser leavesthepage,theAdcanmakeafinalrequesttotheserver

Beforewelookatthebiggerpictureit isimportanttonotethatthetechnicaldeliveryoftheadisnotthesameasanindustryagreedcountoftheimpressionmetric(whichcanbefoundhere:http://www.iab.net/media/file/Global_meas_guidelines.pdf).Howeverasthisdefinitionhascertifiedindustryadoption,itistakenastothetruemeasureofdeliveredadsacrossalladcountingsystems.

Figure1.8:Userrequestschannelupto theadserver’sdatacenterwherecountingandadservingdecisionsaremade.Asmultiplerequestsarebeingmadebymultipleusersforthe sameassets, thead serverdecisionmoduledoesnotneed todistribute the sameadeverytimebutratherletstheCDNdistributethead.

TheroleoftheCDN

TheContentDeliveryNetworktracesalltheflightpathsintheworld,joiningupthedotsbetweenthecitiesandtheISPs(InternetServiceProviders)aroundtheworldandroutingreplicatedinformationontousers’deviceswhere-ever theymaybe.Adsstoredin theadserverneedtobereplicatedintheadserverdatacentres,preparedforthemoststrategicdeliveryoftheassets(adelaycalled“adreplication”)andaprocessthatoccurswhenthecampaign is being “trafficked” (more on trafficking in Chapter 3).When an ad call isreceivedtheadservertransmitstheadstotheCDN’s.Akamai,describedbysomeastheinternet’s backbone, is one of the largest CDN’s and ad servers utilise its servicesconsistently.LargerPublisherslikeMSN,relyonMicrosoft’sowntechnologyratherthanoutsourcetosystemslikeAkamai.TheMSNCDNhasthecapacitytoallowotherinternettechnologiesrequiringCDNdistributiontoutiliseitforsuchneeds.

Figure1.9:CDNcontentdeliveryismostfrequentlyrepresentedincommonconversationbyreferenceto“Thecloud”.Byofferingserversinlocalproximitytouserdevicesinternetaccess speedsaregreatly increasedand files can easily be replicatedacross theworld,thisincludesbothsitecontentandads.

WaitingaroundforadDelivery-theuserexperience

Sowhatistheeffectontheuser?Isitglaringlyobviousthatthemostminisculeofdelaysactuallyhasa shed-loadofprocessesbehind it?For themostpart theanswer isno,butwithsomanyprocesses inplace,once inabluemoon, theuserwillbestuckonapagewhereanadfailstoload.Thepurposeoftheadserversreportingcapabilityistoquantifythe investmentclientsmake inonlineadvertising, includingdemonstratingconsistentaddelivery and performance which the Agency or Advertiser will reference frequently toensuretheirmessagingisreachingitsaudience,ifnotonlytocheckthatthereisnothingclogging up the pipes (more details in Chapter 5 - Reporting). Providing there are noproblemstheoptimumuserexperiencewilllooklikethis:

Pre-addisplay

1. UserarrivesatPublisherpage

2. AdrequestissenttoPublisheradserver,whiletherequestforthecontentgoestothePublisherCMS

3. ThePublishercontentloads

4. ThePublisheradservercallsthethirdpartyadserver

5. Theadrequestarrivesatthelocaldatacenterdependingontheuserslocation

6. Thecountingmodulecountsanimpression

7. Thedecisionmoduleselectsthemostsuitablead

8. TheadserversendsaJavaScripttoinitiatetheloadingofthead

9. The JavaScript loads on the Publisher page and loads in makes calls to all thenecessaryassets

10. TheassetsarepulledinfromtheCDNwhichinturnpullsindatafromtheadservercreativedatabase

11. AnImageorFlashexecutionappearsintheuser’sbrowserwindow

Duringaddisplay

12. Theadservercountsan“AdPlay”-whichisanInstreamVideometric(SeeChapter3-Instream)ordeliveredimpressionmetric

13. Piggybackedthirdpartyimpressiontrackersload(SeeChapter3-StandardDisplay)

14. Viewabilitybeginscounting(SeeChapter5-AdVerification)

15. RichMediaconstantlyreports“interactions”tothecountingmodule(SeeChapter3-RichMedia).

Onadclose

16. Viewability ends counting and sends data to the ad server counting module (SeeChapter5-AdVerification)

17. RichMedia interactions that utilise a timer are reported to the ad server countingmodule(SeeChapter3-RichMedia)

Thisexposestheminicosmosofworkthattheadserverisdoing.Itisclearwhytheworkis all done bymachines in farawayplaces.The next section turns attention back to theonlineadvertisingindustrytolookatwhotheplayersinthisspace,andhowmanypeopleand hands it takes to launch an online advertising campaign andmake themost of adservingtechnology.

Chapter2TheIndustryPlayers

Chapter2TheIndustryPlayers

• Theagencyholdinggroups

• WhoarethebigAgencies?

• RunningaMediaAgency

• CreativeAgencies

• Figure2.1

• Publishers

Toframethecontextofadvertisingtechnologythebooksincludesinformationaboutthevariousindustryplayers.Giventhenatureof theadvertisingindustry toshiftclientsandaccounts,thissectionmaywellfalloutofdatequitequickly;howeverthisisanattempttorepresentitatahighlevel.

Theagencyholdinggroups

The largest and most powerful advertising Agencies sit under holding groups whichthemselves are home to a multiverse of Creative Agencies, PR Agencies and MediaAgencies.WithsomanyAgenciesunderasingleumbrella,carefulprecautionsaretakentoensurethatAgenciescantakeontheworkandbusinessofmanycompetingAdvertiserswithout allowing sensitive information to spread to those competitors or separateddivisions. Since a holding group may house ten or more Media Agencies whichthemselves may house ten or more Advertisers and somewhere down the line twoAdvertisersareboundtofallintothesamevertical.InsomeinstancesMediaAgenciesandholdinggroupshavebeenknowntobuildoutwholenewAgenciesjusttoensurethatsuchAdvertiserdataiskeptveryseparate.Thepoweroftheholdinggrouptendstofallwiththepriceitcangetforpublicationadspace,theeconomiesofscaleofsuchenterprisesdrivesdownthepriceofadspace,withthepromiseofbusinessonanenormousscale.

WhoarethebigAgencies?

ThereareplentyofUp-to-datesourcesonline thatcangiveamore recentand thoroughcoverageofthebreakdownoftheAgencygroups.Asaveryhighleveloverview:

• Publicis Omnicom group houses the media Agencies formerly of theOmnicom group: OMD, OMG, PHD andManning Gollieb and the mediaAgencies formerly of the Publicis group such as Starcom Mediavest andZenithOptimedia.

• Dentsu/Aegis brings to the table Agencies such as Iprospect, Carat andVizeum.

• WPP offers the industry GroupM and its heavyweight’s Mindshare,Mediacom,MaxusandMECtonamebutafew.

• Interpublicgroup(commonlyshortenedtoIPG)includesUniversalMcCannandInitiative

• HavasgrouphousesMPG/MediaContactsandHavas

RunningaMediaAgency

MediaAgenciesaretypicallymadeupofaccountteams(anaccountbeinganAdvertiser).Each teamwillbemadeupofplanners,buyersandaccountmanagers.MediaAgencieshave supplier contracts with technology providers including third party technologyproviders.Theseagreementsvaryconsiderablyandcandifferbetweenregionandcountry.MediaAgencies also have contractual agreementswith Publishers that outline paymentterms.Suchtermsareusuallyconfirmedinwritingbeforeeachadvertisingcampaignviathe Insertion Order (IO) a document which also clarifies the amount of media spacepurchasedaswellasconditionsofthecampaign.LargerMediaAgenciesareoccasionallyaffiliated with creative Agencies or more often have their own in-house creativedevelopersanddesigners.MuchmorecommonisacontractualrelationshipbetweentheAdvertiser and the Creative Agency directly. Media Agencies usually work withAdvertisers in a long termcapacity bypitching for their business by responding to andwinning the chance to respond to a brief issued by the Advertiser. From an onlineperspectiveMediaAgenciestakeabigburdenoffoftheshouldersoftheAdvertiser.TheworkrequiredtoplanandbuymediatoreachtheAdvertiser’sstatedcampaignobjectivesrequiresmanyhandsandlargeradvertiserswillrarelyfillouttheirmarketingteamswiththe headcount to undertake these tasks. It is for this reason that third party ad servingtechnology ismostutilisedbyMediaAgencies,who report their successes to thehiringAdvertiser.

Creativeagencies

Thereareanenormousnumberofcreativeagenciesandthelandscapeisverycompetitive.Acreativeagencywillusuallyproduceamock-upofhowacampaigncouldlookwithlotsof creative ideas,when pitching to anAdvertiser to build a relationshipwith them andattaintheirbusiness.AdvertisersusuallyhavepreferredCreativeAgenciesthat theywillutilise for building creative across variousmedia channels includingTVandPrint.ThemajorityofCreativeAgencies are small companies employing just a fewdesigners anddevelopersbutsomehavereachedenterprisesizesandfallunderthesameholdinggroupsastheMediaAgencies.

Figure 2.1: A single Agency holding group may have clients operating in the samevertical but the teams are kept separate, often in completely different buildings. TheAdvertiserthemselveshasthefreedomtoselectaCreativeAgencywhichmightworkinacompletelydifferentholdinggrouporoperateindependently.

Publishers

Publishers are the content providers. They are web sites that attract a large number ofusers.Togiveupyourdayjobandjustliveoffoftheadvertisingrevenuefromawebsite

your site would need to attract in excess of one hundred thousand “unique users” permonth (dependingon thequalityof theseusers: alsocalled“thequalityof the traffic”).ThismeansthatiftheusersarecomingtothePublisherwebsiteforapurpose,andtheadsshownarerelevantandwell-targetedtothoseusers,thenthequalityofthoseusersismorevaluable.ThisqualitytrafficraisesthepriceoftheadspaceonthePublisherwebsiteandwouldmeanthePublishercanmakeatidyprofitfromthemoneythatAdvertiserswouldbewillingtospendtohaveadsonaPublisherwebsite.BigPublishers(suchasFacebook,Google,MSN andYahoo!) attract a vast numberof users and in turn are almost alwaysincluded on Advertiser “Media Plans”. Smaller Publishers, although not necessarilyattractingbigcrowdshavetheaddedbenefitsofeitheramorenicheaudienceorcheaperadspace.Publishersarethehomeofads;theycontrolthegreatestcosttoanyadcampaign(thecostoftheadspace)andattimesviewalltechnologyintermsoftheirownproperties.

The remainderof this chapterwillbe spent focussingeachof these stakeholdergroups.Eachgroupspendsmoretimeusingadifferentpieceofadtechnologythereforethirdpartyadservingismoreimportanttosomeandlesstoothers.

Chapter2-Section1AdvertisersandMarketers

• WhoistheAdvertiser?

• TheMarketer

• Agencypitch

• RelationshipwiththeMediaAgency

• RelationshipwiththeCreativeAgency

• Figure2.2.1

• Controllingthepagecontent

• ControllingthepagesoftheAdvertiserSite-theCMS

• LandingPage

• BasketPage

• ConfirmationPage

• Figure2.2.2

• CRMPlatform

• OnsiteBehaviouraldata

• Trackingtheusersonsite

• SiteSideAnalytics

• ClicktoLand

• Passingthetokensintheclick-throughURL

• Referral

WhoistheAdvertiser?

Advertisers can be loosely divided into two buckets in the online world: “Brand” and“Direct response”. Brand Advertisers tend to utilise the web to grow awareness andgenerate interest and engagement but rarely have something tangible to “sell” online.Direct Response (DR)Advertisers are usuallymuchmore complex in their advertisingneeds.Theywillhaveane-commercesitetosellgoodsorservicesonlineandoftenneedtoproduceadsandcampaignswithobjectivestotargetusersateachstageinthepurchaseprocess.ThereareofcourseAdvertisersthatcrossintobothdefinitions.Theseareusuallyvery largeDRAdvertiserswhoneeda littleofeverything. In thissectionwewillbeginwithanoverviewofAdvertisers,followedbytheirrelationshipwithMediaAgenciesandthen Creative Agencies. This is followed by an exploration around the predominanttechnologiesanAdvertiserwillbeworkingcloselywith;includingthemakeupoftheire-commerce site and how to integrate such technologies effectively with third party adserving.

TheMarketer

Advertisermarketingteambringstogethertheoverallmarketingstrategymarryinguptheavailabledigitaladvertisingbudgetwiththerightresourcetogetthebestreturn.

ForBrandadvertisers,marketersareconcernedinitiallywithaddeliveryandverification.If the ads have been served they should have been delivered at an optimal “reach andfrequency” with the ultimate aim of hitting the targeting audience with a memorablecreativemessage.InChapter5wewilllookathowtheadserverreportingcandeliverthisinformation.

For Direct response advertisers, marketers are concerned with ROI and want todemonstratelikeforlikeorincrementalsalesfromasetmarketingbudget.Withtherightcombination of delivery techniques, reporting and optimization; digital is a very fertilegroundforDRmarketerswithaneverincreasingbudget,toinvestin.

For themarketer to ensure the advertising budget is best spent, the campaigngoals areclearlylaidoutbeforeacampaignbegins.

AgencyPitch

TochooseanAgencytheAdvertiserreleasesanRFP(RequestforProposal)toashortlistofMediaAgencies.ThisRFPdetails theadvertiser’smarketingobjectivesover thenextfewyearsandaskstheAgenciestocomebackwiththebeststrategicpitchforthework.The process can take quite a long time as the Advertiser is usually looking to sign acontractwithanAgencyoveratermofafewyears;departingfromsuchcontractscanbevery costly to the Advertiser, so the right decision is imperative. Once an Agency issecured,therelationshipcanbegintogrow.

RelationshipwiththeMediaAgency

OnceanAgencyhaswonovertheAdvertisertheAgencyaccountteambeginstheprocessoftransitioningthe“campaignworkflow”process(theoperationalpracticesanAdvertiserisfamiliarusing)toalignwiththepracticesoftheAgencyortheneedsoftheAdvertiser,wheretheAgencyseesfit.ThisstagecaninvolvemovingtheAdvertiserfromonesetoftechnologiestoanotherdependingontheconditionsofthecontractwiththeAgency.Thisprocesscanalsotakeseveralweeksifnotmonthsbut,donecorrectly,setsthestageforabloomingrelationshipinfuturecampaigns.Inthenextsectionwewilllookincloserdetailatcampaignmanagement,aprocesswhichisAgency-led.

AscampaignsareprocessedtheAdvertiserandMediaAgencylearnwhichprocessesandpracticesworkbestforthemandfindwaysofsharingandusingthethirdpartytechnologybetweenthemtosuiteveryone’sneeds.Forinstancesomemarketerspreferamorehands-on approach to the reporting datawhich theMediaAgency is able to grant as long asexpectationsareclearabouthowdecisionsaroundcampaignoptimizationaremade.

RelationshipwiththeCreativeAgency

UnlikeintheworldofTVadvertising,theCreativeAgencytakesabackseattotheroleoftheMediaAgency incampaignsetup (andusually theAdvertiserwillhavea long termexisting partnershipwith either a preferredCreativeAgency or amyriad of them).Themarketerwillusuallytakeakeeninterestinthedevelopmentofanycreativetomaintaina

consistentbrand image (theword“creative” isused todescribe the stateof adigital adbeforeitbecomesanactualadthatcanbeusedbytheadservingtechnology).Thismeanstherewill be a back and forth of communication between theCreativeAgency and theAdvertiserrightupuntiltheadisfullydeveloped,anditismutuallyagreedthatitsvisualelements will not change before the ad is seen by a public audience. Some big BrandAdvertisersaremoredetailorientedthanotherswhenitcomestoadheringtobrandstyleguidelinesandthesign-offprocesscangetextremelylengthy.

Forthisreason,adservingtechnologyisdesignedtoallowtheCreativeAgencytoshareapreviewoftheadswiththeAdvertiserdirectlyandreceivetheirapprovalbeforetheadisprocessed in the ad server for trafficking. The ad preview environment is particularlyusefulforRichMediaandDCO(SeeChapter3).

Withitspartnersonboard(MediaAgencyandCreativeAgency)thestageisalmostsettobegin running an ad campaign. The final piece of the puzzle from the Advertiser’sperspectiveistoensurethatusersarriveatawebsitethatextendstheirinteresttowardsaneventualpurchase.

TechnicalstaffworkingontheAdvertisermarketingteamtendtoplacealltheirfocusandattentionontheAdvertiser’ssitewhichcanbeprettysimple(forbigbrandadvertisersthatdonotoffer thechance topurchaseonline-perhapshaving less than tenpageson theirsite) or vastly complex (such as in the case of a DRAdvertiser like a retail cataloguewebsite-wherethenumberofpagestomanagemaybeinthemillions).

Figure 2.2.1: In Advertiser-world, the Media Agency ensures that users get to theAdvertiser site throughvariouspaid-for channels (trackedacross theboardonlyby thethirdpartyadserver).OncetheuserarrivesontheAdvertisersitethereareseveraltoolsdoingthemeasuring.

ControllingthePagesoftheAdvertiserSite-(theCMS)

There are lots of types of pages on theAdvertiserwebsite but only a few of them are

usefultothemarketer.Pageswhichhelpdemonstratethatamarketingobjectivehasbeenfulfilledneed tobecarefullycontrolled,watchedandmeasured.ACMS tool fulfils thisfirstfunction.

AnAdvertiserwebsitethathassomethingtosellonlineisinvariablygoingtocontainalotofindividualwebpagesandalotofinformation,tobetterinformtheconsumerontheirjourney topurchase.Webpageswhicharebuilt inbasicHTMLcodealoneareseverelylimiting. Thousands of live pages require a solution that will allow the web developerupdating the Advertiser website to make changes in seconds. A Content ManagementSystem (CMS) allows for the creation and update of thousands of pages in seconds byprovidingabackenduserinterfaceforthewebdevelopertochangethesite.TheuserofthewebsitedoesnotseethepresenceofaContentManagementSystembutinsteadseesanillusionof“static”webpages.OneofthecrucialfunctionsoftheCMSisthatitallowsthemarketingteamtoappendthelivepageswithadditionalpiecesofthirdpartytrackingtechnology to monitor page-level “conversion activities”, these are called “ConversionTags” andChapter 4 is dedicated to explaining their purpose and uses in detail. SomemarketersandAgenciesneedfrequentaccesstothepagecodetochangethesetagsbutnotallAdvertisersites(despitethepresenceofaCMS)allowforfrequentchanges.Thiscanand often does directly impact the successful and timely completion of an online adcampaign.

InFigure2.2.1 the threecorepage typeswhichbest reflect the fulfilmentofmarketingobjectives,arestated.Eachofthesepagetypesissignificant tothemarketingeffortandcanbe fully recognisedascampaignyardsticksor“conversionevents” in thirdpartyadserverreporting:

• TheLandingPage(orProductLandingPage)

• TheBasketPage(orpre-confirmationpage)

• TheConfirmationPage(orThank-youpage)

LandingPage

The Landing page is the first page that the user arrives at in the Advertiser’s onlinedomain.Beingthefirstpage,itneedstobedesignedtotheuserexpectations.Thankstothe connected nature of today’s search engines, a user can type in a keyword into thesearchbar,seealinktotheAdvertiser’ssiteandclickthroughtothelandingpage(moreinformationabout theSearchChannel inChapter3).The landingpage should thereforereflecttheverysearchtermthattheuserinputtedintothesearchengineandtheAdvertisershould customise the look and feel of the page to reflect assumptions about the user’sreasonforarrivingonthatpage.

BasketPage

TheuserjourneyonanAdvertiser’secommercewebsiteleadsdown,whatmarketersrefertoasa“funnel”, a lotofusers reacha landingpage,butmanydropouten-route to thefinalpurchasepage.Thisanalogyisoftenusedinonlinemarketingtodescribethechannelstrategyforanonlinecampaign.Theconsumerisattractedintothetopof the“purchasefunnel”withimpactfulbrandadsthatcaptureuserinterest, this isfollowedbycapturinguser intent and finally the user is drawn through the Advertiser site right up to final

purchaseattheendofthefunnel.

The penultimate step is to load up the “shopping basket” with the intended purchase.Althoughthe“shoppingbasket”elementofthewebsiteitselfmayhavemanypages,thecollective reference foroneormoreof thesepages isa“basketpage”.Sinceausercanundertakevariousbehavioursinthebasketpage,suchasdroppingoutoftheprocess,itisimportantfortheAdvertisertomonitorthesepageswithadservertrackingtechnologyandmakesurethattheusersarenotlostforever(seeChapter7).

ConfirmationPage

Theactualsaletakesplacesomewherebetweenthebasketpageandthethankyoupage.The last key Advertiser page type is therefore the “thank you” page (which is also isreferred toas theconfirmationpage);where theAdvertiserconfirms to theuser that theproducthasbeenpurchased.

TrackingactivityontheconfirmationpageiscrucialforDRAdvertisers;aconsumeronlybecomesaconsumeroncetheyarrivehere,sotheveryarrivalontothispage(whichisnotaccessiblebyanyothermeansotherthanbybuyingsomething)isloggedintheCRMandPOS(asshowninFigure2.2.1).ThesesystemslogauniqueidentificationnumberrelatedtothetransactioncalledanOrderID.ItisthissamenumberwhichtiestogethertheCRM,POSand ad servingdata sets as they can all be setup to track the samevaluewhen anorderiscompleted.Theadservingdataexposingthejourneytheuserwentontoarriveonthe landing page and complete the sale - allowing the marketer to see exactly whichcustomersrespondtowhichadsandhowmuchtheyspent.

Figure 2.2.2: The Advertiser can have the web development team working on theAdvertiser site setup the Confirmation Page to pass the value of the OrderID field(declared on the page by JavaScript code) to the ad serving conversion tag. As theOrderID is common to multiple systems it can be used to connect a user’s behaviourtogetherbothonandoffthesitetodeterminehowthedecisiontopurchaseismade.

CRMPlatform

Inadditiontoproductdata,collectingcustomerdataisvitaltothesuccessofthebusiness.

Customer data comes into the Custom Relationship Management tool from customersurveys and data collection points to the POS (Point of Sale) system. The data revealsseasonalbuyinghabits,trends,pricedrivenspendinganditbuildsprofilesofcustomerssothatexistingcustomerswithcertainbuyinghabitscanbemodelledagainstnewcustomerstopredictbuyingbehaviours.

OnsiteBehaviouraldata

To enhance CRM databases, and add extra value to their businesses some Advertiserspresent opt-in tick boxes in their sign up processes which allow the data they collect,aboutthoseusersonthesite,tobeusedagainforotherpurposes.Sometimesitissoldtothirdpartiesdataproviders toaggregatewithotherdatasourcesbeforebeingsoldon tootherAdvertisers.Marketers,whofailtodomorethancollectuserdetailsoffthebackofacampaign,couldfundfuturecampaignsbysellingonthedata.Ofcourseawell-nourishedCRMwillpaydividendslateronwhenansweringcrucialmarketingquestions.

Fromcontrollingthepointofentryandthepathwaythroughthesitetothepointofsale,marketersobtaindataaboutthe“order”itselfbutlesssoabouttheconsumerbehaviourintheconfinesofthesite.ThesecondplatformthatthemarketerwillspendlotsoftimeinontheAdvertiser side are theAdvertiser or Site-Side analytics system, something that themarketerwillhavealmostexclusiveaccesstouse.

TrackingtheusersontheSite-SiteSideAnalytics

Sometechnologiesgetmorepresscoveragethanothersand“SiteSideAnalytics”toolsareoftenconfusedwith thirdpartyadserversbecause theyareoneof the few technologiesthat have their own entire shelves in large book stores. Established Site-SideAnalyticsprovidersincludeIBM’sCoreMetrics,Adobe’sAnalyticsSuite(formerlycalledOmniture)andthepopular,freealternative;GoogleAnalytics.SiteSideanalyticstechnologiesworkbytheAdvertiserplacingasnippetofcodeoneverypageoftheAdvertiserwebsiteandthetoolmeasurestheincomingandoutgoingsitetraffic(mappingthejourneyuserstakethroughthesite).

ClicktoLand

SincethefirsttimetheSite-Sideanalyticstoolseesauseriswhentheyfirstarriveatthelanding page, it is common place for amarketer to try tomatch up the users clickingthrough on ads (as reported by the third party ad server) to those users landing on thelanding page (as reported by the site side analytics tool). Unless the two systems areintegratedthisisaflawedexerciseandwillraisemanydiscrepanciesbecausethecountingmethodologybetweenthetwosystemsisinherentlydifferent.

PassingthroughTokensintheclick-throughURL

Integration is possible where the click-through URL of the ad contains an additionalvariable that thesitesideanalytics toolcanrecognisewhentheuserarrives to thepage.These variables are referred to as “tokens” and when setting up the campaign, thetraffickerwillappendtheclick-throughURLtoeithercarrythroughafixedoradynamictoken.Tokeepmatterssimple,let’sfocusonthefixedtokenscenario:

AclickthroughURLmightlooklikethis:

http://www.advertiser.com/productpage1

Aclick-throughURLwithanappendedtokenmightlooklikethis:

http://www.advertiser.com/productpage1?campaignname=campaign1

In this example the actual click-through URL was changed by the trafficker in the adserverinterface(toseehowthisisdoneheadtoChapter3-AdServingOperations).Thesite-side analytics tool could be setup to recognise the variable called “campaignname”andusetheinformationpassedthroughattachedtothisvariabletogetmoreinformationabout the incominguser (suchas thecampaignname for thead they just arrived to thepage on). In this instance campaign1 is a fixed value; it doesn’t change unless thetraffickerchangesit.Inthescenarioofprovidingadynamictokentheadservermightbeclever enough to have some backend “macros” or shortcut pieces of code that willdynamicallyaddavalueintoafield.

Inthisinstancethetraffickermightinsert:

http://www.advertiser.com/productpage1?campaignname=%campaignname%

Themacroisshownwithapercentagesymboleitherside,andwhentheuserclicksonthead, the clickthroughURL is formulated in real-time so that themacro-portion pulls in amaster-valuefromtheadserverbackendcalled“campaignname”.Thisfeature isahugetimesaverforthetrafficker.

Thetraffickercanadd:

?campaignname=%campaignname%

to all click-through URLs in multiple ad campaigns for the same Advertiser and thecampaign name will dynamically change for the site side analytics tool to read theinformationandcorrectly attribute theuser and their arrival to a campaignnamewhichmatchesthemasterlistofcampaignnamesintheadserver.

Referral

Eitherway, the site-side analytics datamakes for compelling analysis.Essentially thesetoolsmeasureuserjourneysthroughthesitefromentrytopurchase.ThedataisthenusedbythewebdevelopmenttostructurethewebsiteandbythemarketingteamwhousethedatatoenrichtheCRMdatabase.Sitesideanalyticsrevealsimportantinformationabouthow users arrive at the website. This is known as “referral” information and candisseminatebetweentrafficarrivingfrompaidsearchvs.naturalsearch.Wherethesetoolsfall short are in identifying user behaviour before the user arrived at the site which iswheresite-sideanalyticsreportingmarriesupnicelywiththirdpartyadservingdata.

Itistheroleofthemarketingteamtomatchdatafrom:

• TheCRM

• ThePOS

• Thethirdpartyadserver

• TheSite-Sideanalyticstool

…toanswersomeorallofthefollowingquestions:

• Whichadsaredrivingthehighestreturn?Andwhatistheoptimumcustomerjourney?

• Whataffectdocampaignoptimizationshaveoncustomerbehaviour?

• Whatarethebehavioursofthemostvaluedcustomersandhowcanthisdatabeusedtocreateabiggerandmoreloyalcustomer-base?

We have looked in detail about the world from the perspective of the Advertiser andMarketer.ItisnowtimetoturnourattentiontotheroleplayedbytheMediaAgency.

Chapter2-Section2AgencyPlanningandBuying

• TheroleofthePlanner

• Howtoplan?

• Figure2.3

• ResearchTools

• Figure2.4

• DemographicDigitalMediaConsumptionbyPublisherWebsite

• UniqueAudience-SiteOverlap

• ResearchthroughBenchmarking

• In-Agencyplanningsolutions

• Figure2.5

• TechnologyforBuying

• AgencyTradingDesks

• Figure2.6

• TheAgency-sideMediaPlan

• SelectinganAdServer

• Whattechnology-basedServicesdoMediaAgenciesoffer?

• MediaAgencydataanalysts

MediaAgencies bring awealth of expertise to the table but not allAdvertisers employthemtomanageandrundigitaladcampaigns.ThoseAdvertisers thatfeel theyhavethetalent and buying power, utilise technologies and outsource some functions (such astrafficking)toachievethesamegoalsashiringaMediaAgency.TheriskinherentinthisapproachisthatthefullscaleoftheworkundertakenbytheMediaagencyiscloakedbydesign as a positive customer experience so that Advertisers can spend media budgetwithout getting their hands dirtywith the day to day operations. The following sectionlooksatthepredominantrolesandfunctionsundertakenbytheMediaAgency.

Theroleofthemediaagencycanbedividedintofourkeyareas,whichwewilllookatinmoredetail:

• PlanningCampaigns

• BuyingMediaspace

• ManagingCampaigns(locally,regionallyandglobally)

• ReportingandOptimizingtoAdvertiserKPIs

TheroleofthePlanner

DecidingwheretospendmarketingdollarstomeetAdvertisercampaigngoals,isatoughchoreand takesconsiderable experience toget it right.Agencies employ“planners” forthis very role. The process starts by having the planners dissect the campaign briefreceived from theAdvertiser. They then decidewhich sites and channelswill yield thebestresults.

Howtoplan?

Thechannelrecipeforcampaigneffectivenessisnotanexactscienceowingtothebuyingbehavioursoftheconsumer.Strongcrosschannelplanningstrategiesarebalancedbythemarketing principles of “the purchase funnel”, suggesting that Display advertising iseffectiveat thetopofthefunnelforconsumerawarenessandresearch,whiletheSearchchannel ismoreeffective lowerdownthefunnelwhere there isan intent topurchaseorclick. In Chapter 3we look in depth at each channel at itsmerits and how tomanagecampaignsviatheadservereffectively.

Figure 2.3: A generalisation of the paid digital marketing channels available for theMediaAgency toplanandbuyad spaceacross.Broadly speaking the rightmixcanbeformulated through careful research and succinct planning, althoughnew channels andPublishersmaketheoptimummixachallengetogetright.MoreonCampaignSetupbyChannelinChapter3.

ResearchTools

PlannersandbuyersuseresearchtoolstoestablishwhichsitesinapreferredchannelmixwillmeetthedemographictargetingandaudiencereachcriteriatosatisfytheAdvertiserobjectives for the upcoming campaign. Neilsen and ComScore are the two dominantresearchtechnologiesindigitaladvertising.Thesetwocompaniescloselytrackapanelofindividual web users across the globe with their consent (identified as true individualsrather than anonymous cookies), and use this panel as a sample of the overall internetpopulation.Thesampleisthenscaled-uptothepredictedtotalaudiencesizeoftheinternet(calledthe“internetuniverse”)andthedemographicdetailsoftheuniquevisitorspresentoneachlargePublisherwebsite.

Theoutput couldbe compared to a league table ofwebsiteswhere comparisons canbemadeacrosssiteswiththegreatestreach(calleda“uniqueaudience”),thecountofusersand the“pageviews”foreachsite (thenumberof timesallusersare loading individualpages). Given that we are talking about sample sizes the numbers are not precisereflections of all users on these sites, but they do give a clear enough indication aboutPublisherwebsitetrafficforaplannertomakesomedecisions.Theplannerwilllogintotheresearchtoolbeforethestartofacampaigntodecidethattheaudiencetheyareaimingtotargetwithaparticularcampaignbrief;usesacertainPublisherwebsite.Thenextstep

istopasssuchinformationalongtothebuyingteamatthemediaagencywhoinstigatesthepurchaseofavailableadvertisingspaceonthatsite.

Figure2.4:AnagencythatusesboththeNeilsenandComScoreresearchtoolscanexportreportsfromthoseplatformsandcomparethedataforthetwosystems.

DemographicDigitalMediaConsumptionbyPublisherWebsite

IntheexampleinFigure2.4thetopsitesforagiventimeperiodforwebsitesclassifiedinanautomotiveverticalshowthatforthefirstsiteonthelist(microsoft.com)that80/90%ofthetotalaudienceof18-25yearoldmalesoneitherofthemeasuredaudiencepanelsforthe two technologiesvisited thissite. If it is the intentof theAdvertiser (that themediaplannerontheAgencyaccountteamisservicing)totrytoreachasmany18-25yearoldmalesaspossible,itwouldbewisefortheplannertoinformthemediabuyertocontactmicrosoft.com to try to secure ad space for an upcoming campaign. If the plannerwastryingtoreachonehundredpercentofthe18-25maleaudience,bothoftheresearchtoolsagreethatitisnotpossibletodothiswithasinglesite.

UniqueAudience-SiteOverlap

Toreachonehundredpercentoftherightaudience,stillusingFigure2.4asanexample,theAgencywouldneedtoconsultadifferentreportexportedfromthesesystemscalledaUniqueSiteOverlapreport.Suchareportwouldshowtheuniquenessoftheaudienceoneachsiteintherightdemographicbucket.Atthisearlyplanningstagethepurchaseofthemedia space that results from making decisions by looking at this data is not a onehundredpercentaccuratereflectionofwhatwillactuallyoccurwhenthecampaigngoeslive.Onlyoncethecampaignisliveandallusersaretracked(asopposedtousingapanel)will it be clearwhere there isoverlap inuniqueusers.Thead server can reporton thisdirectly (seeChapter5)with itsownversionof theUniqueSiteOverlap report,but theactual delivery of the ads to the demographics stated in this table needs to be verified.Neilsen andComScore offer additional tools based off of their panel data to verify thedeliveryofadstothecorrectdemographicgroup(seeChapter5-Verification)

ResearchthroughBenchmarking

Data collected in the third party ad server from previous campaigns can be extremelyusefulindeterminingthebestPublishersitestochooseforplanningasthehistoricaldatacanbeviewedinawaythatallowsforPublisherstoberankedbyperformanceacrossall

ofthemetricsavailableinadserverreporting.BigAgenciesmayamalgamateallthedataforalltheAdvertiserstheyservicetogiveagoodindicationofpublisherperformancebyAdvertiservertical.Agenciesthathaverunenoughcampaignshistoricallycanfindoutifacombinationliketheaddimensionswillgivebetterperformanceononesiteoveranother.Such toolsmightbedeveloped inhousebyAgenciesbut some independent technologyproviders have alternatives.Mediamind’s Smart Planning tool is an option for Agencyplanners looking to improveonapre-campaignplanningstrategy.Thisad-serverownedresearch tool offers benchmarking across Publisher websites for a large pool ofAdvertisers, therebygiving amore accuratepictureofwhere certain adsmightperformeffectively,longbeforethemediaspaceisboughtandpaidfor.

In-Agencyplanningsolutions

Informeddecisionsabout thecampaignplan, lead toconversationsbetween theAgencybuyingteamandthePublishers.TheAgencybuyingprocesshasevolvedfromitsoriginsby traditional methods of buying media is still used alongside the new technology-supported buying methods. In traditional buying, the Agency buyer makes a call to acontacttheyhaveacquiredatthePublishersitewheretheyareinterestedinbuyingmediaspace. On both sides of the phone, the Publisher and Agency agree on a particular“buying-rate” or “cost model” for the Advertiser’s campaign. The Agency buyer maynegotiateapositionsothatthePublisherisresponsibleforperformanceofthecampaignandindoingsothefinalpriceoftheadspacemaybetiedtothisfigure(moreonactualcost models and costs in Chapter 2 - Publishers). The relationship between the mediabuyerandthePublishersellerinmostinstanceswillhavebeenformedovermanymonthsin which time the Publisher would meet with the buyer to provide updates about thePublisher’s media space and technology integrations. In instances where PublisherperformancemaynothavereflectedwellinresearchtoolslikeNeilsen,suchrelationshipswinthePublisherrepeatbusiness.

The request for ad space ormedia “inventory” from aMediaAgency to a Publisher isrecorded in an “insertion order” or IO (essentially an invoice). The signing of the IOsecuresthedealandthecampaigncanbegin.

Technologieshavesurfaced,createdbyeithertheMediaAgencythemselves,orbythirdpartiestoautomatethesetupofacampaignfollowingtheconfirmationofadealandkeeparecordofthenegotiationproceedings.Thesearecalled“planningtools”.TheDonovanData System (DDS) product called iDesk, is a good example of one such technology.iDeskallowstheAgencytoquickly input thedetailsof thepurchasedadspace,connectthistotheirbillingsystemandautomatically“setupacampaign”insidethethirdpartyadserver interface. iDesk is currently setup to automate this processwith all themajor adservers(seeChapter3-DisplayAdservingformoreinformationabouthowthisworks).InadditiontothisAgencybuyerscancomebacktotherecordsinthissysteminthenextcampaignorbuyingcycleandmakesure that throughoutnegotiations that thePublisherdoesnotattempttochargeahigherpricethanpreviously.

There aren’t a huge range of options forMedia Agencies that are looking to upgrade,replaceorinstallabillingsystemthatcommunicateswellwithplanningtoolsandoutputsdata that allows an agency account team to keep an eye on the pacing for marketingbudget spend across the campaign. The future promises to bring these kinds of budget

managementcapabilitiesdirectlyintotheadserver(partlymanagedthroughtheabilitytoinputcostdataintheadserverinterfaceinthetraffickingprocessasexplainedinChapter3)butforthetimebeingthemostwidelyadoptedsystemforliterallyissuingpaymentstoPublishersandkeepingtrackofallcosts:isMediaBank(alsopartoftheDDSfamilyunderthebrandnameMediaocean).Thisiscrucialbecausethemarketingbudgetmustnotjustcoveroffthecostoftheadvertisingspace,itmustalsocoverthecostofusingtechnologylike3dpartyadservingandadditionalAgencyfees.

Figure2.5:thebreadandbutterofMediaAgencyworldconsistsofResearch,Planning,Buying and Trafficking in a straight-forward campaign development workflow. As theplansforthecampaignbegintotakeholdandIOcontractsaremadewiththePublisherteam,theMediaplandevelopsasalivingdocument,referencedbymultiplepartiesoverthecampaign’slifetime.

ThisbuyingprocessiskeytoMediaAgencyinvolvementindigitaladvertising.ThesizeoflargerMediaAgenciesmeanthatonpaperalargenumberofIO’swillbedouble-signedwith the largestPublishers.Buyingmoread spaceaffects thepriceasPublishers attractMedia Agency dollars through economies of scale. At a certain volume commitment,prices are lower for media Agencies than they would be for Advertisers becauseAdvertiserswillalmostalwaysbebuyinglessadspace.

TechnologyforBuying

Buying from Publishers and Publisher networks directly is now referred to as a“traditional media buy” because as we allured to before: technology has changed thebuying process. Getting on the phone to secure ad space (inventory) from a Publisherleavestoomuchofadependencyonhumancontact,relationshipsandit takestoomuchtime.AutomaticbuyingtechnologiesnowallowforthepurchaseofmassreachinventoryknownasaPublisher“remnant inventory” (literally the ad spacewhich is unsold).Theprocessingandplacingofordersforinventoryisnowdonethrough“adexchanges”whichare trading marketplaces for media inventory (not wholly dissimilar from the stockmarket)whichcanalsooccurinrealtimewiththecorrectenablingtechnologiesinplace(RTB).Thetraditionalmediabuyingconversationsstilloccurforthe“biggestdeals”suchas large Publisher homepage takeovers; the purchase here is for ad spotswith a highervalueorlargeraudienceknownas“premiuminventory”.

AgencyTradingDesks

Therearea limitednumberofknownandpopularAdexchangeswhere the inventory isgroupedandclassifiedbythePublisherbytheaudienceexpectedtobeexposedtothatadspace.At themostbasic level it canbe said that theexchangeadvertises the saleofanaudience. Accessing the Exchanges is hard work and the technology that allows thepurchaseoftheseadspacesandaudiencesareDSPs(coveredinChapter8).The“AgencyTradingDesks”(ATD’s)arespecialistbuyingteamsthatarepartoftheAgencythatutilisetheDSPstoaccesstheExchanges.Theyhavecloseties to theAgencybuyingteamandare utilised to purchase audiences on the Exchanges, a process undertaken when thetraditionalbuyingmethodisnotused.ThisisafairlyrichtopicsoIamgoingtosavethedetails forChapter8.Thebasicpremise is thatAgencybuying teamsdon’t justhave topick up the phone to Publishers anymore; they can buy the ad space through in-housetechnologyteamswhichutilisetheDSPsnow.BigbusinessagencytradingdesksincludeWPPGroupM’sXaxisandthePublicisAudienceonDemand(AOD)platform.

Figure2.6:TheprocessofMediaAgencybuyingandsecuringmediainventoryispossibledown two pathways. The traditional method is a direct relationship-based sellingopportunity. The new methodology; predominantly used to trade remnant inventoryutilisesAdvertisingexchanges.ExpertiseinautomatedtradingismanagedbytheAgencytradingdeskteam.MoreonthismodelinChapter8.

TheAgency-sideMediaPlan

Oncethebuyingprocessiscompletetheaccountteamwillhaveintheirhandsacopyofthe“mediaplan”.Themediaplanisasingledocumentwhichclearlyshowswhichadsaredue to run andwhen.As the document is only completedwhen all parties are satisfiedwiththeinventorypurchase,itcontainsalltherelevantinformationrequiredtobegintheconstruction of the ads themselves. This is due in part to the supply of the Publisherspecifications(detailsofwhichcanbefoundinChapter2-Publishers)andwill includeinformationlikeaddimensionsandwhenthecampaignisduetogolive.Oncecomplete,themedia plan is supplied to the “trafficker” on theMediaAgency side aswell as theCreativeAgency. The process followed by the CreativeAgency is detailed in the nextsection. As for the trafficking process, the complexity of the proceeding processes isdictated by the make-up of the media plan. For instance the media channels that thecampaign is planned and booked to run on will determine on what processes andconsiderationsneedtotakeplaceinthetraffickingprocess(suchasfortraffickingadsduetoruninGooglesearchversusthoseduetoappearonFacebookortheMSNhomepage.Thenuisancesof thischannelbychannel traffickinghavebeen fullydocumented in theChapter3–AdservingOperations.

Once trafficking iscomplete, acampaign isactiveand theMediaAgencyaccount teamwill share the responsibility tomonitor the delivery of the ads and report on campaignperformance using the reporting functionality in the third party ad server. The specificreports that Media Agencies can see are demonstrated and explained in Chapter 5 -Reporting. Finally the Media Agency will complete the process of “campaignmanagement” by either optimizing the ads in the existing campaign (See Chapter 6 -

Optimization)orwillutilisetheinformationcollectedtorunasecondcampaignorextendtheexistingcampaign(SeeChapter7-Retargeting).

The campaign setup process, from research right through to trafficking should be acarefully considered operation, but time constraints around the availability of inventoryand the timelines of theAdvertiser’s ownmarketing plans canmean a rushed process.From speaking with Creative Agency, Publishers, Media Agencies and AdvertisersinvolvedindailycampaignmanagementovertheyearsIhavemanagedtodeterminethatanydigitaladcampaignneedsatleasttwoweekstogettheplanningandsetupjustright.Rarely will a campaign get such a generous window and parties are often rushed ondelivery. Every experienced trafficker can tell stories about ‘Friday night’ trafficking, awell-known burden of the trafficking role in which media plans for the weekend arereceivedatthelastworkinghouroftheweek,butariteofpassageforworkingindigitaladvertising.

During the stages of campaign setup and management (research, planning, buying,trafficking,reportingandoptimization)theAgencyaccountteamwillcontinuallycheckinwithAdvertiser, bringing them up to speed on how close to thosemarketingKPIs, thecampaign is. On some occasions the planners and buyers will manage to secure moremarketingbudgettodelivermoreadsonthemediaplantoreachagoal.TheoptimizationprocessisalsostronglyconnectedtobuyingandAgencybuyersmayhavetorevisetheirIO’swithPublishersiftheadsarenotperforming.

Thisconcludes thecommonrolesof theMediaAgencyaccount team.AswehaveseenMedia Agency’s sometimes have trading desk teams as well. Quite common are alsoTraffickingandTechnologyteams.SuchateamwouldbemadeupofAgencyemployeesthataretechnicalandskilledworkingwithintheenvironmentofthethirdpartyadserver.Where these teams exist, the in-Agency traffickingprocess is slightly different and thisteammight takeon the fullprocessof traffickingamediaplan instead for all theotheraccount teams and their campaigns. Such teams also handle the troubleshooting ofproblematiccampaigns,adsandprovideadvicearoundsolvingtechnicalqueries.Itisalsocommonforthisteamtoconducttheselectionprocessforchoosingnewtechnologiestouseincludingadservers.

SelectinganAdServer

Asthisbookwillreveal,choosinganadserverisnotaneasytask.TheselectionprocessshouldbecarefulconsideredbasedonboththeneedsoftheAgencyandtheneedsoftheAdvertiser.Asthirdpartyadservingtechnologyisrapidlyevolving,trafficandtechnologyteamson theAgencysidewill typically invite thirdpartyadservingcompanies topitchfornewbusiness(whichtendstobemoreforon-boardingnewAdvertisersormigratingwholeAdvertisersfromadifferentadserver,ratherthangoingthroughtheconsiderationprocess for each new campaign). After a period of research the traffic and technologyteamwillissueanRFPwhichcontainsalistoftechnicalquestionsaboutcapabilitiesandhigherlevelbusinessquestionsaboutcostsandplansfortheadserverinquestion.IftheRFPand thepitch that followsmeet theAdvertiserneeds theAgencywill formalise therelationshipandcompleteanadservingcontractwiththeselectedtechnologyprovider.

Whattechnology-basedServicesdoMediaAgenciesoffer?

MediaAgenciessometimesowntechnologysolutionsandselltheseontotheirAdvertiserclients as services.Technologieswhich reportoncampaignperformancearedrawing indata from the third party ad server and other data sources. Technologies for buying ortrading,which connect to exchanges and Publisher networks, are built byAgencies forexclusive use by their Advertisers and are used to attract more Advertiser business.Technologies such as Agency-owned planning tools are integrated with third party adserverstospeeduptheprocessofmovingfromplanningtotrafficking.Suchtoolsaddacompetitive advantage to one Media Agency over another and are an importantconsiderationforanyAdvertiserselectinganewMediaAgencytoworkwith.

MediaAgencydataanalysts

TheMediaAgency roles and responsibilities in campaign setup andmanagement fromresearch and planning through to buying and trafficking help demonstrate why mostAdvertiserschoose toemploy theexpertiseof suchstakeholders.The roleof theMediaAgency does not end when the campaign begins but goes on to add strength throughexpertise in optimization, reporting and the on-going trading of advertising space. It isimportant to state that all media agencies are different, even within the same holdinggroupAgenciesareknowntofollowdifferentprocessesandworkflows.SomeAgencieshave account teams but then split out the responsibility of campaign management bychannel byhaving additional teams inplace such as a search teamor an affiliate team.Further to this some Agencies who have a particular focus on data-heavy strategies,employateamofexpertanalysts;whobuilddataprocessingsystemsandretrievethemostgranulartypeofreportingfromthead-serversduringandpostcampaign.MediaAgencyteamsthathaveanalyticsresourcesareontheriseastheindustryventurestowardsgreateruseofwhathasbeen termed“bigdata”:makinguseof largedatasets suchasadservercookielevelreportingtomakebusinessdecisions.

ItisattheAdvertiser’sdiscretiontoemploytheMediaAgencyandbythesametoken,theAdvertiserchoosesaCreativeAgency.CreativeAgenciesarealsonotessential toeverydigitalcampaign;campaignswhichrunonchannelslikeSearchdonotrequireadesignerordeveloper toproduceanaestheticvisualasanadvert.ThosechannelswhichdohavethisneedsuchasVideoorStandarddisplaybringtheCreativeAgencyintothecampaignsetup process at the point that the Media Agency issues the media plan. Since thisdocument details the specifications that the creative Agency will need to build to, theCreativeAgencycannotbegintheirworkwithout it.ThenextsectionwillgointodetailabouttheroleandworkflowsundertakenbytheCreativeAgency.

Chapter2-Section3CreativeShop

• TheCreativeShop

• RelationshipwiththeAdvertiser

• RelationshipwithPublishers

• RelationshipwiththeMediaAgency

• MarketingbudgetforCreativedevelopment

• RolesandResponsibilities

• Figure2.7

• GettingtheDesignright

• DevelopmentandProduction

• Flashauthoring

• Figure2.7.1

• Actionscript

• Figure2.7.2

• Components,buildingandpublishingads

• Adstoragesystems(AssetManagementSystems)

• Localisation

• StandardQA

• Figure2.7.3

• UploadingCreative

• What’sinaname?

• CreativeInspiration

ThissectiondefinestheroleoftheCreativeAgencyandexaminestheresponsibilitiesthatexisttheretosupportthedigitalcampaignsetupprocess.Inadditionwewilltakeadeeperlookintothedefinitionofanadorcreativetounderstandtheskills,toolsandsystemsthatare mandatory for the Creative Agency to support the Advertiser, Media Agency andPublisher.

TheCreativeShop

Visualadsaredesignedandcreatedbya“creativeteam”.SuchateamcaneithersitinsideaMediaAgencyormorecommonlysitswithinanindependentAgencyofitsownright;the“CreativeAgency”(sometimesreferred toasa“CreativeShop” in theUS).Whenamediaagencyoradvertiserchooses toplanorsetupadigitaladcampaign thatdoesnotutilisecomplexvisualelements(suchaswithcampaignssetupinthesearchchannel)theservices of the creative Agency are not required. As the possible digital advertising

channels have evolved, so has the role of the Creative Agency: as we will see later,channels like Dynamic Creative Optimization (DCO) have completely changed theperception and the expectation around Creative Agency involvement throughout thecampaign.

Creative Agencies build their predominant relationships with the Advertisers whoemployee them. Relationships with Media Agencies and Publishers, which are just ascrucialtomaintainforsuccessfuladcampaignsexistfordifferentpurposeswithseparatedialoguerequired.Let’slookatthisbalancemoreclosely:

1. RelationshipwiththeAdvertiser

To firstworkwith anAdvertiser, aCreativeAgencywill attempt to advertise theskillsoftheirdesignersanddevelopersbyobtainingtheAdvertisercampaignbriefandpreparingmock-upstoshowcasetothepotential“client”.Mock-upswillbethecreative realisation of the brief into the story-board of the campaign in anaestheticallyinnovativeway.ThisworkisundertakenforfreeatcosttotheCreativeAgencythatcanhopetowintheattentionoftheAdvertiserandifsuccessful;theirbusinessforthecampaignandbeyond.SuchtacticsmakethecreativeAgencyworkverycompetitiveandtheprotectionandlimiteddistributionofthecampaignbrief,crucialtotheexistingCreativeAgency.

IndependentCreativeAgencies have very close relationshipswith theAdvertiser,more so than theywouldhavewith theMediaAgency in fact and this all comesdowntothedevelopmentofabrand.Inbranddevelopment,theCreativeAgencyisemployed to birth theAdvertiser’s brand standards and guidelines. By keeping aconsistentloyaltytothoseguidelinestheconsumerexperienceoftheAdvertiserandthe ads that they encounter enable far stronger brand and product recall. TheAdvertiserthereforeinvitestheCreativeAgencytoapproveanddevelopallcreativein digital ad campaigns as well to adhere to the creed of the brand. Havingdeveloped such standards, theAdvertiserwill look touse theCreativeAgency toapproveworkundertakenbytheMediaAgencythatmayimpedeontheperceptionof the brand. Alongside this auditing process the actual output delivered by theCreativeAgencyisreferredtointhedigitalworldas“thecreative”andconsistsofimagesandcontentintheformofdigitalfiles.

2. RelationshipwithPublishers

Publishers own the infrastructure of theweb, the sites and ad spaces themselves.ThePublishercanrestrictthetypeofadsthatcanbeusedontheirwebsites.Theselimitationsor“specifications”andmostlyuniquetoeachwebsiteandlegallymustbeobservedinorderforthecreativeto“golive”onaPublisherwebsite.Thelegalagreement to adhere to such specifications is laid out when the Media Agencycommittothepurchaseoftheinventory.

Publishers will check the creative thoroughly before the campaign goes live.Creativedevelopersthatmaytrytoside-stepthespecificationsusingclevercodingwillhave theircreativecaughtby thePublisheroperations team.OntheonehandtheinclusionofanylimitationstotheCreativeAgencymayseemlikeabarriertoexpression and creativity, but the Publisher must look out for its own interests.

These interests include the preservation of its own design and the specificationsensurethatthecreativeisasvisuallyeffectiveasitcanbeinthepublishercreatedenvironmentoftheirwebpage.

CreativeAgenciesandPublisherswilloftenworkhandinhandonexpressiveandelaborate executions that, for one campaign or two, are granted permission tobreach Publisher specifications. Such exercises canmean good publicity for bothparties,aswellasimpactfulbrandingfortheAdvertiser.

3. RelationshipwiththeMediaAgency

Commissionanartisttopaintandhewillpaintuntilthemasterpieceiscomplete,hisowninterpretationandstyleatthehelm.Fortheartistatthecreativeagency,apartythatisnotdoingthecommissioningcouldbeviewedasahindrancetothecreativeprocess.However,whenaCreativeAgencydevelops the“creative”withoutsomecollaborationfromtheMediaAgency,thingsgowrong,crucialdetailsaremissedortheintentofthecampaigngoalscouldbemisunderstood.

Theanswerisnottoinvolvebothpartiesateverystepinthecreativedevelopmentprocess. The relationship between Media Agency and Creative Agency is abalancing act, because the Creative Agency needs to retain creative directorshipwhilestill invitingfeedbackandcontextualcampaigninformationfromtheMediaAgency.TherightbalanceisbeststruckbyworkingwithbothaMediaandCreativeAgency that have a strong existing relationship. These optimum relations givesuccesstoacampaignfarbeyondtheuseofthetraditionalMediaAgency-CreativeAgencyengagement;theuseoftheMediaPlan.

Aswehaveseen,amediaplanisanimportanttooltoaligntheMediaAgency”buys”withthe creative that needs to be developed for them. Laid out in a template thatmight beforeigntoanewCreativeAgency,aweakrelationshipcantrulyimpactthetimelinesforthecampaign.Problemsthatcanarisefromnotcollaboratinginclude:

• TheCreativeAgencymaynotdelivercompletingadsintime,meaningalatestarttothecampaign

• TheMediaAgencycouldmakelastminutedemandsforcreativespecificationsthattheCreativeAgencydidn’tplanfor;thechangescanleadtoalatestarttothecampaign

• TheMediaAgency could add Publishers, sites and placements to themediaplanatthelast-minute,makingtheCreativeAgencylateondelivery,canleadtoalatestartofthecampaign.

The best advice for all parties is tomaintain a consistent contact with all stakeholdersbefore campaign launch and during campaign delivery, to ensure that the bookedadvertising space is utilised effectively. If new creative can be developed within themarketingbudgetandmediaplantimelinesthentheprocesstobuildmorecreativewillgoahead. [As a side note: these situations are as relevant to creative teams at creativeAgenciesastheymightbetocreativeteamsworkingwithintheMediaAgency.]

Inanidealworld,MediaAgencyandCreativeAgencyshouldworkhandinhandthroughthewholecampaign.CreativeAgenciesoftenrequestthattheMediaAgenciescreatethe

MediaPlanswithmoreinputfromthecreativedeveloper.Ideally,theworkflowshouldbethat theCreativeAgencyandmediaagencywork together tocomeupwith thecreativeidea and then theMedia Agency would book the suitable placements for that creativeexecution.

Let’snowlookcloserattheworkbeingundertakenbytheCreativeAgencyandwhatsortsofrolesexisttheretofacilitatethatwork.

MarketingbudgetforCreativedevelopment

Visualads,animationsandimagesaresomeofthemostexpensivecomponentsofadigitaladvertisingcampaign.Thecampaignbudget issetasideearlyonforthedevelopmentofcreativetothespecificationsofthefinalmediaplan.InturnCreativeAgenciescanchargehourly ratesor fixedcosts reflecting theoutputofads,eitherby tallying themuporbychargingbasedoncomplexityoracombinationofboth.Campaignswhichhavemoreadsormorecomplexadfunctionscanthereforecostalotmoreandhitthecampaignbudgetharderbeforethecampaignhaseverbegun.

Sincecampaignscanalsochangeovertime,theneedcanariseformoreassetsandads,soitissensibletosetasidepartofthemarketingbudgetforadditional,unforeseencreativecosts.InadditiontothiscreatingnewadsforeverynewadcampaigncangetverycostlyfortheAdvertiser.OveranAdvertiser’slifetime,millionsarespentonbuildingadsfromscratch.

Third party ad serving technology can assist in this area: the technical solution is todevelopcreative thatutilisesDCO(DynamicCreativeOptimization).Wewillcover thisoff in full detail inChapter3 butCreativeAgencies that aim to developDCO ads canprovideaverysignificantsavinginthelongtermforAdvertiserslookingtorunnumerousdigitalcampaignsoverthecourseofayearormultipleyears.Millionscanbereducedtohundredsofthousandsofdollarswithoutsacrificingcreativity.

Whenthebudgethasbeensetaside,theCreativeAgencyteamscanbeginworkingonthecreativeforthecampaign.

RolesandResponsibilities

A creative “build” requires the efforts of two distinctive roles at the Creative Agency.Theserolesarethedesignerandthedeveloper(orproducer).ThefullCreativeAgency’sstructureisformedasfollows:

• The Account Management Team: This is the team that handles therelationshipwiththeAdvertiserdirectlyandcreatesthebriefsforthecreativeteam. This team is in charge of managing expectations between theAdvertiser(referredtoastheclient)andthecreativeteam.

• TheCreativeTeam:FormedoftheArtDirectorandtheCopywriter.TheArtDirector (designer) is in charge of producing the graphics and leads thedesigning of creative containing animation. The copywriter has theresponsibilityofcomingupwiththetext(or“copy”)intheads.Theserolescollude to come up with the overall concept for the final creative output,“story-boarding”themfortheproductionteamtodevelop.

• TheProductionTeam:Theproductionteam’sstructurevariesfromAgencytoAgency. Some big Creative Agencies (like Saatchi & Saatchi) tend tooutsourcemostoftheirproductionworktospecialist“ProductionShops”.Ingeneral, a production teamwould have front-endDevelopers and back-enddevelopers.Front-enddeveloperswould in chargeof coding the “look-and-feel” part of the Advertiser’s websites using Flash, JavaScript and HTMLamong other languages. The back-end Developers in charge of coding inserversidelanguagesforAdvertiserwebsitesorincomplexin-adlanguageslikeActionscript - whichwewill cover off inmore detail in the next fewpages.

TheDeveloperswouldreceivethe“art”or“creative”fromthecreativeteam,have a discussion to agree on feasibility, to code and develop the finalisedoutputwhichisusuallyaFlashfilecalleda*.swf

Insomeinstances,standarddisplaycampaignsareeasyenoughtobedevelopeddirectlybytheartdirectorsinthecreativeteamwhogenerallyhavetherequiredFlashanimationskillstodothework.Herethereisnoneedtoemployeetheworkoftheproductionteam,whomightjustbeinvolvedinmorecomplexexecutionslikeRichMedia.

ThecreativeAgencyteamsworktogetherinthecreationofcreativeforthecampaign.Atahighleveltheprocessisasfollows(thisprocessmaydifferfromAgencytoAgency):

1. TheAccountManagementteamsolicitsthecampaignbrieffromtheAdvertiser

2. TheAccountmanagementteamreceivesthemediaplanfromtheMediaAgency

3. TheAccountManagementteamsuppliesthesedocumentstotheCreativeteam

4. TheCreativeteamdesignthecreativeandbuildouttheassets

5. TheCreativeteamsupplytheassetstotheProductionteam

6. TheProductionteamauthor,developandbuildthefinalfunctioningcreative

7. TheProductionteamQAthecreative

8. TheProductionteamuploadthecreativetothethirdpartyadserver

9. TheProduction teamsupply thereference to thecreative, to theMediaAgencyfortrafficking

TheAccountManagement teamtend tohave lessengagementwith the technology; it isthe Creative and Production teams that undertake the production-line style processingbringingadigitaldisplayad to life.The journey isone fromDesign toDevelopment toQA(testing)andthentoupload,theremainderofthissectionwillexplorethetechnologiesthatsupportthiscreativeprocess.

Figure2.7:ThefourcrucialstagesinthecreationandpurposingoftheCreativeAgencyoutput.

GettingtheDesignright

The creative teamdesign and create the images and assets for use in the final piece ofcreative.Imagesareoftencreatedtohighresolutionssothattheycanbeusedformultiplead dimensions. Designers achieve this feat using what is referred to as vector-baseddesign. Vector based design allows images to be scaled down or up without a loss ofimagequality.DesignerswilluseprogramslikeAdobePhotoshoptoproducetheseassetsbut theyarenot free todevelopveryhighquality images forallads.Aswe touchedonearlier, this is due to Publisher specifications and almost all Publishers across thewebuniversallyagreethatstandarddisplayadsshouldbenomorethanaround40-50kbinfilesize.Thecreativedesignerneedstobeverycarefultweakingimagesandassetstoensurethatthecreativeremainsaestheticallypleasingwhilestillmeetingtherequirementsonfilesize.

Richmediaoffersdesignersamoreliberalopportunitytoincludemuchhigherresolutionimageryastheconstraintsonfilesizearelifted(althoughthesetoowillbesubjecttosomeupper limitations). Video creation also falls into the remit of the designer and thecombinations of images, video and even high resolution imagery are the resultingcreationsthatthedesignerwillpassalongfordevelopmentoranimationtotheproductionteam.

The final output of the creative development process will be one of the following fileoutcomesdependingonwhatthemediaplanrequires:

StandardDisplayCampaign-Aminimumofone*.giffileperadrequired.Amaximumofone*.giffileandone*.swffileperadrequired.

RichMediaCampaign-Aminimumofone*.giffileandone*.swffileperadrequired.Amaximumofmultiplefilesofmanydifferenttypesincludingvideofiles.

In Chapter 3; Campaign Setup by Channel, we will look at why these preciseconfigurationsarerequired.Forthethissectionwewillfocusonexactlywhat*.giffilesareandwhat*.swffilesareandhowtheyarecreated.

Gif’sasmostreadersmightbeawarearejustordinaryimages.Theyareusedasafiletypebecausetheyhavealegacyforhavinggoodcompressionrateandpacknicelyintoasmallfile for the Creative Agency to workwith and deliver. The gifs would be built by thecreative team in a program such as Photoshop. The creative team will work on otherimagesanddrawingsaswellbutthesearenotoutputtedtoafinalfile.Insteadtheymakeup part of the production process known as creative authoring. It is the full process ofauthoringwhichoutputstheotherfiletypementioned:the*.swffile.

DevelopmentandProduction

The production teamworkwith the functionality of the creative images that have beenpassed along by the creative team. Creative for ad campaigns is rarely a static (still)image; it has lots of moving parts, it might be animated or contain controls forinteractivity.Itmayrespondtoexternalstimulusorinformation,andbehaveaccordinglyundercertainconditions.Theproduction teamproduce the final files, typicallyworkinginsideAdobe’sFlashauthoringprogramsothattheadcontainsthisfunctionality.Sowhatisflashauthoring?

Flashauthoring

AswetouchedonbrieflyinChapter1,AdobeFlashisaprogramthatallowsthecreationof dynamicmultimedia; typically outputting an animated file that can be played in anybrowser where the flash player “plugin” has been installed. With vast dominance onalmostallwebaccessingmachines(althoughlesssointhenowgrowingtabletandmobilespaces):Flashhasbeenthedevelopmentenvironmentofchoicefordigitaladvertising.

TheproductionteamdeveloperswillhavecopyofFlashinstalledontheirlocalmachinesand will spend many hours using it to create the desired functional effects inside thecreative.Theareadesignatedfortheassemblyofthecreativeisreferredtoasthe“stage”,whichdisplaysthecurrent“frame”thatisbeingworkedon.Framescanbelayeredontopofone another or becomehidden and as a default theFlash animationwould contain aseries of frames representing the progression of the animation. Just as with an imagecreatedusingPhotoshop, the programhas a selectionof artistic tools, visual filters andeffects.Amongthemostusefuloftheseeffectsare“tweens”(pre-codedanimationpathstochangeshapesorprovidemotion)andtheavailabilityofActionscript.

Theauthoringtoolalsoutilisesaspecialstorageareaofimportedimages,shapesandcodecalledthe“library”whichcanbeuniquetoeachanimation.Itemsinthelibraryusedintheanimationare formallycalled“assets”.Today theseare theatomsof thead servingandbuildingworld, there isnoentitysmaller toplaywith,moveand transport than“assets”buttodaythesearemostlytrappedintotheauthoringtoolandcannotbereleasedfromthecageoftheresultingcreative.InChapter3-DynamicCreative,wewilllookathownewformsofcreativedevelopmentarechangingthis.

Figure2.7.1:Amock-upof theFlashauthoringtool, thecreativedeveloper’sworkshop.Many hourswill be spent inside this interface to compile the creative in an *.FLA filebefore it is published inside the same tool to a *.SWF file. The tool is very flexibleallowing for theuseofdesigner-ready toolsaswellasActionscript code,whichcanbetriggerbyanythingfromthepassageoftimetotheclickingofabutton.

Actionscript

Actionscript is a cousin of JavaScript,muchof the code looks similar andwith a goodgraspofJavaScript,Actionscriptisnothardtopickup.Itisaveryversatilelanguagethatcan be used to mathematically produce images and animation rather than importingcreative assets.The advantageof thismethod is that theoverall fileweight of the finalcreativewillbereduced(toleaveroomforcreativeflarewithmorecodeormoreassets).

The creative flare thatActionscript provides thedeveloperwith shouldbemetwith the

objective toofferauserexperience thatmeetswhatwecall in the industry:“the lowestcommondenominator”.Stolen from thepagesofahigh schoolmaths textbook,LowestCommonDenominatormeansthatcreativeshouldbedevelopedtoallowevenuserswiththeslowestcomputersandinternetconnectionstohaveagreatexperience.ThePublisherspecifications are designed to ensure this so when building to those specifications thedeveloper compares the specifications across all of the sites on the media plan andproduces creative that adheres to the strictest guidelines. For instance; if the strictestspecificationsdemandeda2mbfilebutotherpublishersweremoreliberal,allowingfilesizesupto5mbthenthedeveloperwouldbuildthecreativetoa2mbspecificationsothatthesamefilecouldbeusedacrossallofthepublishers.ThisisfarmoreefficientonthecreativesidethatcreatingaversionforeachPublishertocorrespondingspecifications.

OntopofthePublishersdemands, thedeveloperisrecommendedtoconsiderguidelinesthatallPublishershavemissedfromtheirspecificationsbutdoaffecttheuser’sexperiencewith the ad. One such example is the reliance on Actionscript within the creative to“draw”objects and shapes in the creative in real-timewhen theuser encounters the ad.Althoughthismethodreducesfileweight,itcanbeaheftydrainontheRAMoftheuser’smachineandslowdowntheperformanceoftheirbrowser.Suchadslookoutstandingonhighspecificationcomputers,butasabestpracticetheLowestCommonDenominatoroftheuser’sowndeviceisanimportantconsideration.

So farwe have described the design and build of the creative as a freehand operation,undertakenbythedeveloperinsidetheFlashauthoringtool.Thirdpartyadservershavealwaysgiventhisfreedomtoflashdevelopers;allowedthemtouserawcodeandstartthebuildprocessforthecreativefromscratchforeachproject,providinghelpfiles,trainingandmanuals toassist thebuild.As the industryhasevolved, theskills required toworkwithcodeinthismannerhavebecomeevermoredemandingandcomplexsothenumberofdevelopersintheindustryhasshrunkwhilethenumberofdesignershasincreased.Aswe have seen this has resulted in the out-sourcing of the development work to otherAgencies.Anothergrowingtrendistooutsourcetofreelancedevelopers.

To assist less-skilled or busier developers, the third party ad server provides assistanceinside the Flash authoring tool itself by supplying pre-built templates and using smallmodulescalled“Components”togivetheFlashauthoringtooladirectconnectiontothethirdpartyadserver.Thisappliestoalltheprocesseswehavelookedatalready:DesignandProduction,butalsoextendstoQAandfacilitates;aneasyuploadintothethirdpartyadserverUserinterface.

Figure2.7.2:ComponentsopenasaseparatepanelintheFlashAuthoringtoolandallowfor the fast creation of creative to ad server specifications quickly and at scale forcampaigns where creative needs to be produced in bulk. Creative producers that arealreadyfamiliarwithcomponentslikeDG/Mediamind’sworkshoptoolcanbeexpectedtodelivercreativefasterthanthosecreativeanddevelopersthatarenot.

Components,buildingandpublishingads

Third party ad serving providers offer a range of components to make it easier fordevelopers tobuildads.Componentscome inpackageswhichareusually free to installand use. The DG Mediamind package is called “the workshop” and installs onto thedevelopersmachine.Updates are usually supplied every fewmonths just like anyothercomputerprogramtoensurethebestcreativekitisavailabletothedeveloper.Amongthecomponents will be a pre-built video player as well as a pre-built data capture form.Perhaps the most useful among the components is a library of pre-designed templates(something thatDGMediamind refers to as “Blocks”). Loading one of these templateswillunpackontotheauthoringstageanear-completeFlashproject.Clickingafewsettingsandmaking some adjustments allows the developer to tweak the template tomatch thePublisherspecificationsforthecreative.Thedeveloperwillthendraganddroptheassetsorimagescreatedbythedesignteamintothesameprojectandassemblethemsothattheframesofthefinalcreativecanbeplayedthroughinrunningorder.

Thedescribedprojectthatthedeveloperisworkingonisactuallyafilecalledan*.FLAfile. An FLA file cannot be used outside of the Flash authoring tool but stores thedesignerswork in the design stage andwhen the file is passed along to the productionteam, it stores their work too.When the creative has been finished, the developer canchooseto“Publish”thecreativewhichallowstheFlashAuthoringtooltosavethefiletothedeveloper’smachineasa*.SWFfile.The*.SWFfilesarethecreativeoutputthataresentontothenextstageinthesetupoftheadcampaign.*.FLAfilesshouldnotleavethecreativedeveloper’smachinebut theyshouldberetainedincaseproblemsariselateron(moreinformationonthiscanbefoundinChapter3-Richmedia).*.SWFfilescannotbeadjustedlateronwithoutopeningupthe*.FLAfileandchoosingthe“publish”optionintheFlashauthoringinterface.FLAfilesareusefultokeepholdof.Shoulda*.SWFeverbemisplacedbyanyparty, theFLAcanbeusedtogenerateanothercopyof the*.SWFfile.

SometimesanAdvertiserwill runcampaignsquite scarcelyand to lowercosts,willuseand adapt old creative for future campaigns. For flexibility to use other developers andCreativeAgencies theAdvertiser should ensure that all filesused in the creationof thecreativearesuppliedtotheAdvertiserdirectlyattheendofthedevelopmentprocess.ThisincludesallFLA’s,SWF’s,Images,VideosandFonts.(Font’s,whichfacilitatethedesignofwrittencontentinacreativeactuallyonlydisplaythatwayonthedevelopersmachinebecausetheyareloadedintotheFLAfileseparately).WhentheproductionteamconvertsthefilesfromFLAtoSWFthefontsbecomeembeddedintheSWFsothattheydisplaycorrectly.Ifaproductionteamweretoout-sourcedevelopmentworktoafreelancerwhosuppliedallfilesexcepttheFontsattheendoftheproject,weretheproductionteamwishtomakeadjustmenttothecreativeby“re-publishing”theSWFfromtheFLAfile,anytextused foranuncommon fontwoulddisplay incorrectly.The freelancerwouldhave tobesuretosupplythefontfilehehadusedattheendoftheproject.

Adstoragesystems(AssetManagementSystems)

High performing creative production teams are those that are most organised when somanyfilescome intoplay. Ifeachcreativehas threefiles for instance,and tendifferentdimensionsarecreated,intendifferentlanguages,thereare300filesbeingdesignedandbuilt.Thisdoesn’ttakeintoaccountthemorerawassetslikeoriginalvideofilesthatthedesignteammayhavebeenworkingon.ClearlythereisaneedforintelligentfilestoragesystemsandintheworldofaddevelopmentthesearecalledAssetManagementSystems.

Somepartiesunfamiliarwiththeinnerworkingsofanadservermaybeunfamiliarwiththestoragefacilitates that ithasavailable.Thethirdpartyadserver isnotdesignedasastorage facility for creative. Creative uploaded to the ad server cannot easily bedownloaded again, and the ad server does not support the uploadof file types used fordevelopmentsuchasFLAfilesorPSDfiles(thefilesusedbythedesignerinthecreationof layered images inPhotoshop).There ishoweveranareacalled the“creative library”.This is tree of folders inside the third party ad serving interface that can visualise andarrangecreativebefore it ispickedupfor the traffickingprocess(moreontraffickinginChapter3-Displayadserving).Thecreativelibrarycanbeseenas‘thedropoffpoint’forthecreative,butit isanarearelativelyrudimentarycomparedtotrueAssetmanagementsystems.

Assetmanagementsystemsarenotdesignedformassaddistribution likethethirdpartyadserverbutprovidesecurearchivingsothathistoricaladscanberetrievedlater.Thesesystemsalsoprovidetheabilitytotagorlabelcreativeinintuitiveways(whichiscalledthe application of “meta data”) so that they can be easily retrieved, recovered andassembled later on. Assessment Management systems are also essential to CreativeAgenciesthatworkacrossmanybordersinmanydifferentcountries.

Localisation

BigadvertiserswithcampaignsthatruninmultiplecountriesandlanguageswillemployoneCreativeAgencytoproduceaselectionofcore templatesatacentralpointgloballywhichis thendistributedtoCreativeAgencies in localcountrieswheretheadcampaignwill be running. This process is called “Localisation”. The Assessment Managementsystemallowsforcontrolledstorageanddistributiontotheselocalteamswherethelocal

creative designers will adjust the creative both for language and to local Publisherspecifications. This way nuisances in themedia plan at the local country level are notcompletelyrestrictedtostringentbrandingguidelinesbuthavesomeflexibility.TheAssetManagement system allows the globalised Advertiser to retain a degree of control byfeeding brand guidelines through to local teams and approving localised creative tominimumbrandguidelines.Globalcampaignsco-ordinatedinthiswayoftenfollowalongchainofbrandguidelineapprovalfromtheoriginalglobalmarketingteamtoensurethateven for the smallest Publishers; creative and brand guidelines are adhered to. AssetManagementSystemsarestillintheirinfancybuthelptheproductionteamtokeeptheirdeskstidywhenthenumberoffilesindevelopmentescalates.

StandardQA

Once the production team have finished developing and “publishing” the *.SWF theymust test it to make sure it works accordingly. This process is called “QA” (QualityAssurance) and it can be a complex undertaking if a lot of files are involved and theAdvertiserwants to be very thorough about the final creative delivery. To be thoroughtherewouldbefourdifferentQAprocessesthatwouldoccurinthelifetimeofanad:

1. LocalMachineQAbythecreativeproducer

2. QAinthethirdpartyadserveradpreviewinterface

3. QAbythePublisher(tocheckitagainstthePublisherspecifications)

Thereasontodosomanychecksisthatthe*.SWFisbeingtransportedandembeddedinan environmentwhere there are a lot of other things going on (the Publisherwebsite).ParticularlywhendevelopingRichMediaads,morecangowrongsoalongerandamorestringentQAprocessisrecommended.WhenitcomestothecreationofStandardDisplay*.SWFs the development cycle will be much shorter and the ads far less complex. Ininstanceswheretherearethousandsofadstocheck,step(2.)mightbemissedoutentirelyor the work out-sourced or undertaken by spot-checking the files. Typically a goodcreativeproductionteamwillundertakethefirstQAcheckonthislist.

WhenperformingaQAtheproductionteamarecheckingthatthecreativeisfunctioninganddisplayasexpected.Onalocalmachinethiscaneasilybeachievedbypublishingthe*.swffileandviewingitinthebrowserwindoworintheauthoringtoolpreviewscreen.AtrueQAinvolvedchecksthatalltheinteractivefunctionalityisworking,thattheadhasnoexcessloopsorglitchesandthatitmeetsthelowestcommondenominatorspecificationsforthePublisher.Themostcrucialcheckisthefunctionalityoftheabilitytoclick,andthedeveloper will encode this into the Actionscript of the creative. This specific script iscalled“theclickTag”andthethirdpartyadservercomponentsensurethattheclickTagisalways embedded in the creative from the outset of the production process behind thescenes.Despitethis,theabsenceorincorrectcodingoftheclickTagisthemostcommonreasonthatadsgetrejectedintheQAprocessin(2.)andthisisbecausetheclickTagcodeis stillnot standardisedamong theadservers. It isa littledifferentdependingonwhichonegetsused.

Figure2.7.3:An example of a clickTag coded in Actionscript 2.0. Ad servers are veryparticularaboutthepreciseusageoftheirspecificclickTag,creativewhichfailstoincludethe correct clickTag code can be rejected by the ad server. This failure can causesignificantdelaystothestricttimelinesofadigitaladcampaign.

Uploadingthecreative

WiththecreativeQA’dbytheproductionteamitcannowbeuploadedtothethirdpartyad server. Once uploaded the creative will live in the creative library until the MediaAgencyassigned“trafficker”selectsitforuseinthesetupoftheadcampaign.Thethirdparty ad-server supplied components have built-in the functionality which allows thecreativetouploadthefinal,testedcreativetotheadserver.Todothisthedeveloperneedstohaveausernameandpassword setupwhichwill ensure thecreative isuploaded to adesignatedprivatecreativelibrary.ThethirdpartyadserverwillissuethesedetailstotheCreativeAgencywiththeapprovaloftheAdvertiserorMediaAgency.Thefinalcreativecanbeuploadedassinglefilesorinbulk.Someadserversallowcreativeinseveralfilestobezippedupinanarchivefileorzipfileanduploadedtothethirdpartyadserverthatway.The upload process causes the third party ad server to solicit a reference number.Thisiscalledan“AdID”or“CreativeID”andthecomponentwhichfacilitiestheuploadofthecreativedisplaysthisIDtothedeveloperfortheirreference.Thecreativedeveloperwill then supply this number to the Media Agency trafficker. Some ad servers offeradditional QA services, just to be sure that the creative at least meets ad serverspecifications(whichwillmeanthatthecreativecontainsthespecifiedclickTagandthatitcontains all the required methods that any creative built using the ad server suppliedcomponentswillcontaininanycase).

What’sinaname?

In the process of the uploading the *.SWF file, the developer should apply the samenamingconventionthatwillbeusedwhennaminganduploadingalltheother*.SWFfilesto the ad server.Thebestnamingconventions are reached in collaborationbetween thedeveloperand theMediaAgencybecause thecreative filenamewill eventually find itsway into ad server reporting, and ad performance can be analysed by the creative filename.Ifa*.SWFfilenameissetuptocontaininformationlike“creativemessage”,thenthe performance of creative using a specific message can be determined later on forexample.

Theroleofproductionteamendswiththeuploadingofthecreative.Theassistanceofthisteammaybecalleduponatalaterdatetoundertakethesametasksandproducefurther*.SWFfilesoradaptstheexistingcreativebyrepublishing.

The Creative Agency are occasionally asked to participate further in the setup of thecampaignintheadserverinterfaceitselfbuttheroletoplayhereisanegotiationbetweenthe role of the Media Agency trafficker and the creative control held by the creativeproducer. With the advent of DCO (see Chapter 3 - DCO) creative functionality anddesigncanbecontrolledoutsideoftheFlashauthoringenvironmentandintheadserverUIinstead.Itisthechoiceofthecreativeproducertotakeonmoreofatraffickingroletocontrolthebirthofnewversionsofthesamecreative.

CreativeInspiration

Thecanvasoftheadmaynotsoundlikethemostglamorousoflocationsforthebirthofamasterpiecebutincrediblecreationsareborneverydaytheworldoverandthesegemsareincreasinglyrecognised,adoptedandcopied.Suchexecutionscanwinawardsleadingtofame for designers and production teams.Themarriage of cutting edge technology andcompellingartworkmakesforsometrulyinspiringadsthatputcampaignsinthehistorybooks and make Agencies and Advertisers think twice about employing the mostformidablecreativetalent.ThisisthereasontomakeawiseconsiderationforaCreativeAgencyandtoharnessalongtermrelationshipwithtalenteddesignersanddevelopers.

Nowthatwehavelookedattheendtoendprocessofcreativedesignandproductionitistime to turnattention to the roleof thePublishers, theyare the last stakeholderwewillfocuson.Powerfulandextremelydifferentinthewaytheyapproachandutilisetheweb,theyaretheallythateveryAdvertiserquicklyrealisesitneedstohave.Afterall,withouttheadspace,thereisnowheretodisplaytheartproducedbytheCreativeAgencyandnoaudiencetoshowitto.

Chapter2-Section4Publishers• WhatarePublishers?

• Figure2.8.1

• HowdoPublishersselltheiradspace?

• CPM

• CPC

• CPA

• PublisherNetworks

• Figure2.8.2

• NicheNetworks

• BlindNetworks

• Publisheradservers

• Figure2.8.3

• Inventory

• PublisherSpecifications

• PublisherQA

• PublisherAdOps

• Figure2.8.4

• Adprioritisation

• Frequencycapping

• OverandUnderDelivery

• BrowserTargeting

• Figure2.8.5

• HowdoesPublisherreportingdifferfromthirdpartyreports?

• Discrepancies

WhatarePublishers?

Theonlineadvertising industryrefers to themajorityof thewebsiteson theworldwidewebasPublisherwebsites.Thesearethewebsitesthatattractinternetusers;theyarehubsfor both audiences and content, often pumping vast sums ofmoney into creating suchcontent, to attract audiences.The audiencehas an experienceon aPublisherwebsite ofmostly free, rich information which brings the internet alive; the illusion is that theinternetsomehowoffersaworldofinformationfornexttozerocost.TheunderlyingtruthofcourseisthatcontentofthePublishersitescostsmoneytocreate,tostore,tomanage

and it is through theeyeballsof thoseaudiencememberswhichgenerate the truevalue.The audience attention rests on the content, but within peripheral vision sits digitaladvertisingwhereAdvertisersspendvastsumsofmoneyinthehopethatadswillbeseen,clickedandpotentiallyusedasaconduittomakeapurchase.

Figure2.8.1:ExamplePublisherPageswithemptyadspotsinvaryingsizes.

Figure 2.8.1 shows an example of three webpages making up a very small fictionalPublishersite.AsaPublisherifIwanttomonetisemywebsitewithoutcharginguserstoaccess thesethreepages thenIhavea totalof6adspotsIcansell. IfmyPublishersiteonlyhasoneuservisiteachpageonceoveraperiodofamonththenit’ssafetosaythatI’mnotgoingtomakeanymoneybutifmyPublishersiteisverypopularwithmillionsofusersvisitingthesethreepagesinamonththensuddenlyIcanselltheadspaceandmakemoney.HowmuchadspacedoIhaveavailabletosell?Theanswerislots;notjust6adspaces,buteverytimethepageisloadedIcouldshowtheuseranewad.Ifeachuserseesjust one page and there are a million users a month then my “inventory” is a millionimpressions.

ThecatchtothisisthatIactuallyhavetopredictmyhowmanyusersorhowmuchtrafficIwillhaveinthefuturetoknowhowmuchadspaceIhavetosellandifIwanttochopupthis inventory and sell it tomultipleAdvertisers, how can I do it? These questions aresolvedbyPublisher-side ad serving technology.Sellingoff the inventory impressionbyimpressioninthiswayisjustoneofthemethodsPublishersutilisetoselltheiradspace.

HowdoPublishersselltheiradspace?

ThePublishersellstheadspaceas“inventory”ontoMediaAgencybuyers(seeChapter2-AgencyPlanning)acrossthreecore“pricingmodels”.Themodelsare:

1. CPMorcostper ‘mille’,offeringa thousandadviewsor“impressions”ataprice.CPM reflects the value of the audience seeing the ads; the further down thepurchasefunneltheyare,thehighervaluetheadspace;iftheadspotisseenmoreoften,byalesstargetedaudience,thelowertheCPM.AlsothemorevaluablethePublisherisperceivedtobeorthemorenichetheaudience;thehigherthecosttobuy the space. How does a Publisher decide on the CPM pricing? This is morecomplexthanitmaysound.Pricesforadinventoryvaryenormouslybetweensites,pages and Publishers. Considerations for coming to a reasonable and fair priceincludeanyvariablethatmightinfluencetheuser’svisualexposuretoanadonthefinal purchase with that Advertiser. This will include: the Format of the ad, theaudience type that it is shown to, the total amount of traffic that the ad spot isexpectedtobeexposedto,thetotaltimeofexposure,thestrengthofthecontentandwhethertheAdvertiserislookingtoselldirectlythroughtheadorjustconductanexercise in branding. Beyond these considerations the pricing needs to remain

competitive with other publishers. In some cases Advertiser and Agencies helpPublisherstodeterminethefinalpricebasedonthepricingofsimilaradspotsinthemarketintheprocessofnegotiatingthecostoftheinventory.CPMsarenotanexactscienceoruniformacross theweb.ForAgencies andAdvertisers it pays toget aquoteforadspotswithseveraldifferentPublisherstogetagoodguideonamarketprice.

2. CPCmodelsdominatedtheearlyinternet;forcingtheAdvertisertopayforclicksnotthe number of times the ad is seen. Clicks are more valuable because the userdemonstratessomeintenttoengagewiththead,productorbrand.ButclicksdonotworkinfavourofthePublisher;ifthecreativeisbadlydesignedthePublishertakesthehitbyshowinganadthousandsoftimeswithnoclicks.AswewillseelaterinthissectionthePublishercanalsotrackthead’sclickbycalculatingthelikelihoodforastandarduser toclickontheadover thecourseofacampaignthePublishercandeterminewhatagoodpricewouldbetochargeperclick.IfaPublishercangetawaywithservingjustafewadsbutexpectsaclickforalmosteveryuser,itwillbemore favourable to charge on a CPC. Where CPC models exist “long tail”Publishers have been known to try and dupe the ad serving counters employingsmallarmiesofinternetrobotprogramstomimicusersclicking.Suchtechniquesdonottakeintoaccountthatadservershavefilteringlayersinplacethatdonotpassfraudulentclickactivityontothereportingpartoftheadserver.

3. ThelastmodelisCPAorcostperacquisition.ThepresenceofaconversiontrackerontheAdvertisersite(suppliedbyeitherthePublisherthemselvesorthethirdpartyad server) ties a conversion back to the ad and publisher which drove thatconversion, this can also be extended to a lead or sign up (more details on thepathwaytoconversionarecoveredinChapter4).Forthismodela“conversion”justmeanstriggeringaconversiontag”sothiscanmeanthecompletionofanyspecificaction.Duetotheflexibilityofthisdefinitiontherearepricingmodelsthatexistlike“CPL” (Cost Per lead) where the action of generating the lead will fire a“conversiontag”-sointhissenseCPAandCPLarethesame.

Running a business selling ad space is never quite as simple as three pricing models.OtherswhichIhaveencountered:FixedPrice,CPMcombinedwithafrequencycap,FOC(freeofcharge),CPMV(costpervisual impression),CPE(CostperEngagement),Rev-sharemodels(Revenue-Share),Sponsorship(amodelparticularlypopularintheChinesemarket)andawholerangeofdealswhereonlineadspaceispackagedwithadspacefordifferentchannelssuchasTV.Newtechnologiesinadservingwillmeanlotsmoreoptionsforinventorypurchaseinthefuture.

PublisherNetworks

With thesemodels inhand thePublishercanapproach their themediaAgencybuyerorAdvertiser.Publishersmustmonetise theirwebsites tooffset thecostsofproducingandmaintainingcontentandwithoutsubscriptionthiscanonlybeachievedthroughthesaleofad inventory. The Publisher typically has a sales team that packages up the inventorybasedoncommonknowledgeor,onoccasion,assumptionsabouttheaudienceviewingit.ForinstancetheAutosSectionofawebsite is typicallyclassedasinventorysuitableformiddle agedmen in themarket to purchase a vehicle. The Publisher sales teammight

choose tocontactAutosAdvertisers (orMediaAgencies representingAutosAdvertisersdirectly)tosell,closingthedealbyhavingthenewcustomersignanIO(InsertionOrder).

Figure2.8.2:CampaignSetupWorkflowbetweenAdvertiserorMediaAgencyandPublisher

The completed IO is processed by a Publisher ad operations team who use a“Publisher ad server” to load ads and ad tags onto the Publisher website. APublishernetworkactsasasalesandoperationshouseformultiplePublishers,andnormallyonlytakeonPublisherswithacertainamountofavailableinventoryandaudiences which are in demand by Advertisers. For a Publisher, signing up tobecomepartofanetworkmeansthesecurityofoutsourcingmostoftheadvertisingcampaign operations, and in some instances contractual agreements are made toguaranteerevenue.

Publishernetworkshugelyreducewastageofinventorybybeingabletoselectfroma greater pool of Advertisers and capping the user’s exposure to a particularcampaign or Advertisers messages across multiple sites. Networks having beengoingthroughbigchangesinthelasttwoyears,ThebiggestchangeofallhasbeentheadventofAdvertisingExchanges.AdvertisingExchangesandSSPsarecoveredinmoredetailinChapter8-Programmatic.

NicheNetworks

ANetwork isa term thrownarounda lot in thePublisherspace.MobileNetworkand Video Network are examples of Publisher Networks that are more channelspecificwhich are becomingmuchmore common in online advertising. AnothertypeofnetworkistheAffiliateNetworkanditscomplexitygivesitachannelallofitsown.Tounderstandthedifferencebriefly;Affiliatenetworksoperateundermoredynamic costmodels than standard Publisher networks and offer a differentiatedreachoftheonlinemediaaudience.AsaffiliationisactuallyconsideredaseparatechanneltobothSearchandDisplay;thisbookdedicatesacompletesectiontoitinChapter3-Affiliates.

BlindNetworks

Networks sell the available inventory but aren’t always completely transparentabout exactly which pages the ads will appear on or even which sites in theirPublishernetworktheywillbeservedon.WithablindnetworkbuythereisnowayfortheAdvertisertogotoapagetoseetheiradwhenthecampaigndoesgolive.

Onlyreportingwillrevealthedeliverystatisticswhichiswhyusingathirdpartyadserverforthese“buys”isessential.

The relationshipbetween thePublisherandmediabuyer is facilitated through theagreementofanIOandsinceseveralPublisherswillbeontheMediaplanforanygivencampaign,thejobofdeterminingexactlyhowmuchtospendwithwhocanbedeterminedbytherelationshipthatthePublishersalescontacthaswiththatMediaAgency buyer. Once the relationship has been established it’s time to utilise thetechnologytoassistwiththecompletionofthebuyingprocess.

Publisheradservers

Thereare twosortsofAdserver.Firstly is the thirdpartyad server (alsoknownas the“Advertiser ad server” which we have been referencing to date) and secondly is “thePublisheradserver”.The twoarecommonlyconfusedbutvery fewcompaniesproduceboth technologies under the same branding; the exception to the rule isDoubleClick,offeringDFA (DART for Advertisers) as the Advertiser ad server andDFP (DART forPublishers)asthePublisheradserver(andmergingthetwounderarebrandingexercisewhichistakingGooglesomeyearstoachieve).

A Publisher ad server is a platform that measures and manages the ad spaces on aPublisher web site, sites or domains. It is typically rolled out only to single sites ornetworks as opposed to Advertiser ad servers that have a much larger cross-Networkreach. Publisher ad servers are most commonly accessed by the advertising operationsteamattheNetworkorPublisher.

InitiallythePublisheradservercanbeusedtoestablishhowmanyadspotsonthesiteareavailable to receive advertising currently and in the future. Sales houses and reps forNetworksandPublishers,utilisethetechnologytopackagetheavailableadinventoryforsale.ExamplesofPublisherAdServersincludeAdtech,DFAandOpenX.

Figure2.8.3:ManagementofPublisherSites-ContentVs.Ads.AdsarethenservedbythePublisheradserver(suchasDFA)beingservedthemselvesbyeitheralocallyservedadorathirdpartyservedad.

Inventory

Once available inventory has been sold to an Advertiser or Agency; the advertisingoperationsteamonthePublishersideawaitsthedetailsofthesignedIO,includingtheadtags“trafficked”fromtheAdvertiseradserver.Sometimesadtagsarenotsupplied,suchas in instances where Publishers have sold inventory and agreed to serve the assetsdirectlyfromtheirPublisheradserverordomain.Thisisalsoreferredtoas“siteserving”and,insomeinstancesoffersaloweroverheadtotheAdvertiserwhilecompromisingonthebenefitsofde-duplicationandthethirdpartyreportverificationreceivedfromthethirdpartyadserver.In2014intheUK,sometraditionalTVbroadcasterscontinuetoonlysite-serve “instream” video ads rather than receive third party tags (more on Instream inChapter 3). Ads which eventually appear on Publisher pages must be built to explicitspecificationssoasnottoinhibittheuserexperienceonthePublishersite.

PublisherSpecifications

PublisherswillpubliclyreleasespecificationsaccordingtowhichadsneedtobebuiltbytheCreativeAgency.PublicavailabilityensuresthattherecanbenoexcuseforaCreativeAgencywith themediaplan inhand todevelopanadwhichwillnotberejectedby thePublisherbeforethecampaigngoeslive.AvailableinventoryisalsosoldunderthestrictconditionsthattheCreativeAgencyadherestothepublishedspecifications.ForStandardDisplayadsthespecificationsusuallyconformtotheindustryacceptedspecificationsforDisplay ads, set out by the IAB(http://www.iab.net/guidelines/508676/508767/displayguidelines ). The IAB advise thatPublishers place restrictions on file size (40-50kb is a standard file size for a standardDisplayad)aswellasthenumberofloopsthattheadisallowedtohaveandlimitationson “automatic play” and audio volume. Specifications also clearly provide creativeguidelinesforthecontentofthead,foritssuitabilityonthePublishersite,aswellasrulesfor ad dimensions (please see Chapter 3 - Display Ads and RichMedia ads for moredetails).

PublisherQA

At thestartof thecampaignsetup thePublisheradoperations teamwill receive theadsfromtheMediaAgencyorAdvertiser(eitherasthirdpartyadtagsorrawcreative).ThefirststepinthesetupprocessistoundertakeaperiodoftestingcalledthePublisherQAprocess.Thiswillinvolveloadingtheadsintoatestenvironmentwhichwillreplicateasclosely as possible the onsite experience of the loading of the actual ad. This testenvironment,alsocalleda“stagingarea”looksandactsverymuchlikethelivePublishersitebutallowsthePublisherto:

a. Check theadconforms to thePublisherspecificationsbycheckingpage loadspeedorperformance,loopingandsizing

b. Checkingthattheaddoesnotcontainanymaliciouscode(includingmalware)

c. Checkingthattheaddoesnothaveabadinteractionwithotherelementsonthepage(failuresof theadtoperhaps loadontothecorrect“layer”of thepagecanbeoverriddeninthePublisher’sCMSsystem).

PublisherAdOps

The advertising operations team load the ad tags or assets into the Publisher ad server

(providingtheyhavepassedQA)andapplyvariousrulestotheirdeliverysuchtheagreedgeographic location of the users IP address, ad prioritisation and frequency capping.Specific ad delivery ensures the ads appear where the Advertiser has paid for them toappear.Forsomeagreementsthismaynotbeaspecificpagebutrather“runofnetwork”or RON which infers a randomised delivery to any ad spots (of the matching addimensions) anywhere on the network of sites connected to the Publisher ad serveraccount.Thissamecollectionofsettingsgives theoperations teamthechance toensurethattheadisdeliveredonlytospecificbrowsertypesortospecificIPaddresses(orgeolocations).

Figure2.8.4 :Publisher/NetworkAdOperations.On therightof thediagramthePublisher team (sales and operations) liaise with the media buyer to share theavailableinventoryandcompletethepurchaseinadvanceofaparticulardate.Thissame team can access the Publisher ad server interface from their machines todeterminewhatisavailableandtocontrolwhatgetsdeliveredandwhen.

Adprioritisation

APublisherhas thepower toprioritisewhichads to show tousersbasedonhowhighaprioritythoseadsarecomparedtootheradsonthenetwork.PrioritisationisusedtooverridethedeliveryofadswhichareworthmoremoneytothePublisherover a shorter periodof timeor to showadsof a lowerworthwhen inventory isunsold.Suchadsarereferredtoas“houseads”andmaynotbeadsatallbutvisualassetsusedtodirectuserstoparticularpartsofthePublisherwebsite.

Frequencycapping

Frequency capping is amore powerful feature in the Publisher ad servingworldthan for the third party ad server. Set the frequency cap in the latter, andbackupimages will be displayed when the cap is reached (as the third party ad serveralways needs to serve something). But set the frequency cap in the Publisher adserver and upon hitting the cap, the user will be shown an ad for a completelydifferent Advertiser; therefore more effectively capping the user exposure to asingleAdvertiser.Frequencycapsareusuallysettolimittheuser’sexposuretojustacampaigntoahandfulof“impressions”overagiventimeperiodoftwentyfourhours or less. Publishers are able to use such settings to their advantage to pushmoreads for aparticularAdvertiserover a shorterperiodof time. It is important

therefore for the Advertiser to specify to the Publisher when booking theircampaign the requirement of an effective frequency cap. Here the third party adserver plays a strong role by allowing the Advertiser to report on the effectivenumber of times a user should see an ad in a given time period to ensure theyconvert,clickorperformadesiredfunction.Notenforcingafrequencycapcanseetheagreedpurchaseof inventory“burned through” inaveryshortperiodof time.Alsosettingthecaptoolongmayimpacttheeffectivereach(usersmaynotseeanadoftenenough)soitisanessentialbalancingacttosetthecapcorrectly.

OverandUnderDelivery

APublisher ad server is supposed to be an effective tool inmaking sure that thenumberofadspotssoldtotheAdvertisermatchesthenumberofadsdisplayedtousers during the agreed time frame. Due to the complexity ofmany different adcampaigns(inventorylookups,pagedynamicsandsuddensitetrafficspikesaswellas ad loaddiscrepancies) a campaigncanunder-deliver (which is aburden to theAdvertiserwho is looking to display all the ads they paid for) or over-deliver (aburden on the Publisher which may see their agreed margins on sold inventory,diminish). Over and Under delivery can have knock-on effects to future bookedinventory by cannibalising available space or leaving valuable gaps where therewerepreviouslybooked campaigns.Contractual compromises betweenPublishersand Advertisers are common place and now an accepted, but still a frustratingoccurrenceintheplanningandbuyingofcampaigns.

BrowserTargeting

Publisheradservershavetheability to targetusersbasedontheirbrowserandIPaddress.Themost commonuse for this functionality is to ensure that the ads arebeingdeliveredtoIPaddressesinthecountrywhere“audience”soldtothebuyerisdescribedasbeinglocated.Thisensures thatusers thatarriveto thePublishersitefromsomewhereelseintheworldarenotshownadsbyAdvertisersthatmayhavelittleornopresenceinthatcountry.

Once thePublisher adoperations teamhasprocessed the IOand inputted the campaigndetailsintothePublisherAdserverinterface,theadcampaignwillgoliveatthesettimeandthePublisherAdServerwillworktodelivertheadsevenlyoverthegiventimeperiod(orhoweverthedeliveryhasbeensetup).Atthelaststageinthesetupofeachad,theadoperations team will include small pieces of code called “macros” which allow thePublisherad server to trackclicksandpreventads fromcachingon the localor remotemachines as the ads are delivered. [As a side note: caching is a concept whereby acomputerwillstoreacopyofanimagetosavetimeandbandwidthbyhavingitloadfromlocalmemoryratherthatpullingtheinformationdownfromtheserverforeachload.Itisimportanttopreventthisfromhappeningwithadservingbecauseifthetraffickermakeschanges to the setup of the campaign or the ad itself, the change will not appeareverywherebecausethehistoricalimagewillstillbedisplayed].ThemechanismusedtopreventcachingisaconceptthatexistsinthethirdpartyadserveraswellasthePublisherad server.Recording clicks is also something duplicated in both systems.When the adoperations team is finished with the trafficker’s ad tag it can end up looking like this(nuancesofthisparticularmethodofaddingmacrosdiffersamongthevariousPublisher

adserverssothisexampleisnotaccurateforallPublisherfacingsystems):

Figure2.8.5:InthisexampletheadopsteamreceivestheadtagfromtheMediaAgencytrafficker(thetraffickerhassentthistagfromthethirdpartyadserverMediaplexontothePublisher inquestion)Theadops teamamendsthe tagwithacachebuster(%n)andaclicktrackingmacro(%c)othercharacteramendmentsaremadetomakethetagsuitableforservingfromthePublisherAdserverDFA

To ensure that clicks are tracked accurately, some Publisher sites and networks includeinstructionsinthePublisherspecificationsthattheclickTagiscodedinaparticularway.WehavetouchedtheissueofclickTaginclusiontheprevioussectioninthischapter(TheCreativeAgency)butitispossiblethatthePublisherrequirementsregardingtheuseoftheclickTag and the ad server requirements conflict so that it is not possible to fulfil bothrequirements.In this instancetheMediaAgencymust informtheCreativeAgencyas towhichsetofspecificationstobuildto,essentiallyallowingthereportingofclicksontheadservertype(thethirdpartyadserver)tooverallthecountingofclicksinthePublisheradserver.WhenindoubtitisalwaysrecommendtobuildtotheadserverspecificationsontheuseoftheclickTag.Asthecampaigngetsunderway,theadservers(ofbothtypes)begintostoredataandgeneratereporting.

HowdoesPublisherreportingdifferfromthirdpartyreports?Aswiththethirdpartyadserver,attheendofacampaignorduring,thePublisheradservercandeliverareporttoshowhowtheadsperformedandwhatwasdeliveredtotheuser.Theactualtechnicalmechanism that counts these statistics actually counts them at slightly different timesbetween these different ad server types (differing anywhere between one hundred andseveral thousandmicroseconds).Thismeansthat ifyouweretocollectadeliveryreportfromthe thirdpartyadserverandcompare itwith theone fromthePublisheradserveryouarelikelytoseeadiscrepancybetweenthenumbers.ItistheroleofthePublisherAdOperationsteamtosupplythereportingtotheMediaAgency(whomayormaynotuseasareferencetooptimizetheircampaign).TheAdOperationsteamcanalsopullthereportsfromthePublisheradservinginterfacetocheckthatthecampaignisdeliveringcorrectlyandmakemanualadjustmentstothedeliverysettingstobringtheactualdeliveryclosertothenumberspromisedintheIO.

Discrepancies

TheIABhascreatedstandardsaroundacceptablelimitsofdiscrepancy(betweenthefiguresreportedbetweendifferentcountingsystems)whichstandatfivetotenpercent (plus or minus). Larger discrepancies can either be investigated, or the

AdvertisercancontractuallyagreewiththePublishertoalwayspayoutonthethirdpartyadservingfigureswherethecountingmethodologyofthethirdpartyadserverhasbeenverifiedbytheIAB.DiscrepancyinvestigationsbythesupportteamatthethirdpartyadserveralongsidethePublishercantakealongtime,as thedatasetsneed tobeas comparable aspossible and the technologiesmayneed to be testedside by side over time to uncover potential issues. The most common cause ofdiscrepanciesbetweenPublisherandAdvertiser technology ishumanerror,wheretagsareimplementedincorrectlyorreportsarepulledagainstthewrongtags,datesor placements, all of which are easily remedied by tighter processes and bettercommunicationbetweenallpartiesinvolvedinthecampaign.Manythirdpartyadservers will allow Publisher to have their own access to third party reportingdirectlysothattheycankeepaneyeondeliverythatway.PublishersaccessingthirdpartyadserverreportingsystemswillbeabletoseeallofthecampaignsrunningontheirnetworkorsitewhichmayincludemultipleAdvertisers,campaignsandMediaAgencies.

Providing the ad campaign goes off without a hitch, the ads get delivered to thePublisher’s audience over the agreed time frame and under the agreed conditions, theMediaAgencyisastepfurthertowardsmeetingtheadvertiser’scampaigngoals.Wehavebegunlookingintothesetupofadigitaladvertisingcampaignbyexaminingeachofthefourmajor stakeholders and their connection to each other. The underpinning technicaldetail that achieves the feat of digital ad campaign delivery requires a much deeperunderstandingontechnicaldetail,flushingouttheworkthatthetraffickermustundertake.Each channel is setup differently but for a third party ad server to be used to its fullpotentialitmustbeusedtotrackallchannelsthatmakeupanAdvertisersmarketingmix.

PartBCampaignSetupbyChannel

PartBCampaignSetupbyChannel• Figure2.8.6

• HowdoestheendtoendCampaignManagementworkflowwork?

• Billing,ElectronicIOandAdservingrates

Wehave lookedat eachof the stakeholdergroupsassociatedwithanonlineadvertisingcampaign and some of the technologies and processes they use in the lead up to thecampaign setup, most of which occurs outside of the third party ad server. The actualprocess of setting up the advertising campaign and distributing the ads and creativethroughtothePublishersisdoneinachannelbychannelprocessbyloggingdirectlyintotheadserverinterface.Butwhatdoesitmeantosetupacampaignchannelbychannel?

Channelsevolve,changeandmatureovertime.Therehasbeenaneffortonthesideofthetechnology providers andAgencies to try and identify the propertieswhich distinguishchannels from one another and although on the surface they seem to be different, theunderlyingtrackingtechnologyisessentiallythesame.Essentiallytherearethreetypesofbasictrackingtechnologywhichallowforreportingintheadserver,theseareimpressiontracking, click tracking and interaction tracking. In various combinations andimplementationsthesetrackerscanbeusedtotrackadvertisinginallmediachannelsbutitisthepropertiesofthesechannelswhichadtothecomplexitybehindthesetup.TheruleisthatthemorecomplextheMediaAgencywouldliketohavetheirdatareportedbytheadserver,themorecomplexthesetupneedstobe.

With these complex insights it becomes easier to overcome new tracking challenges tofeedadvertisingcampaignswithmoredatatoattainhigherROI.Thefeaturesavailabletoaccomplishprecisiontrackingintheadserverarenowhighlyevolvedandallowforvastchangesfromasingleswitchoradaptationstotrackinginchannelsthatmightspringupliterallyovernight.

Which channels does an Advertiser need to run a digital ad campaign across to besuccessful?Thequickansweristhatthereisnomagicmix.Channelswhichhaveyieldedthebestresultsinthepastbecomefamiliarterritorytoplannersandbuyersandareover-used, driving up the price ofmedia (such as in the SearchChannel). Constant channelexperimentationisrecommendedsincetherightmixwillrequiresanadaptablestrategy.

Figure2.8.6:Areminderofthecurrentstateofplayforthechannelsavailabletomediaplanners andbuyers onwhich to planandbuy inventoryanddrive the target audiencedownthepurchasefunnel.Thisexamplediagramisaguideratherthanarule.

HowdoestheendtoendCampaignManagementworkflowwork?

Toreview:theoverallcampaignworkflowlookslikethis:

1. theAgencyorAdvertiserwilltypicallymakeadealwithoneormorePublishersfortheirinventory

2. theCreativeAgencywillbuildtheads.

3. TheAgency/Advertiserwillthentakethereinsthemselvestomanagethecampaign,channelbychannel.

Itisonthislastitemthatthethirdpartyadservercomesintoitsown;carefullytrackingthewholeadcampaignandwithittrackingpotentialcustomersaswellasdeliveringtheads to user machines, churning out the tracking data into reports (to understandperformance)and thengiving theAgencytheoption toadjustoroptimize thecampaignsetupforthebestreturn.ThisprocessiscyclicalthroughoutthelifetimeofthecampaignandthenthesamelessonscanbeappliedforfuturecampaignsuntiltheAgency/AdvertisertreatstheprocessasaKPIhittingproductionline,andgreatexpertisecanbebornoutofrinseandrepeat.

Apparent exponential technology growth means that this production line changesdramaticallyeverythreeyearsorsoandthereforeahungerforlearningandadaption(trialanderror)makesforthemostsuccessfulAdvertisingcampaignsandthebestdigitalmediaAgencies.

Billing,ElectronicIOandAdservingrates

Thirdpartyadservingisacostthatsitsontopofcostsforthemediaspaceandalthoughitis essential to keep costs down for every campaign, selecting a new ad server is notsomethingthattypicallyhappensoncampaignbycampaignbasis.Gettingthebestdealon“adservingrates”isunderstandablyimportantforallpartiesthathavetobearthecost,butadservingratesthemselvesarecontractualandreviewedoveraperiodofafewyears.Adserving rates are usually set a price on the delivery ads across the ad server reportingmetric “impressions”. However, compared to media costs, ad serving costs aresignificantlylowerandforthemostbasicadservingandtrackingshouldveryrarelytakeasizable percentage of total campaign spend. Bolting on other technologies to improveinsights,performanceandtolowercosts(suchasDCO)willraisecostshigher,but thenthegainsaresignificant.MediaAgenciesmanagingcampaignsonbehalfofAdvertiserssometimespassthecostofadservingdirectlyalongtotheAdvertiserbutthisisnotaruleforallMediaAgenciessoitpaystoshoparoundandaskquestionsaboutadservingrateswhenanAdvertiserisselectingaMediaAgencytoworkwithlongterm.

Nowthatthebillsareoutoftheway,itistimetointroducetheconceptof“trafficking”.

Chapter3AdservingOperations

Chapter3AdservingOperations• Trafficking

• ThethirdpartyadserveruserInterface

• UserAccessforcampaignsetup/Permissions

• Usingmultiplethirdpartyad-servers(double-tracking)

• Figure3.0.1

• Figure3.0.2

• Thegoldenruleofadserving

• How does the trafficker check the ads are working? (Setup verification using anHTTPtracer)

Trafficking

Thecampaignsetupprocessisreferredtoas“Trafficking”andisundertakenbyeither:

• TheMediaAgency

• TheAdvertiser

• Thethirdpartyadserversupportteam(outsourced)

• Aspecialisttraffickingcompany(outsourced)

Theemergenceofspecialisttraffickingcompaniesisnotanewconceptintheworldofadserving.BigplayersincludeTheorem,TrafficRich,andOperative.Attheseorganisationslargenumbersof technically trainedstaff (usuallybased incountrieswhere staff canbehired at a low cost) will complete the trafficking process according to the media plansuppliedbytheMediaAgency.ThesecompaniesprovidesupportformostoftheknownadserversandcanbeessentialforlargeAdvertisermarketingteamswhodonotemployanAgencyanddonothavetheresourcestocompletetraffickingthemselves.

Thespeedofthetraffickingprocessdependsonthecomplexityofthecampaignandtheexperienceofthe“Trafficker”(thepersondoingthetrafficking).Mediaplanssuppliedtotraffickers on a Friday afternoon for a campaign live date of the weekend all toocommonly experience a delay owing in part to problems such as incomplete CreativeAgenciessuppliedcreativeorproblemsmeetingPublishersspecifications.

ThethirdpartyadserveruserInterface

The rest of this Chapter will make reference to the Trafficker using the third party adserver to setup and control the tracking of the digital ad campaign. The third party adserverhasauniversaluserinterfacewhichcanbeaccessedonanycomputerconnectedtotheinternet.Onceloggedintothisinterfacethetraffickercanperformtherolenecessarytoconductthesetasks.Althoughtheinterfaceitselfdiffersinlook,feelanddesignbetweenthe thirdparty ad servers on themarket, the features available in the tracking setup foreachchannelarealmostidentical.Wheresuchfeaturesareuniquetoaparticularadserver,

thisbookwillmake reference tosuch factsand indicatewhich thirdpartyadserver thefeature isunique to.As iterated in the introduction to thisbook, someof these featuresmaybe universally available in all ad servers or adapted by the time this bookgoes toprint or in the future. It is important to check with third party ad server technologycompaniestogetthemostuptodateinformationaboutfeatureavailability.

UserAccessforcampaignsetup/Permissions

Atraffickerhastheappropriateaccesslevels insidethethirdpartyadserverinterfacetomakechangestoallaspectsofthecampaign.ItisadvisedforAgenciesandAdvertiserstobefullyawareofthetrafficker’saccesspermissionsbeforetheyembarkonthetraffickingprocess.Itisimportanttobeawareofwhatpartsofthesetupcouldbesubjecttohumanerror,iftasksarecompletedinadequately.Itisforthisreasonthatthesetechnologieshaveveryspecificlevelsofenablerpermissionsinplace,so thatsomesettingswhichmaybetieddirectlytoafeeorcharge,areoutofreachofinexperiencedtraffickers.

Nowbeforeweexploreintotheintricacyofeachsetupbychanneltherearetwotopicstotouchonthatapplytotraffickingregardlessofwhichchannelisbeingtracked:

1. DoubleTracking

2. SetupVerificationofliveads

1. Usingmultiplethirdpartyad-servers(double-tracking)

Sometimes Advertisers will employ multiple Agencies to work on acampaign by dividing thework up by the channel that they belong to (forinstance some Agencies might be specialists in running Search campaignswhere others usually manage Display). When these Agencies are usingdifferentthirdpartyadserversorsiteservingthroughPublishersdirectly,itisimportant to ensure that each ad server (or the main ad server used forreporting)cantrackalloftheonlineactivityasclearlyaspossibletopreventtheduplicationofaudiencesandusers.

What thismeans inpractice is that a thirdpartyad serverwilldropa thirdparty cookie on a user when it serves them an ad (see Chapter 1). If twoseparate Media Agencies are reporting their campaign results back to thesameAdvertiserfromtwodifferentadservers,itwillnotbecleariftheyaredescribingthebehaviourofthesameindividual.Furthermoreitwillbeverydifficult todescribehowausercame intocontactwithbothsetsofactivityoverthesametimeperiodinordertodistinguishwhichadorMediaAgencywasresponsibleforpushingtheuserfurtherdownthepurchasefunnel.

Figure3.0.1:Theproblemathand;workingwithtwoseparateMediaAgenciesorat least two separate ad serving systems means that ads existing outside of thesystemareessentially invisible to thatAgencyanda truepicture in single report,with a consistent methodology, does not exist. Building the results frommultiplesystemreportsisextremelychallenging.

EssentiallytheactivitytrackedinadserverAisinvisibletotheactivitytrackedinad server B and visa-versa. On several occasions I have sat down with puzzledmedia planners who cannot understand why the search channel activity is beingcontrolledby their sistermediaAgency, is not showingup in their standard thirdpartyadservercampaignreports.TheanswerisofcoursethatthesisterAgencyisusingadifferentthirdpartyadserver(orasectionofthesamethirdpartyadserverwhichiscutofffromthemainadcampaigntracking).

Figure 3.0.2: The solution: double-tracking. Cheap tracking pixel technology isgeneratedandsubsequentlyimplementedtorevealtheinvisibleadactivitytooneoftheAgencies(AgencyA)whowillnowbeinapositiontorepresentallactivityinasinglereport(ReportA).

To resolve this problem the traffickerswill need tomake sure that the ad server

fromwhichthefinalreportsarebeinggenerated,createsextratrackingcodetogointotheactivitybeingcompletedbythesecondaryadserver.

Thiswayalltheactivityistrackedtogether(becauseembeddedtrackingcodefromthe primary ad server in the second ad server reveals those ads so that they arerecognisedbytheprimaryadserverasnativeadsbelongingtotheadcampaign).

ThesameshouldbedoneforPublisherservedcreative(thecodeneedstobecreatedin the third party ad server and added to the ads when they deliver form thePublisheradserverifsite-served).Toallowforthistracking,thetraffickercreatesaseriesoftrackingpixelsintheadserver(aprocesswewillcoverinmoredepthinthenextsection).Thiswholeprocessmakesde-duplicatedreportingavailable(alsocalled “De-duplication”) but there is an ad serving cost to tracking things in twosystemsreferredtoas“double-tracking”.

De-duplication is a golden rule for successful ad serving; alwaysmaking surethat theentireaudience foralladsbelonging toanAdvertiser,are tracked inasinglesystem.

Double-trackingisamoreadvancedconceptbutaneverydayrealityforAdvertiserswho have a lot of digital activity to run and a need to bring their reporting datatogethertomakesenseofit.Thisisoneoftheadvantagesofusingasinglesystemfor datamanagement, although a single third party ad server does the trick, newtechnologies are now emerging that appear to facilitate this need such as DataManagementPlatforms(DMPs).Thisissuewithsuchamethodisthatthetrackingmethodology between twowholly different ad serving systemswill cause unseenerrorsinthecollectionandmeasurementofthedata.ThisismuchmoreadvancedsowewillsavefurtherdiscussionaboutDMPsuntilChapter8-Programmatic.

2. Howdoesthetraffickerchecktheadsareworking?(SetupverificationusinganHTTPtracer)

ItisimportanttoknowoncethetraffickingprocessiscompletethattheadsareinplaceandbeingservedasexpectedonPublisherpages.ThiscanbedonewithouthavingtoaccessthereportinginterfaceinthethirdpartyadserverandforpeaceofmindsomeAdvertiserwillwanttobeabletoperformthischeckthemselves.

AnHTTP Tracer is a small program that runs alongside the browser andmonitors thepacketsofdatatravellingtoandfromtheuser’sbrowser.Thisisenormouslyusefulwhenchecking to see if code has been implemented on a page correctly. By loading thepublisherpageupwheretheadisdisplayedandhavingthetraceropentheservercallscanbeviewedinrealtime,demonstratingthattrackingisindeedinplace.

Mosttracersarefree.Fiddler(astandaloneprogram)andFirebug(aFirefoxadd-on)arepopular choices. As well as seeing if ads are clicking, redirecting and loading upparametersorother files, the tracers cannote if eventhandlers arebeing fired forRichMediaadsanddisplayvariablesthatarebeingpassedontotheserver(forfurtherdetailsaboutsetupverificationpleaseseeChapter3-Displayadserving).

Ondayoneofcampaignlaunch, thetraffickercanutilisemonitoringtoolsandthenpullreportsfromthethirdpartyadservertocheckondelivery,performanceandtotalsalesor

conversions(seeChapter4).

These are two generalised (non-channel specific) considerations for the seasonedtrafficker,butthefulldisplaytraffickingprocessliesahead.

Chapter3-Section2Displayadservingoperations

• Dayinthelifeofthetrafficker

• FoundationLeveltraffickingforstandarddisplay

• Creatingaplaceintheadservertosetupthecampaign

• Thirdpartyadserverhierarchy

• Figure3.1

• Settings at Agency and Advertiser levels - SiloCapabilities

• Figure3.2

• Settings at Agency and Advertiser levels - SharingCapabilities

• Structuring thecampaign setup to replicate thePublisherson themediaplan

• MediaPlan

• Figure3.2.1

• TraffickingSheetsforoutsourcetrafficking

• Sites

• Identifytheformatsoftheadsandcreatingtheplacements

• AdFormats

• Banner

• Leaderboard

• Skyscraper

• MPU

• RoadblockandSkins

• PlacementsandPackages

• CostPackages

• Assignthecreativeandsettingrotationcontrols

• NamingConventions

• Ad&Creative

• AutomatedQA

• AdPreview

• Addingtheclick-throughURL:clickTag

• 302Redirects

• Addingmultipleclick-throughURLs

• Adclassifications/CustomCategoriesandsearching

• AdDuplication

• Figure3.2.2

• Figure3.2.3

• ReadyingtheAdsforAssignment

• Backupgif

• PossiblePlacementAssignments

• Figure3.2.4

• Adassignment

• Multipleadstoasingleplacement

• Asingleadtomultipleplacements

• Asingleadtoasingleplacement

• Figure3.2.5

• AdjustingtheAdControls

• FrequencyCapping

• AdRotation

• Sequencing

• Weighting

• AdPrioritization

• TimeBased

• PostCapGif

• Thetwolevelsofcontrol

• Figure3.2.6

• TraffickingtheTagstothePublishers

• AdTagsandTagFormats

• Figure3.2.7

• Flighting

• ThoroughlyCheckinganAdislive

• Intermediateleveltraffickingforstandarddisplay

• HowdoesatraffickermanagecampaignswithhundredsofAds?

• BulktraffickingthroughExceltemplates

• BulktraffickingwithExcelastheUI

• AutomatedtraffickingsetupviaadserverAPI

• Figure3.2.7.1

• TheJavaScriptcarriagecapability

• SurveyvendorCodeandthirdpartyscripts-inads

• Figure3.2.8

• Trackingbutnotserving

• IsitpossiblethataPublishercannotsupport thirdpartyadserving?

• Figure3.2.9

• AnoteonClickCommands

• AdReplication

• AdDeliveryandReplication

• Server-SideCaching

• Whatexactlyisatag?

• MakingabiggerimpactwithDisplayCampaigns

This section runs through the trafficking process from end to end, using the DisplayChannelasanexample.CampaignsforMobile,Instream,RichMediaandDCOgenerallyfollow the sameworkflow (i.e. the creation of theMediaPlan, the adding of Sites, thecreationofPlacements,theassignmentofAds,theadditionalofAdControlsorWeightingrulesandthesendingoftheadtagstothePublisher).AswearefocussingontheDisplayChannel,andthemajorityofthelegworkisdonebythetrafficker,itisthisrolethatwillundertakethemajorityoftasksthatfollow.

Dayinthelifeofthetrafficker

Priortothetraffickingprocess,themediaplanmakesitswayaroundtheMediaAgency,before being handed over to the designated trafficker. In an ideal world when thisdocumentiscompleted:thePublisherssitesarelisted,thenumberofbookedimpressionsis highlighted against each one and the creative names are filled in to an agreedconvention.MediaAgenciesarerarelyabletobringtogethersomuchconciseinformationin time, and more often than not will pass along the media plan to the traffickerincomplete.Media plans should come as close to a standardised document as possible(being the same document layout used in all Agencies around theworld) but currentlymediaplanscomein lotsofshapesandsizes.For this reason traffickingcompaniesandsomeadserversissueatemplatecalleda“traffickingsheet”.Thetraffickingsheetisjustamediaplan,filledoutinaspecificlayoutandthebenefitisthatthisallowsforhighspeedhumanormachineprocessing.

The traffickerwillwork from the layout in trafficking sheet, attempting to replicate the

requiredsetupintheuserinterfaceofthethirdpartyadserver.Informationmissingfromsuch a document slows down the trafficking process and ultimately the campaign.Adheringtothestrictinstructionscoupledwiththetraffickingsheettoaiditscompletion,should be followed to the letter to keep the correct pace for the digital ad campaigndeliverytimeline.

Thetraffickerwillkeeptothefollowingworkflow:

Foundation-levelTraffickingforStandardDisplay

1. Creatingaplaceintheadservertosetupthecampaign

2. StructuringthecampaignsetuptoreplicatethePublishersonthemediaplan

3. Identify the formats of the ads from the supplied plan and create theplacementstomatch

4. Assignthecreativetotheseplacementsandsetrotationcontrols

5. Trafficthetags

Therestofthissectionwillfocusonthissimpleworkflowandthenlookatthetraffickingofmore advancedmedia plans, showing how the trafficker can take advantage of bulktraffickingtechnology,workingdirectlyfromexcel.

Intermediate-levelTraffickingforStandardDisplay

6. Bulktraffickingand4thpartytracking

7. AdReplication

8. MakingabiggerimpactwithDisplayCampaigns

1. CreatingaplaceintheadservertosetupthecampaignThesuccessofacampaignandthedatathat thethirdpartyadservercollects,notonlydependsonacleanandeffectivetraffickingprocessandcampaignsetup,butalsoonconsiderationsmadebythetraffickingteamlongbeforeacampaignbegins.Theseconsiderationsincludetheorganisationofallcampaignsatahigherlevelinwhat is termed“theadserverhierarchy”.Planningandimplementinganoptimumhierarchysetupwillsavehoursofworklateron.

Thethirdpartyadserverhierarchy

Thehierarchyofthethirdpartyadserverisregrettablynotuniformacrossallthirdpartyadserversand isa reason thatmanyMediaAgencieshaveapreferred thirdpartyadservingprovider.Switchingtoanotheradservercanmeanare-educationintohowtheadserverhierarchyisconstructed.

Toreviewfromthebeginningofthisbook;whenatraffickerfirstlogsintotheuserinterface of the third party ad server they will note that the whole interface isdivided into Campaign Setup versus Reporting/Analytics. Some ad servers issuedifferent access credentials to the trafficker to divide these areas up, restrictingaccesstothosethatdonotneedtoseeparticularpartsoftheinterface.Shouldthetrafficker wish to begin the process of Campaign setup, they navigate to the

campaignsetupsection.

The first thing to note about campaign setup is that access is limited to theAdvertisersandCampaignsthatthetraffickerhasbeengrantedpermissionstobothreadandappend.Tocoinaphrase from theworldofcomputerprogramming, thetraffickerissaidtohave“readandwritepermissions”.Withthecorrectpermissionsinhand,thehierarchythatthetraffickerinitiallyarrivesathasmanylevels.Eachoftheselevelsservesapurpose.

DoubleClick (DFA) hierarchy is structured so that the highest level is called“Network”, the next level down is called “Sub Network”. The traditional Atlashierarchy begins with the level “Company” followed by a level called “Branch”beforemovingdownto“Client”andthento“Advertiser”.

Mediamind offers an “Advertiser” and “Advertiser-Brand” level at the top of thehierarchy. Each of these folder-like entities has a rather unique set of rules. Themore folders there are in a hierarchy (both horizontally - across to multipleAdvertisersandvertically,downtocampaignlevel),themoresituationsthatathirdpartyadservercancaterfor,theseincludeveryadvancedusecases.

Figure 3.1The trafficker initially logs into the third party ad serving campaignmanagement and setup interface and drills down into the standardised hierarchy.TheverytoplevelsofthehierarchyarecommonlyhiddenfromthetraffickersothetraffickerappearstobeginthejourneyinafolderfortheirownAgencydisplayingalltheAdvertisersthattheymanage.InturntheAdvertiserfoldercontainsalloftheCampaigns that have been setup or are being managed under that Advertiser.PleasenotethatwearefocussingonthesetupforStandardDisplaycampaignsinthissection.Hierarchynavigationisnotcompletelyconsistentforallchannels,mostnotablythearrangementisdifferentforthe“Search”channel..

SettingsatAgencyandAdvertiserlevels-SiloCapabilities

Thetoplevelentitiesinthethirdpartyadserverinterface,offertheirownsettingswhichcanbeappliedtothefoldersandentitiesthattheycontain.Thisisusefulforapplying bulk settings to the hierarchy that sits below. Examples of the uniquecapabilitiesofthesetoplevelentitiesareasfollows:

a. AllowingfortheclassificationofthewholeAdvertiserintoaverticaldefinitionsuchas“Automobile”or“banking”(this information is fed intobenchmarkreportingandpubliclypublishedbytheadserveronarollingcadence-SeeChapter5-Reporting)

b. BillingSettings(sothatallactivityassociatedwiththeaccountcanbecapturedbytheadserverbillingsystemsandtheadservercostscanbeinvoicedtothepayingclient)

c. Contact Settings (usually Publisher contact details for the distribution of adtags-seelaterinthissectionunder‘Flighting’)

d. Bulkapplicationofthirdpartytrackingorthirdpartyapplications(suchasthe

deployment of the Ad Choices badge to all ads - more detail of which isavailableinChapter7-Retargeting)

e. Advertiser Universal settings such as Verification settings, Time Zone orlocationsettings

This last item applies a rule to the Advertiser as a whole to establish how datashouldbecollected, storedandprocessedby thead server.Thepresenceof thesesettingsconstrictsthedatacollectionprocesswhichinformsthereportingsystem.Inessencetheyactasfiltersandhaveabigimpactonreportingwhenadjusted.Giventhe more complex nature of these settings they are discussed in more detail inChapter5-reporting,andareomittedfromthissection.

Adserverhierarchiesthatsilotheirdataandtrackingaremorecommonplacethanthosethatofferthealternativesharingmodel.

Figure 3.2The two types of account structure or hierarchy are shown: The Silomodel and the SharingModel. Siloedmodels; althoughmore simple, are a littlerestrictivewhenbigAdvertisersneedtoextendtheinfluence,controlandcountingmethodology across to other Advertisers and Agencies that may be workingtogetherunderajointventure.

SettingsatAgencyandAdvertiserlevels-SharingCapabilities

Underthesharingmodel,thetopofthehierarchycanbecustomisedtoshareinformationwith other top level entities. This is a feature common place in the originalAtlas thirdparty ad server structure and allows for campaign-collaboration between multipleAgencies and Advertisers. The most common settings which are shared utilising this‘bridge’are:

• ConversionTags(seeChapter4fordescription;thisisanalternativemethodofdata-sharingtothemorepopular“piggybacking”method)

• Double Counting (Duplication andDe-duplication settings) The presence ofthis setting allows for the tracking of channels like Natural Search (seeChapter3-Search)

• Threshold and Time-zone Settings (porting Silo settings across to otherAdvertisersthatmaybenefitfromstandardisingaparticularsettingamongstthem).

Controlofthesesharedsettingsandentitiesisgiventothesourceentity/account.Itisseenasabenefittobeabletouseasingleaccounttocontrolsettingsinaremoteandseparatehierarchy.Issuescanoccurwhereatraffickeramendsaparticularsetting,unawarethattheadjustmentaffectsother(seeminglydisconnected)accounts. It isan importantfeatureoftheAdServeruserinterfacetherefore,toclearlybroadcasttothetraffickerifanypartofthehierarchyisbeingsharedtootheraccounts,andpreciselywherethatisoccurring.

2. Structuring thecampaignsetup to replicate thePublisherson themediaplan

After the trafficker has logged into the ad server user interface (and is presentedwithalistofalltheAdvertisersthattheyhavebeengrantedaccessto),thecorrectAdvertiser folder is selected and clicked on, bringing the trafficker to a list ofCampaignsthatarestoredunderthatAdvertiser.

In these early rafters of the ad server hierarchy, the trafficker will not have therequiredpermissions to creatednewAdvertisers orAgency’s.Thisworkmust beundertaken by the ad server support team (to prevent hierarchies from beingcrowdedwithnewAdvertisers).Owingtothis,AdvertisersnewtoAgenciesornewtothirdpartyadserverstakeextradaystobesetupforthefirsttime.OnlyonceanAdvertiserentityisinplace,cancampaignsbecreatedwithinit.

MediaPlan

AscoveredinChapter2,themediaplanisamanufactureddocumentationkeepingthe record of all Publisher bookings for multiple Publishers within a singlecampaign. From the Media Agency’s perspective, the Media Plan is thatdocumentation; however “media plan” is also a third party ad server hierarchyentityinitsownright.ThismeansthattheMediaAgencyversionofthemediaplanneedstobeprocessedbythetrafficker(oradaptedintoatraffickingsheet)sothattheplanisfullyreplicatedintheadserver.

The trafficking sheet, is predominantly in a spread sheet format with each rowrepresentingtherelationshipbetweenaPublisherwebsiteadlocation/spot(calleda“placement”)andthecreative/adthatistoappearinthatspot.ThemediaplanalsolooksmuchlikeaprojectplanorGanttchartasitfeaturesatimelinewhichclearlydemonstrateshowlongtheadwillbeassignedto theplacementanddisplayedonthePublishersitefor.

Figure3.2.1Aexampleofamediaplan.

Since physical media plans rarely look like the end result in the ad server userinterface, some Media Agencies are asked to convert their media plandocumentation into the “trafficking sheet” layout or template referenced at the

beginningof this section.Such sheets are particularly useful because they canbefedintoautomatedsystems,whichcangeneratetherequiredsetupwithoututilisingthetimeandresourceofatrafficker.Somein-agencyplanningtoolscoveredoffinChapter2arecapableofundertakingthisrole,sidesteppingtheneedforasmuchtraffickingheadcountontheMediaAgencyside.Thesesystems(suchasAdazzle)have an API connection with the third party ad server in order to enable suchautomatedprocesses.(Soasnottodivergefromthecoretraffickingprocessthisisexplained at the end of this section under this section in Automated and Bulktrafficking.)TraffickingSheetsforoutsourcetrafficking

When outsourcing trafficking to an external trafficking company, supplying acompletedtraffickingsheetwillbemoreassistivethansupplyingtheoriginalmediaplan. This is because the traffickers at these organisationswill be processing themediaplansofalargenumberofAgenciesandAdvertisersinaday.Thisrequiresuniformitytocompletetheworkeffectively.Agenciesthatdonotemploytraffickersand rely on account teams to construct the media plan, are responsible forcompletingthetraffickingsheetinfullorfacetherejectionofthedocumentationbytheoutsourcedtrafficker.Incompletemediaplansandtraffickingsheetscontinuetoplayhavocwithdigital campaigndelivery timeswhere such rules are not strictlyenforcedandrejection,duetoincompletesheets,iscommonplace.

Sites

Since each row of the trafficking sheet references a particular Publisherwebsite,thisinformationmustbereplicatedinthecampaignsetup.Thetraffickerbeginsbycreatinganewmediaplanentityandthendrillingintothatentitytoaddanewsitetothemediaplan.AmasterlistofallknownPublishersitesappearsonthescreenandthetraffickermakesthecorrectselection.

As rosy as this may sound, the master-list of all known Publisher sites is acentraliseddatabaseofthethirdpartyadserverthatrequiresregularmaintenance.Auniversal agreement between all traffickers, ad server users and Agencies existsthroughtheenforcementofanti-replicationrulesintheadserversothatonlyoneofeachsiteappearsonthelist(thosesitesthatappeartobeabsentcanbeaddeduponrequesttotheadserversupportteam,butthisprocesscantakeadayormoreandshouldbefactoredintothetimelinesforthecampaignsetup).

Theduplicationwhichrealisticallyappearsinthesedatabases(becauseeachmasterlist of sites database is unique to each third party ad server) is chaotic. Somepopular sites have so many names to represent them it is impossible for thetraffickertodecidewhichonetochoose.Thebestpractice?Topickasinglenameand continue to use that same name every time media is booked on and adstraffickedtothatwebsite.

Whycan’tthesitemaster-listbeuniquetoeachaccount,AgencyorAdvertiser?Ifitwas, therewouldbe far lessclutter, thedatawouldbe far “cleaner”and itwouldtake less time to setup and traffic a campaign. The reason to retain a singleconsolidated list isbecausePublishersalsoaccess the thirdpartyadserver.When

Publishersaccessthethirdpartyadservertheyseealistofalloftheirsitesandtheactivity being run across them by all Agencies and Advertisers. Traffickers thatselectthewrongsitewhensettingupthemediaplan(despitethefactthatthenamemight look right) are inadvertently messing up Publisher-reporting from the adserver interface (as seen from a different user perspective). If the trafficker isunclear about exactly which name to use when setting up the Media plan theyshouldcontactthePublisherforthatsiteandcheckinwiththemtoseewhichsitenames listed in theadserver interfacematchwith themediaboughtand listedontheplan.

A typical media plan contains upwards of one Publisher website, some bigcampaigns feature twentyormore sites andmay also include Publisher networks(which also are selectable as ad server “site” entities when setting up themediaplan).OncethePublisherandsitesareselected,thefoundationsofthemediaplanarecomplete.Thenextstepistocopythetraffickingsheetfurtherwhichwillshowthateachsitehasatleastonecorresponding“placement”.

Placementsaredependentononepieceofinformationaboveallothersandthatisthe dimensions. Without the correct dimensions declared, ads will always bedisplayed incorrectly; theplacementmustbeconsidered tobe theframeof theadandtheadmustalwaysfit intotheframe.Togetthedimensionsoftheplacementright, the site needs tobematchedupwithwhat theCreativeAgencydeveloped.The collective name for all the creative produced to standardised sizes is called“formats” and the correct formats need identifying to proceed with an accuratetraffickingsetup.

3. IdentifytheformatsoftheadsandcreatingtheplacementsWhen Publishers decide to sell ad space they take a long hard look at the theirwebsiteandestablish theopportunity topackage theavailable inventory tosell toAdvertisersofAgencies.Bylimitingthedimensionsoftheadssothatacrossasitetherearethousandsofadspotsavailablewithdimensionsoftypeoneandthousandsavailablebydimensionsoftypetwo,thereareonlytwototalsetsofdimensionsandthereforeonlytwopiecesofcreativethattheCreativeAgencyneedstobuild.Thesetwo sizes, are referenced by the Publisher as its “available formats”.Dimensionshoweverarejustasinglepropertyofwhatdefinesaformat.Themostbasicformatis an adwith dimensions 300 pixelswide by 200 pixels high butmore complexpropertiestothesedimensions(suchastheabilityfortheadtocastthewholescreentooneside)changesthenameoftheformatfromtheindustryrecognised“MPU”tothe industry recognised “sidekick”. It is the standardisation of formats for allPublishers right across the web which has made ad campaigns manageable andscalable.

Adformats

ThewebisfullofadspacebutinmajormarketsliketheUK,theIABhasdefinedmostof thesespacesasbeingcapableofoccupationby“standard formats”whichare represented by various creative dimensions as well as other properties asdiscussed. This standardisation has helped Creative Agencies to build ads faster

(andmoreeffectively).IthasalsohelpedPublisherseasilymonetisetheirinventoryofferings in the wider marketplace. Ads built to the IAB specifications anddimensionsofastandardisedformatcaneasilybetraffickedtoallsitessupportingIABapprovedformats,withthestandardisationescalatingtheoverallprocess.As previously mentioned in Chapter 2- Creative Agencies, the IAB also issuesguidelines for ad creation and acceptance, tomaintain a gooduser experience onPublishersites:

• Standardflashadsforinstanceareadvisednotto“loop”theiranimationmorethanthreetimes(repetitivemessaginginsideasinglepieceofcreative)whichhasbeenshowntodeterthepositiveengagementofauserwithabrand.

• Foradsthatmightcontainvideo,itisadvisedtokeeptheaudioofthevideomutedwhen the ad loads, even if the Publisher permits the video to begin playingthroughwithout it having been initiated by a user (sometimes referred to as an“auto-play”asopposedtoa“userinitiatedvideoplay”)

• Highcontrast colourusageand flashing imagerywhichmayposea risk touserswithphotosensitiveepilepsyhavevanished

• And perhapsmost famously is the strict guideline against the use of JavaScript“pop-ups”and“pop-under’s”whichplaguedtheearlywebandhadasignificantlynegativeimpactonuserexperience.

Although the IAB do not get involved in the ad development process with theCreativeAgencies; thePublishers themselveshaveadopted theseguidelines intotheirownstrictspecificationsso thatadscontaining toomanyloopsfor instancearerejectedbeforethecampaigncangolive.

The use of the term “format” can be understood by considering a “pop-up” adformat.Thepropertiesoftheadarealmostthesame,suchthatwhenthe“pop-up”formatisrunonanysite,everyoneknowswhatitisandhowtogetridofit.Thisis the ad format standardisation that the IAB have been largely responsible for;bringingorder to an immature internet circa2004.Formatnamesareoneof themostimportantconceptstograspbeforethetraffickingprocesscanbeginbecausetheyarethecontentofthemediaplanor“traffickingsheet”.

Banner

Thebannerformator468x60wasverypopularwhenitwasoriginallycoinedasastandardformatbuttoday,asscreenshavegrowninsize,thisformathasbecomeless and less popular because the ad itself is really too small to push a creativemessage and remain impactful, on a computer and remains too large for moreportabledevices.

Leaderboard

Theleaderboardisa728pixelsby90pixelsadformat,andhastakenthelimelightawayfromthebannerformatasitincreasedinpopularity,simplybecausethereismorespacetoworkwith.SomePublishersofferasuperleaderboardformatwhichwill typicallybemorethan900pixelsacrossandissometimesreferredtoasthe

“masthead”.

Skyscraper

The skyscraper (120 pixels by 600 pixels) remains a very popular ad size onPublishersitesalthoughitisbeginningtobeovertakeninpopularitybythelarger“super skyscraper” at 160x600. Once again because screen sizes in non-mobileand non-tablet devices are growing, so the super skyscraper just presents moreinformation.

MPU

TheMPU (mid page unit) is the squarer sized ad format at 300 pixels by 250pixels.Stillaverypopularformat,theMPUisverycommonlyusedasaspacetohostRichmedia video as the square size lends itself nicely to building a videoplayer. MPUs are commonly more expensive spaces as well, as they are morelikely to appear closer to the top of page content as a single unit, than theLeaderboardandSkyscraperformats(whichoftenappearmultipletimesonapageandaremorelikelytodropbelowthefoldofthebrowserwindow).

RoadblockandSkins

ARoadblock is thenamegivenwhenaPublisherpagehasall itsadspots takenover by the sameAdvertiser and ad campaign.Roadblocks consistingof twoormoreadscanbesyncedtogetherusingRichMedia(seeSyncAds),butevenwhenunsynchronised, still offermore impact by repeating the samemessagemultipletimesonthesamepage.Sometimeswebpagesalsocentreallofthecontentofthepage(morecommonwiththemajorPublishers)whichleavethesidesor“gutters”of the page available for advertising. A “skin” is giant image or collection ofimagesthatcovertheseareas:alsoreferredtoasthe“Wallpaper.”

Perhapswehavetakensomethingofadetourfromtheworldoftraffickingintotheuniverseofformatsbuttobringthestoryhome,eachrowofthetraffickingsheetormediaplanwilllistoutaplacementofaparticularsize,(thisisthehousingfortheformatof thecreative itwillbepairedwithanddisplay,when thead loads).ThetraffickerselectsthesiteinthemediaplanintheadserverUIandistakentoanewblankscreenwhich the traffickerwillpopulatewithall theplacementsunder thatsitewhichappearonthemediaplan.

Placement&Package

InsideeachsiteorPublisherentityinthemediaplanthetraffickercreatesaseriesof“placements”whichwillmatch the corresponding ad formats purchased from thePublishersintheplan.ThetraffickercancheckthephysicalPublishercampaignIOtocheckthesetupcorrectlymatcheswithwhatwaspurchasedandwhatappearsontheMediaplan.

Placements belonging to Publishers at themedia plan level in the third party adserver interface, can be grouped together by the trafficker into “Packages” ifrequired, so that some bulk settings can easily be applied to a large number ofplacements.Thesesettingsusuallyrefertothemonetaryvalueofthemediacostofeachoftheplacementswhicharelabelledas“Costs”intheadserver.Toallowfor

added flexibility, packages support different costmodels and are the reasonmosttraffickers decide to keep a pocket calculator handy on their desks during thisportionofthetraffickingprocess.ForthisreasonPackagesaresometimesreferredtoas“costpackages”.

CostPackages

The inputted information in the cost fields in a cost package is for referencepurposesonlyanddoesnotcarrythroughtothePublishersthemselves.Thismeansthatifthetraffickerweretosetthecostsofaplacementandthenamendthemlater-the agreed IOcost and contractwith thePublisher is not changedorbroken: thisinformation is never conveyed to the Publisher themselves. Instead it is used tocalculate extra cost fields in the ad server reporting, based on information likedeliveryandperformance statistics.For instance, thecost fieldcanbeaggregatedacrossallofanAdvertiser’scampaignsrunningoveragivenperiodandthenbrokendownbycountrytocalculatethecostamountbeingspentinallcountriesacrossallcampaignsatasetperiodintime.InthisexampletheAgencyorAdvertisercouldpull this information from the reporting section of the ad server interface (moreinformationinChapter5-Reporting)

Ascostinformationadditionsareextraworkforthetraffickertheyoftenomitthisbutthenfindtheneedto-rebuildreportsmanuallylaterontocalculatethesepost-campaigncosts.Itishighlyrecommendedthereforetoincludethesefiguresduringthetraffickingprocesssothattheadserverdoesthelegworkofcalculation.SomeadserversalsohavetheoptionofdisplayingadservingcostsandAgencymark-upcoststoreflectthetruecostofeachpackagetotheAdvertiser.

Forsomeadservers,placementscannotbecreatedwithoutfirstcreatingapackagetoput them in, and themediaplan screen looks likea compressedversionof thetraffickingsheet.Here,placementsarehiddeninsidethepackagesasfoldersanditis thepackageswhichareassigned to thesites.Oncepackagesettingsarecreatedthey are rarely changed, so packages are often hidden to the trafficker in thetraffickingscreensaftercreation(toremoveclutteronthescreen).

Once Publishers, Packages and Placements have been setup in the ad serverinterfacebythetrafficker,thenextstepistomatchupthecreativewiththeads.

4. AssignthecreativeandsettingrotationcontrolsInparalleltothetraffickersettingupthecampaignintheadservinginterface,theCreativeAgencyconnectstotheadserverviathesuppliedcomponentsintheflashauthoringenvironment. It ishere theCreativeAgencycanname the files that thedevelopersupplies.Asuccessfuluploadwill returnaconfirmationmessage insideFlashwithauniqueidentifierforeachuploadedfile.

These identifiers can thenbe sent on to the trafficker,who is able to retrieve thecreative.InDG/Mediamind for instance these identifiersarecalled“AdID’s”.AdID’s themselves are a poor indicator for what a creative actually looked like orwhereitcamefrom,sothefile-namesuppliedbythedeveloperisquitesignificant.Despite the fact that thecreativedeveloperwill supply thecreativewithacertain

name, this name can be altered by the trafficker at the start of the traffickingprocess,soastofilterdowneffectivelytoreportinglateron.NamingConventions

Creativenaming conventions (as in the filenamesgiven to theFlash creative andGIFs)canmeanthedifferencebetweenefficientandinefficienttrafficking.Creativenaming conventions should be agreed between the Creative Agency and MediaAgencybefore theadsarecreated,so thatwhentheyareuploaded, thecontextofthefilessuppliedisperfectlyclear.

Someinformationthatthenamecouldbemadeupofmightinclude;

• thecreationdate

• abriefdescriptionofthecreative

• thecampaignname

• theAdvertisername

• thenameoftheCreativeAgencydeveloper

• information about the intended location of the final ad (if known) such asplacementorsiteinformation

• an indicator of some of the functionality of the creative such as “expand” for aRichmediaexpandablead

• anindicatorofthead’sintendKPI(suchas“servicesignup”or“brandawareness”)

Although most of these data points reference things that are added later in thecampaign,suchinformationreallyhelpsthetraffickerworkinginthethirdpartyadservertomatchtheadtothecorrectsite,adspotandcreativefolder.

Downtheroadthecreativenamingconventionisalsoalifesaverinreportingwherethe“creativename”fieldcanbeaddedtoanykindofreportrevealinginformationabout an ad that had been misplaced, or even a creative theme that was moresuccessful thanothers. Inmanyinstances, thecreativenamecanbeusedasquickfix where perhaps a third party ad server has reporting limitations (the namingconvention is used to ‘carry’ creative information through to data analysts toimprovecampaignoptimization).

Todemonstratetheeffectivenessofthis,acreativenamecanbetransformedfrom:|Creative3_1108_dave

|to

|Unilever_Lynx_Axe_brandawareness_summer2015_akqa_themealpha

Any analyst reading a report containing any information about this creative willnowatleasthaveanindicatorabouttheAdvertiserandbranditisreferringtoowingtothereferencesinthenamingconvention.

Ad&Creative

TouseCreativeinacampaignitmustturnedintotheentityreferredtoasan“Ad”.

Someadserversdothisautomatically,othersrequireacoupleofextrastepsonthepartofthetraffickertocreatetheads.Thiswillbethefirsttimethatthetraffickerhas exposure to how the adwill look andbehave, and as such, the trafficker canconductsomebasictestingtoensurethattheCreativeAgencyhasbuilt theadstotherequirementsofthebrief.

AutomatedQA

Some ad servers automatically process the creative when it arrives in from theCreativeAgency. These checks forQualityAssurance (QA) check a short list ofcreative properties that should be present in all creative supplied, so these arestandardisedpropertiesandnotchecksforuniquefunctionality.CheckswillincludelocatingthepresenceoftheclickTag,checkingthecreativeformaliciouscodeandensuring that the click portions of the ad actually clickwhen such user action istaken.OnceadshavepassedthisautomatedQAtheycanbeusedforfurthertestingby the trafficker.As a sidenote: somead servers arenotusingmachines for thisprocessbuthireteamsofindividualstomanuallycheckeachandeverycreativethatpasses into the ad server interface, as such the initial QA can take up to aroundtwentyfourhoursbeforetherestofthetraffickingprocesscanbeundertaken.

AdPreview

The newly created ad can be viewed in the ad server interface as it would bedisplayedwhenlive.Thispreviewareacomesclosetosimulatinghowtheadserverwill display that ad once it is live on the Publisherwebsite. The preview screenallows the trafficker to visualise the finished ad for the first time, check thatinteractionsarefiring(seeChapter3-Richmedia)and that thead isbehavingasexpected.TheadcanthenbeviewedagainstthebackdropofalivePublisherpage.Finally, a copy of the preview can be sent to any other Advertiser or Agencycontacts for final sign off (it is not common for the sign off to be a mandatoryprocess as this has proven to slow down the trafficking process). This sharedpreview produces a URL which often remains live long after the campaign hasendedso that it canbeusedasa reference towhat thead looked like.This sameURLcanbesuppliedtotheCreativeAgencywhowouldstoreitfortheirownon-sitecreativearchiverecordssothatthecreativecouldbere-viewedasanadatanypointinthefuture.

Oncethetraffickerissatisfiedthattheadisbehavingasexpecteditistimetoaddalocationfortheclick.

The main difference between ‘Creative’ and ‘Ad’ is that a usable ad has acorrespondingURLsothatwhentheuserclicksontheadwhenitisonaPublisherpage, the user we are redirected through to the destination designated by thetrafficker.

Beforemovingforwardsitisworthrestartingthisclaim:Acompletedor“finalised”ad (as it is referred to in theAtlas third party ad server) is the marriage of thecreativeandtheclickthroughURL.ThetraffickerofcoursewilltaketheURLfromtheonesuppliedonthemediaplan.Thisiswhycreativeisnotassigneddirectlytoplacements,sincecreativedoesnotcarryaclick-throughURL.

The recommendation for any Advertiser is that the landing page (that theclickthroughURLdirects to), should in somewaybecontextually relevant to theuser.Some traffickersworkwithAdvertiserwebsitemanagement teams toensurethattheclickthroughURLisdynamicallypopulatedwithavaluethattheusermayhave inputted into thead (suchaszipcode).ThecorrespondingAdvertiser site isthensetuptoprocessthisvalueandamendthelandingpageinsomeway,suchasby using the zip code to conduct a search and then displaying the search resultsspecifictothatzipcodeasthelandingpage.Thesemorerelevantexperiencesareacampaign considerationwhichwill undoubtedly show improvements in campaignperformanceifsetupcorrectly.

Addingtheclick-throughURL:clickTag

RemembertheclickTagconceptfromChapter2-CreativeAgency?Asareminder,allcreativeneedstobeembeddedwiththecorrectclickTagbytheCreativeAgencytobeapprovedforuseinsidethethirdpartyadserverforthecampaign.

The clickTag is actually a very simple piece of technology that prevents theCreativeAgencyfrom“hard-coding”theclickthroughURLintothecreativeitself.Now,whenauserclicksonanad,theclickTagintheflashfilequeriestheadserver,sending the user to the complete URL inputted by the trafficker rather than anyaddressinputtedbythecreativedeveloper.

The purpose of removing the power of the click-throughURL from the creativedeveloper and placing it instead with the Media Agency trafficker means that,shouldtheclickthroughURLchangeduringthecampaign,theSWFfileproducedbytheCreativeAgencydoesnotneedtobeadaptedand“re-published”.

Thistime-savingfunctionalityisoneofthereasonsaclickTagissoimportantandalsomeans that creative can be developed and published even if the final click-throughURLisnotknownuntilthedaybeforecampaignlaunch.ItallowsfortheflexibilityofchangingaURLmidwaythroughcampaign.TheotherprimaryreasonfortheclickTagisthatwithoutit,the3dpartyadserverwouldnotrecorda“click”whentheuserclicksonthead-moreonthisinChapter5-Reporting.

302Redirects

The responsedelivered from thead serverback to thebrowser commonlycarriestheHTTP status code of 302 (most readerswill bemuchmore familiarwith thecode404fromthesamefamilyindicatingamissingwebpage).Bycomparison302isusedtoinformtheuser’sbrowserthatthewebpagehasbeentemporarilymoved.Thisinstructionallowstheadservertousethe302calltologanimpressiononthecountingmodulecomponentoftheadserverwhileservingthecreativeor1x1fromthecontentserverorCDN(seeChapter1-AdServing).

Addingmultipleclick-throughURLs

Asaside-note,thebigthirdpartyadserversconsiderastandardflashadtobeonethat contains a singlepieceof animationor flash file,with a single click-throughURLtocorrespondwiththeclick-ableareas.Insomeinstancesthereare:

1. multiplelocationswithinanadtoclick,directedtothesamepage(which

isstillconsideredasingleclick-throughURL)

2. multiple locations within an ad to click, directed to different pages(multipleclick-throughURLs)

Inthefirstinstancealltheclickswillbeattributedtothesame“clicks”bucketwhenreporting on a single ad, to break out which click-location inside the ad wasgenerating themostclicks thesecondscenariohas tobeadopted(ii).For (ii) richmedia technology is usually utilised so that the second click is actually a custominteractionandthesupportingreporteitherreports“other”clickthroughsascustominteractions or redefines them in a custom report as an additional set of clicksalongsidethesamead.

Alongsidetheclick-throughURL,itisworththetraffickeraddingextracontextualinformation to theadso that theperformanceof thatadandotherswith thesamepropertiescanbegeneralisedinthereportinginterfacelateron.

Adclassifications/CustomCategoriesandsearching

Byconvertingacreativeentity intoan“ad”entity,newoptionsbecomeavailable,options which are available to all ads said to sit at the “ad-level” unique to theAdvertiseraccount.Oneofthekeyoptionsavailableattheadlevelistheabilitytoaddfieldstothedescriptionofthe‘ad’intheadserver.Oneofthemostusefulofthesefieldscommontothebigadserversiscalledthe“adclassification”field.Thismaynotbeasinglefieldbutmultiplefields.Eachadclassificationfieldisdesignedtocarryanewpieceofinformationabouttheadentity,withthatad,onitsjourneyfromcreationthroughtoreporting.

Forinstance“adclassificationfield1”maycontainacreativemessagetype.IfIama big automobile Advertiser my possible creative message types might rangebetween 1)Model awareness 2)Model interest 3)Model detail. Each ad createdfromacreativecanbeclassifiedintooneofthethreetypes.Itisuptothetraffickerduring the trafficking process therefore to classify the ads by appending the textwritten in “ad classification field 1” to suit one of these options.When the timecomestoreportontheperformanceoftheadsintheadserveranalyticsinterfaceitwill be clear by displaying this field,which creativemessagewasmost effectiveregardlessofotherfactorssuchastheadsize,thePublisheroriftheadcontainedanyvideoforexample.

Currentadservingtechnologyoffersadclassificationfieldsasopentextfieldsthatthetraffickerisinchargeofupdating.Itisimportanttoofferthisflexibilitywhichhasbeenusedcreativelybymanyoftheindustry’sproblemsolverstoimprovethequalityofreporting.

Otherfieldswhichsitattheadlevelincludethelengthandwidthofthecreative.Inolderadservers thesefieldsalsoneed tobemanuallyupdatedby the trafficker toreflect the correct size of the creative but some accurately detect the size andpopulate these fields automatically.Aswith the ad classification field, the lengthandwidthoftheadcanbeaddedtoreportstodetermineifadsofaparticularsizearebetterperformers.

Thereareawholeselectionofotherfieldswhichcanbepopulatedmanuallyattheadlevelanditisthisdatainputtingtaskforeachadthatisresponsiblefortakingupsomuchofthetrafficker’stime.

Nowthattheadhasbeenwelldescribedintheadserverinterface,anymemberoftheMediaAgencyteamthatmightgointotheinterfacewillbeshowninformationaboutthatadandfinditfareasiertoidentify.ThenextstepwilladdresshowtheadwillrelatetothePublishersonthemediaplan.Thisrestsonensuringthatthereareenoughadstogoround.

AdDuplication

In order to keep campaign costs down a minimum amount of creative will beproducedbytheCreativeAgency;afterall,whatisthepointinpayingtheAgencyforacreative foreachPublisheron themediaplan, if thead tobedisplayedwilllookexactlythesameoneverysite?

Insuchaninstancethecreativeisfirsttransformedintoanadentitybythetraffickerand then replicated so that each ad assigned to each Publisher’s placements isunique(eveniftheonlyvisuallyuniquethingaboutitistheadID,generateduniqueforeachadcreated).UniqueadassignmentisimportantbecauseiftheuniqueadIDis analysed by itself later on for its performance, the unique properties of itsassignmentwillrevealwhereitsatinthemediaplanandthereforewhichPublishertheperformancecouldbeattributedto.

Figure3.2.2makingfulluseofanexpensiveCreative-Agencyproducedcreativeinaspecific format.TheCreative issimplyduplicatedasanAdentity. In this imageeachadisthenuniquetoeachPlacement.

Inturnthetraffickerisrequiredtoassigneachadtoasingleormultipleplacementsbasedon the informationsuppliedby theCreativeAgency(acreativenameoradID)andthemediaplaninfrontofthem.

ItisnotunheardoffortheCreativeAgencytosidestepanyconnectiontothethirdpartyadserverentirelyandjustsendthecreativeassetsdirectlythroughthetothetrafficker.ThetraffickercanlogintotheadserverinterfaceanduploadthesefilestoanareawithintheAdvertiser’s“account”calledthe“creativelibrary”atanytime.

ThisisthesamelocationthattheCreativeAgencywoulduploadtheirfilestoiftheywere using the automated upload facility available in the flash-based authoringcomponent.

Figure3.2.3TheCreativelibraryispresentattheAdvertiserlevel.Creativecanbearchived,arrangedandstoredinthecreativelibraryfacilityoftheadserver.Intheplacementcreationandadassignmentscreensthetraffickerchoosethecreativeoftherightdimensionsandmarryitwithacorrespondingplacement.

ReadyingtheAdsforAssignment

Once ad entities have been created and the classification fields have beencompleted, the ads can be assigned to or coupled with a single or multipleplacements. To enable the process of assignment, the placement needs to meetmandatoryrequirements:

• Theadhasacorrespondingclick-throughURL

• Theadisthesamedimensionsastheplacementrequiresittobe(therearecaveatstothisforRichMediaads)

• Theadhasafall-backdisplayoption,incasetheuser’sbrowsercannotdisplayaflashfile.

The suggestion of a fall-backoption is a new concept, but oneworthy of furtherexplanation:

BackupGif

Standard ad serving and Rich Media offer a fail-safe, so that usermachinesordevicesthatdonothaveaflashplayerinstalled(orhaveanolder version than the one required to play the creative) still see an ad.Thefailsafeadisa“backupgif”andmustbesuppliedwheneveraflashfile is supplied by theCreativeAgency. Since the third party ad servermustalwaysservesomethingifanadrequestisreceivedbytheserver,itwillshowthebackupgifwhentheconditionsofabsenttechnologyonthelocalusermachineorbrowseraremet (suchas theabsenceof the flashplayerplugin).

Servingbackupgifsisamorerelieduponfailsafeinrecenttimeswiththelack of flash support for mobile devices. In instances where a mobile-specific campaign is not setup, and a user browsers to a Publisher siteintendedforaPCuser;aniPadforinstance,failstoreceivetheflashfileand a backup gif is displayed instead. Such failsafe or fall backtechnologiesaretoplayawiderroleinthedevelopmentofadsformobileandflashalternativessuchasHTML5.

When the campaign goes live the placements should rarely be showingthebackupgif since thewhole point is to be showing the best possibletypeofcreativetotheenduser.Itispossibletomonitortheseoccurrencesfrom within the third party ad server reporting interface; occurrencesmightbeconfinedtoasingleplacementorPublisherwhichwouldrequirefurther investigation in conjunction with the Publisher ad ops teamdirectly.

PossiblePlacementAssignments

Asarule,theassignmentofadstoplacementsisnotexclusivetothesetupof‘anad(withclick throughURL)withbackup image’.Thereareother setups thatcanbeassigned to Placements, the backup image can be assigned on its ownwithout abetteradinplaceandthecampaigncanstillgolive.

Figure3.2.4SofarwehavediscussedthesetupshownbelowforPlacementA(thead toplacementassignment forAd1).PlacementBshows the setup forchannelsand sites requiringmorea trackingonly solutionwithout full thirdparty serving.Thiswillbecoveredofflaterinthischapterunder‘Trackingbutnotserving’

TheassignmentshownforPlacementCinFigure3.2.4isanabsolutelastresortforany trafficker to have to do; to allow them tomove forwardwith the traffickingprocesstheymaybeforcedtoonlyassignabackupimagetothePlacementwithnocorresponding Flash creative as an ad. In this instance the Placementwould stillrequirethatthebackupgifbesuppliedwithaclick-throughURL.Theresultisthattheuserwill seeonly themostbasicbackupgif.ThiscanoccurwhenaCreativeAgencyfailstosupplytheFlashfiletothecorrectdimensionsintimebutabackupimage(perhapsfromaprevious campaign) is available for the trafficker to selectfrominsidetheadserverinterface.

AdAssignment

InFigure3.2.4(assignedheretoPlacementA),itisnowmandatoryforjustasingleadtobeassignedtoaplacement.Indeedtheflexibilityisbuiltintotheadservertoallowformultipleadstobeassignedtoasingleplacementandviceversa,saving

thetraffickeralotoftime:

Multipleadstoasingleplacement

Aplacementdisplayswithinit,anadofthesamedimensionsbutitisnotalways the same ad that is displayed.Refresh a Publisher page and thesameplacementmight bedisplaying another adby the sameAdvertiserwith a different piece of creative. This is achieved through themanipulationofthecontrolsontheplacementleveltochangewhichadisdisplayed inside the placement. By having multiple ads assigned to aplacementthetraffickeronlyneedstosendasingletagtothePublisherattheendofthetraffickingprocess.Thismightbesupportingawholeheapofadsbehindthescenes,controlledbythetraffickerinthethirdpartyadserverinterface.

Asingleadtomultipleplacements

In some instances the same creative needs to be used across multiplePublishers but few ad servers are proficient in allowing traffickers toapply the same creative to multiple placements. Instead the ad serverforcesthetraffickertocreateacopyorduplicateadbeforeassigningittosubsequent placements with the same creative dimensions, even acrossmultiplePublishers.Thedownsidetothislimitationisthatany“changes”made to thecreative (suchas thecreationof anewcreativemessageorthebuildingofanewcreativeentirely)meancreatingawholenewadandun-assigningtheoldadacrossmultipleinstances.

Asingleadtoasingleplacement

Asimplesoundingassignment,butonethatatscalecanspelltroubleforthe trafficker,whomayhave to spendhoursmaking the assignments inaccordancewiththesuppliedmediaplan.

To serve the interests of time-saving, ad servers supply tools which allow thetraffickertoperformthese(laboriousonalargescale)tasksinashortspaceoftimeacrosshundreds, thousandsorhundredsof thousandsofadassignments (seebulktraffickingsectionlaterinthissection)

Alotofgroundhasbeencovered,sohereisasummaryofthetraffickingprocesssofar.

a. ThetraffickerhasloggedintothethirdpartyadserverinterfaceandnavigatedfromtheAdvertiserleveldowntothecampaignlevel

b. The traffickerhas setup themediaplanat this levelby selecting thecorrectsites

c. Thetraffickerhascreatedplacementsforeachofthesites

d. The trafficker has found the creative supplied by the Creative Agency andcreatedadsfromthemwheretheydonotalreadyexistasads

e. Thetraffickerhasaddedinformationtotheadsincludingclick-throughURLs

f. The trafficker has assigned these ads with backup images to the correctplacements,followingtherequirementsofthemediaplan

Figure3.2.5Thebasictraffickingworkflowundertakenbythetraffickerinsidetheinterfaceofthethirdpartyadserver.

Beforemovingontothelaststepinthebasictraffickingprocess,asshowninFigure3.2.4(trafficadtagstoPublishers),thetraffickerhastheoptionofadjustingaseriesof the ad controls, after which it is recommended to conduct a “QA” (qualityassurance)ortesttheplacementsas“tags”beforetheyaresenttothePublisher.

AdjustingtheAdControls

AdControls are essential for placements that havemore than one ad assigned tothem.Theplacementwillnotbedisplayingalloftheassignedadsatoncesotheymustbeshowninturn.Adcontrolsareasetoflogicalrulesthattheadservercanfollowwhenarequestisreceivedbytheservertodisplayanad.

FrequencyCapping

Bylimitingthenumberoftimesauserseesaparticularadthetraffickerensuresthattheuserisnotover-exposedtothesameimages.Whencombinedwith“adrotation”(definitionshownbelow)thisensuresthatusersonlyseeaparticularcreativeasetnumberoftimes(calleda“frequency”)overaparticularperiodoftime.Frequencycappingcanbeappliedatdifferentlevels,forthepurposesofthissectionitcanbeappliedattheadlevelbutitcanalsobeappliedtothecollectionofadsonasingleplacement(calledadeliverygroupinDG/Mediamind)orevenacrossanaudience(asintheAtlasAdserver).AlthoughusuallyseenasthefeatureofaPublisheradserver, frequency capping on the third party ad server allows for the controlleddelivery of more expensive formats such as rich media ads, balancing formatdelivery to best manage advertising budget while still delivering impactfulexecutions.

AdRotation

TheAdvertiseradserverallowsthe trafficker toensure that theuser isnotseeingexactly thesame imageallof the time.Byputtingseveral imagesorcreative intorotationonthesameplacement,theuserhasthepotentialtobeexposedtodifferentimages and fresh content each time the same placement loads. The followingdescribesdifferenttypesofadrotation:

Sequencing

Sequencing is the ability to storyboarda seriesof ads so thatusers thathave seen the first ad in the sequence are exposed to the second in aseries. Sometimes this is restricted to ads of the same dimensions andformat, but ad server improvements are allowing formore flexibility inthisareainthefuture.Sequencingusesinformationstoredintheadserveragainst theuser’s cookie todeterminewhichad to showbasedonwhatadshavebeendeliveredtothembefore.Sequencingisrecommendedfor

Advertiserwholiketostoryboardtheirads,takingtheuseronajourneythrougheffectivecreativemessaging.

Weighting

Adsinrotationcanbeweightedsothatthosewithahigherweightingareseen more often. Weighting is expressed as a percentage and the totalweighting for all of the ads on a placement must always equal onehundred per cent. Weighting is particularly useful when optimizing acampaigntoensurebetterperformingadsplaymoreoften.

AdPrioritization

Forsomeadserversweightingcanbeexpressedasawholenumberratherthan as a percentage. The lower the number, the more often it will beshown. Typically ad prioritisation exists within the band of one to onehundred.This is especiallyusefulwhen ads are commonly takenout ofrotationandnotalways replaced, ifweightingwasused in this situationthenthenewadweightswouldneedtobecalculatedeachtime.

TimeBased

A day-part analysis of a campaign (See Chapter 5 -Advanced reports)willrevealtemporaltrendsaboutbestperformingadsbrokendownhourbyhour.Thetimebasedsettingallowscertainadstobeshownatcertaintimesof theday.Thiscanhelp significantly inavoiding“wastage” (adsbeingdeliveredtoanirrelevantaudience)bysettinganadtoserveduringquieterservingperiodsthatperhapscostslesstoserve(suchasaGIForastandard ad showing at 4am instead of a rich media ad). Also,performancemay improve on different days such as at theweekend sothis feature can be useful then. It is important to make a carefulconsideration about performancewhen adjusting day part delivery.Adsthat fail to receive clicks at night may still be providing a hiddencontributiontofinalsale(seeChapter4-Conversions).

PostCapGif

Someadrotationsarebalanced:inwhich,ifoneadisnotshown,anotherisshownin the same rotation (such aswithWeighting) but others like Frequency cappingpresentsituationswhereausercouldhaveseenthepresentedadsinafiniteseries.AdServersofferthepresenceofan“otherbucket”inthissituation(alsopresentinTargeting)where, if the ad request ismadeby thePublisher page and all the adshave been shown in the series, then a backup image is shown. This image istypicallyaGIFandisalwaysrequiredinsuchsetups.Someadserversdonotforcethe Advertiser or Agency to upload a single backup or default image for eachplacement, but instead allow them to upload a universal backup image at theAdvertiser level (a non-campaign-specific piece of creative) which is just shownwhenevertheconditionsaremet.

Thetwolevelsofcontrol

Adcontrolscontrolonlywhatadisdisplayedintheplacement.ThePublisheralso

hassuchcontrols in theirPublisheradserverasexploredinChapter2,but theadopsteamareusingthesecontrolstochoosewhichadvertiser’splacementshouldbedisplayed.Thisisancomplexconceptbutvitaltotheunderstandingofaddelivery,these are facts that often need to be re-stated during a troubleshooting processbecausetheterminologyusedtodescribetheMediaAgencytraffickersadcontrolsand the Publisher ad ops team’s controls are the same. For instance there are aminimum of two frequency caps that can be set for a placement; one in thePublisheradserverandoneinthethirdpartyadserver.

Figure3.2.6IntheaboveexampleboththeAdvertiseradserverandthePublisheradserverareusingfrequencycappingtocontrolwhichadwilleventuallygetseenby the end user. TheAdvertiserAd server can only controlwhat goes on in thePlacement (referred to in this diagram as a “tag”) that it delivers to Publisher adserver.ThePublisheradservercannotchangethesettinginsidethattagbutitcanchoose not to serve that tag and serve another instead. In this example it doesn’tmatterwhat the trafficker on theAdvertiser ad server included in the ad controlsbecausethedefaultcreativewillnevergetseenaslongasthefrequencycapinthePublisheradserverisshorterthanitisontheAdvertiser’sside.

The ad control interface in the third party ad server is where the rules for addelivery aremanaged so on top of these options is the ability to setwhich ad isshowntowhichspecifictypeofaudience.

AsaudiencesarenotarequirementforthebasicsetupofadisplayadcampaignithasbeenomittedfromthissectionandaddedtoChapter4-Audiences.

Afterthetraffickerhassettheadcontrols;thead,theplacementandthecompletedsettingsarereadytobeusedas“tags”.

5. TraffickingtheTagstothePublishersAdTagsandTagFormats

Thecreativehasbeenturnedintoanadandtheadassignedtoaplacement,nowthe“tags” can be generated and sent to the Publisher. The “ad tags” themselves areshortpiecesofcodethatsimplyreferencethelocationoftheadonthethirdpartyadserver.ThiscodeismostcommonlygeneratedasJavaScriptalthoughitcanalsobegenerated as an iframe (more HTML friendly script) or for added flexibility,

standardHTML.

Tagstypicallyhaveareferencenumberlocatedinsidethemwhichnormallyreflectsthe placement or Ad ID and are a fast track method to identify the tag fortroubleshooting purposes, should problems occur later on. (The individualtroubleshootingwouldsimplytaketheIDandinputit intoasearchboxinsidetheadserverinterfacetolocatetheplacementregardlessofwhereitwaslocatedwithinthehierarchy).

An ad tag can be generated for each placement on themedia planwithin the adserverinterfaceoncetheplacementhasbeenpopulatedwithads.

Figure3.2.7Thiscodewillrendertheadinsidetheplacementforwhichithasbeengeneratedwhenitispassedontoawebpageandloadedfromwithinabrowser.Itispossibletorunacheckonthetagonalocalmachinebygeneratingtheadtagandthenopeningtheresultingplaintextfileinawebbrowser.Bothrenderedimageandclickportionsoftheplacementarerepresentedinthiscode.

Flighting

Onceadtagshavebeencreated,thetraffickerworkinginthethirdpartyadserverneedstogetthetagstothePublisheradserverandthePublisheroperationsteam.Few technology companies offer an integration between theAdvertiser ad serverandaPublisheradserver(withthemostnotableexceptionbeingDoubleClick)soadtagscreatedintheAdvertiser-sidetechnologydonotautomaticallyappearinsidethe Publisher technology. The ad tag code is therefore delivered to the PublisheroperationsteamviaemailandtheoperationsteamreceivetheadtagsandputthemintothePublisheradserverinterface.

The third party ad server interface offers a screen at the end of the traffickingprocess to attach the ad tag to an email sent from the ad server directly. Thetrafficker inputs the contact details (email address) of thePublisher adoperationsteam and may attach a note, more code or other files to the email containinginstructions for implementingand testing thead tag.The instructionsalsocontainreference informationsuchas theplacementdimensionsandnamingconventions,sothatthereisnoconfusionabouttheadtagthatisbeingsent.

Once the trafficker sends the tag, the third party ad server keeps a record of thesend. These records are utilised by large trafficking teams on large scale displaycampaigns to determine which tags have already been issued. By comparingreportingdatatotheserecordslateronitispossibletodeterminewhichPublishershaveneglectedtoimplementthesenttagsontime.

Uponsendingthetags,thePublisheradoperationteamreceivestheadtagtotheirinboxandwillcopyandpastetheadtagcodeintothetechnologyontheirside(thePublisheradserver).Thiswillsettheadtogolivein-linewiththeinformationtheyhavereceivedontheIOfromtheplanner/buyerattheMediaAgencyorAdvertiser.This informationwillmatch the line itemon theMediaAgencies in-housemediaplan.Iftheadtagisnotreceived;someadserversofferPublisherstheflexibilitytologintothethirdpartyadserverdirectlytore-downloadtheadtagsfortheirsites.In doing so the Publisher ad ops team will have third party ad server logincredentials,withdifferentpermissionstothetrafficker.

ThoroughlyCheckinganAdislive

Once an ad tag is put into a Publisher ad server and is live on a Publisher site,traffickersarekeentocheckhowtheadslookintheliveenvironmentonthesite.Some Publishers return screenshots or send over theURL of a page back to thetraffickerwhere the live ad can be seen. To check a correct implementation it isadvised that the trafficker load the Publisher page in question while running anHTTPtracerprogramontheirlocalmachine.ProgramslikeFiddlerorHTTPWatcharesuitabletorevealifthecorrecttagisloadingontothepage(bycross-referencingthePlacementID)andthatboththeloadingoftheimage(orflashfile)anduseofthe click-through URL are generating real-time server calls. Server calls whichmake requests referencing the placement or ad ID usually clarify that the rightplacementisonthepage.IfthewrongplacementIDhasbeenimplementedontothepagetheadserverreportingwillnotreportbackaccurateresultsandthePublisheradopsteamshouldfixthetagonthepagewiththecorrectplacementID.

Inadditiontorunningthesemanualchecks,thetraffickercanusethethirdpartyadservertoverifyactualdeliverytotherightuseraudiencesuchasusersinthecorrectgeography. This is called Verification and further information about this and themonitoringofdeliveryinformationisavailableinthethirdpartyadserverreportinginterface(seeChapter5).

OncethecorrecttagsareliveonthePublisherpagesthecampaignsetupiscompleteandthetraffickerwillaccessthereportinginterfaceonthethirdpartyadservertobegintheprocessofOptimization(seeChapter6).

7. HowdoesatraffickermanagecampaignswithhundredsofAds?BulktraffickingthroughExceltemplates

With thousands of ads, placements, sites and many assignments, un-assignmentsand amendments to be made, bulk trafficking tools are indispensable for thetrafficker. The trafficker can download the ad server’s trafficking sheet from themediaplan level toa localcomputerandfill it in rowbyrowbeforeuploading it

back to the ad server via an error checking and queuing system. The traffickingsheetisdesignedsothatitcanbereaddirectlybyamachine.Atanytimethefullmediaplancanbeexportedfromtheadserverinterfaceandappendedbeforebeingre-uploaded so that bulk changes can be made. The ad server makes a log ofchangesatscalebyspecifictraffickerssothatifmultipletraffickersareworkingonthe samemedia plan at any time, there is little confusion about which traffickermadethelastchangestothemediaplan.BulktraffickingwithExcelastheUI

SometoolsliketheAtlas ‘MediaAdd-in’meanthetraffickercancompletealmostevery taskof the traffickingprocess from insideExcel includingexploring theadserver hierarchy. Like the Flash components supplied by the ad server to theCreative Agency developers; an additional piece of software can be downloadedfromtheadserverandinstalledlocallyonthetrafficker’smachinetoenabletheuseof thebulksoftware.Essentiallyaskinorpluginaccesseddirectly fromtheexcelinterface,permitstheexecutionofvariousoperationswhentraffickercredentialsareverified.Thegain inoperational scaling isveryapparentbut thedownside is thatthe interface of an excel spread sheet is a tenuous replacement for the usabledesigns of ad serving interfaces. As of publication, the marriage of a true bulktraffickingsolutionandanadservinginterfacehasyet toannouncetheireffectiveengagementbeyondaspreadsheet-styledesign.

AutomatedtraffickingsetupviaadserverAPI

ItmayseemstrangethatitispossibletotransformExcelintoauserinterfacefortheadserver,butaslongasaconnectionisestablishedwiththecorrectusercredentials,excel andotherprogramscanbeused to replicate theactual ad serving interface.Thisismadepossibleduetotheadserver“traffickingAPI”.

AnAPI(orApplicationProgrammingInterface)allowsanad,webpageorprogramat a distance (external to the server in question) to request only a select piece ofinformationfromtheserverandhavethatinformationreturnedinordertofulfillafunction.

Services that offer an API quickly ensure the propagation of their informationaroundtheinternetandsoonbecomeindispensabletotheworkingsoftheweb.TakeforinstanceFacebookwhichallowsdevelopers tosignup to theirAPI to retrieveinformationaboutusers(withtheirconsent).OrGoogleMapswhichallowsanyonewhohasrequestedafreedeveloper‘key’(atypeofpassword)todisplayaGooglemapbasedonthelongitudeandlatitudeinformationsupplied,passingthedeveloperkeyinarequesttotheserverdirectly.

Figure 3.2.7.1With just read access anAd or any other application canmake arequesttothebackenddatabaseviathegatewayoftheAPI.TheRequestforMapisjoinedby theAPIkeywhichgrantsaccess to thebackenddatabase.Thebackenddatabase processes the request and if theAPI command is correct and the key isaccepted,itreturnstherequireddetailswhichinthisinstancearethedetailsofwhatthemap looks likewhicharesubsequentlyshown in thead.WithwriteaccessanapplicationsuchasaMediaAgencyplanningtoolcanissueacommandto theadserverdatabaseviatheAPIsuchastohaveanewplacementtrafficked.IftheAPIkeyandthecommandareacceptedthenthedatabasemakestheamendmenttothesetupof the ad campaign and returns a confirmation to theplanning tool that thetaskhasbeencompleted.

AsFigure3.2.7.1 shows,withanad server traffickingAPI, theexternalprogramcan issuecommands to thead server fromadistance to request that anactionbeexecuted or for the submission of information into a particular section of thehierarchy.Thistwo-wayfunctionalityofbothaffectingsetupintheadservermediaplanandreturningconfirmationofthechangesisastepfurtherthanthe“read-only”accessofpublicAPIsofferedbyGoogleandFacebookinthepreviousexamples.

The ad server trafficking API supports a multitude of commands and can returnusefulpiecesofinformationaboutthecampaignsetup.Thisishowitispossibleforaremotecopyofexceltomakerequeststotheadserverandfetchacurrentcopyofthemediaplanwithoutrequiringthetrafficker to leavethetraffickinginterfaceoftheexcelenvironment.Inthefuture,strongadserverAPIssupportingallfunctionsof the ad server will permit application developers to create simplified userinterfaces and dashboards which could be installed on mobile devices allowingtraffickerstotraffic,monitorandcontrolcampaignsonthemove.ForthetimebeingtraffickingAPIsarepredominantlyutilizedby in-Agencyplanning toolsand toolslikeiDesk(astouchedoninChapter2-AgencyPlanning).ThemediaplancreatediniDeskmakestherelevantcallstotheadserverusingit’sAPIandentitiessuchastheplacementscanbecreatedfromafar.Inturntheadservercanreturninformationsuchasalistofavailable“site”entitiessothatiDeskcanselectoneandreturnthatrequestbacktotheadserver:traffickingatadistance.

Other API capabilities exist to support the ad server capabilities such as forreporting.MoredetailscanbefoundinChapter5-Reporting.

TheJavaScriptcarriagecapability

Anotherofthemorecommonadvancedcapabilitiesofthestandarddisplayadisitscapability to launch additional code on the condition of certain JavaScript basedactivities resulting from real-time user action. Such activity may include theJavaScript parameter of “Unload”; whereby on the condition of the ad beingabandoned(suchasbyleavingthepageorclosingthebrowser)theJavaScriptcaninturnlaunchanon-screensurvey.Thisparticularexampleisthedeliveryofwhatisreferredtoas“SurveyVendorcode”,buttherearemanyactionsthatcanbetaken,allowingthetraffickertomanipulatetheJavaScriptcapabilitiesfromthecomfortoftheadserverinterface.

Surveyvendorcodeandthirdpartyscripts-inads

Survey Vendors emerged in the world of ad serving technology because trueAdvertising effectiveness is a subjective concept. Although an ad may haveperformedwellaccordingtoadserverreporting(receivingclicks, interactionsandevenconversions(seeChapter4) it isstillnotonehundredpercentclear ifuserswereinfluencedbytheaditselftoperformtheseactions.Itisentirelypossiblethatiftheadwasshownintherightplaceattherighttime,thatthatadmerelyactsasaconvenientpathwayfortheusertocompleteanactiontheyhadeveryintentionofdoinganyway.

Furthermoreitisnotclearifthemessageinsidethecreativeitselfhadanimpactonthe user. Since we cannot see what is going on in a user’s head, collecting thisinformationrequiressomequalitativesurveystobeconducted.

ResearchcompaniesorSurveyVendorssuchasDynamicLogiccreateshortsurveysthat can be launched as part of the delivery of the ad, to collect this informationfromusers.When thead loads, isclickedorclosedasmall survey loadsonto thescreen.ThissurveyoverlaystherestoftheinformationonthePublisherpageandisdesignedtoquicklyasktheuseraboutadeffectiveness,withoutdisruptingtheuserexperience.

This information, once submitted in the online-form, is sent on to the SurveyVendor’sdatabaseonanexternal server. Inorder for the survey to loadup in thefirstplace,asmallpieceofcodeneedstobeprovidedbythesurveyvendortothetrafficker long before the campaign goes live. The trafficker then loads this codeintothethirdpartyadserverinterfaceattheadlevelwhentraffickingthecampaign(addingthecodetowhichadsitisrequiredtofireoffagainst).Pre-createdfieldsintheadserverinterfaceallowforthesecodestobefiredduringoneormoreoftheJavaScriptactivitiesontheplacement.

TosetupaSurveyfromtheoutsetofthecampaign,theAgency/Advertiserneedstoreachouttothesurveyvendorwhenthecampaignisbeingdevisedandthesuppliedcodeby thevendorshoulddeliver someunique identifier in its returnad-call thatwillallowthedatafromthevendortobetiedtothedatafromtheadserverlateron.A Placement ID is suitable for this purpose. Therefore an optimal setup processwouldbe:

a. ThePlacementsshouldbecreatedbythetraffickerfirst

b. Then the trafficker should export themedia plan in the bulk trafficking tool

(displayingallofthePlacementIDs)andsenditalongtothesurveyvendor

c. The code corresponding to each of the placements should be created in theSurvey Vendor interface (not found inside the ad server), with amatchingPlacementID

d. The survey vendor should update the “survey vendor” field in the suppliedexportedmediaplanbeforesendingitbackthetrafficker

e. The trafficker should ensure that the code is set to fire at the right time (byeitheraddingtheinformationinthemediaplanandthenuploadingthemediaplanbackintotheadserverinterface,orbymakingthenecessaryadjustmentsoneachplacementoncethemediaplanhasbeenuploaded).

The various nuances of the traffickingworkflow are designed to facilitate cleveradditionstotheprocess,suchastheadditionofsurveyvendorcode.Theavailabilityofthebulkmediaplanfunctionalitymakesitmucheasierforthetraffickertoworkwiththeseadditionaltraffickingrequirements.

Figure 3.2.8 Placements andAds all have the capability to support or provide a“piggyback” to other blocks of code.These codes can be added by the traffickerduringthetraffickingprocessandappeartotheenduserasseamlessfunctionalitytothead,iftheyareseenatall.Thesemethodsallowadstoactaschannelswhichcanpourinformationtomoreexternalsources.Publishersoftenkeepakeeneyeonadsfor this exact purpose because ads pouring information to other networks cantarnish thePublisher’sownvaluable audience - aprocessknowndarkly industry-sideas“dataleakage”(seeGlossary).

We have touched on the ability of the ad or the placement to support the“piggybacking” of additional code (albeit the survey vendor code for now). Thesame fields in the ad server interfacewhich allows a JavaScript activity such as“onLoad”tobeselectedontheadlevel,beforefiringadditionalcode,existattheadlevel, placement level and higher up levels such as the campaign level. Thecampaignlevelallowsthesamecodetobeappliedtoalladsinthecampaign.TheAdvertiser level allows the code to be applied to all ads in all campaigns for thewhole Advertiser. If the trafficker pastes code in from a different system, likeSurveyVendorcode(called“thirdpartycode”or“4thpartycode”)intothesehigherlevelsof thehierarchy, theiruseallowsforthebulkimplementationofsuchcode,feedingthefieldsintheentitiesthat trailbelow.ThispavesthewayforaconcepttouchedoninChapter1-Doubleadtracking.

RecallingfromChapter1;doubleadtrackingmeansthatasingleadservercanhavea view over activity running in more than one ad server. This is made possiblebecause the ad tag code from one ad server is exported and implemented by thetraffickerintothesamethirdpartycodefieldsthatwouldbeusedforsurveyvendorcode.

Itwouldnotmakesensetohavetwowholepiecesofcreativefromabothadserversappearinthesameadspotatthesametime,sothetraffickerinthelesseradserverneedsonlytosupplythemostbasictypeoftracking.Thecreativewillthenjustbeservedfromtheonesystem,butthereportingoftheactivitywillappearinboth.Inthissenseanadserver’sprimaryfunctionistoservetheadbutinaclosesecondisthefunctionoftracking.ThissamemethodcanbeusedwhenaPublisherdecidestoservethecreativefromtheirownsystem.

Trackingbutnotserving

It isquitecommonindisplayadserving tohavea fragmentedmediaplan;wherethetraffickerhascreativeformostof thesetupbutmustuse“trackingpixels”forthe rest.Thecreative isusuallyserved from the thirdpartyadserverdirectly,butsomePublishers‘donotsupport thirdpartyadserving’and insteadpermitting theMediaAgencytraffickertosupplyatrackingpixel(foreachofthePublisher-hostedads)sothatthetraffickercanbesurethattheactivityistrackedorrepresentedinthethirdpartyadserver.

IsitpossiblethataPublishercannotsupportthirdpartyadserving?

Publishershave less togain from theuseofa thirdpartyadserver inacampaign than Advertisers or Agencies do. Implementation of coderequiresoperationalresources,extratrainingandideallythepresenceofaPublisheradserver.ThesemaynotbeavailabletosmallscalePublishersorPublisherswithcontenthoused inveryuniqueenvironments (suchastheemergenceofadsinmobileappscreatedwithoutastandardisedSDK-see Chapter 3 - Mobile Ad serving for more information). Clashesbetween code supplied from an external system could in theory alsoconflictwith the Publisher page, particularly if there is a poorCMS inplace. However given the demands of Advertisers to utilise third partysystemstoverifydeliveryandperformance,poorsystemsinplaceonthePublisher side are bad excuses for not offering total transparency.Publisherswhichrefusetheuseofthirdpartyadservingtagsareriskingdeeper investigation by the wider community to uncover reasoning toavoidtransparency.

Permanenttrackingpixelsneedtobecreatedintheadserverbythetraffickerandtheiradtagssentandinsertedintocreative(servedbythePublisherdirectlybythePublisher ad ops team, from the Publisher ad server). This brings the countingmethodology as close as possible to the third party method. In this instance thetrafficker can still send the ad tags to the Publisher at the end of the traffickingprocess,butsinceitisnotstrictlydisplayingthead,thecodelooksalittledifferent.Someadserversrefertothiscodepackageasa“TrackingPlacement”anditismade

upofatiny1x1trackingpixelimageandasinglelineclickthroughURL(aclickredirect).ThetraffickerwillneedtocreateoneforeachcorrespondingitemontheMediaplan,wheretheadsorassetsarebeingservedbyanysystemotherthanthethirdpartyadserver.Tokeepthingsniceandtidy,thetrackingplacementshouldbetreatedlikeanyotherplacementintheplatformandbeassignedtothecorrectsite.Given the nature of tracking placements it doesn’t make sense to add the “adcontrol”optionstotheseentities.

Figure 3.2.9 another ad assignment option for the trafficker: assign a trackingplacement(atrackingpixelwithclickredirect)toaplacement.Theflexibilityofatracking placement allows the tag to be trafficked to a second systemwhere theactualadorcreativemightbehosted.When the tag for the trackingplacement istraffickeditisconsidertobea“4thpartypixel”inanythirdpartysystem.

AnoteonClickCommands

DFA (theDoubleClick third party ad server) allows traffickers to createmultiple“clickcommands”(ClickCommandsarejustclickre-redirects)butcruciallyallowstheassignmentofmanyoftheseentitiestoasingleplacement.WhenthesetagsaretraffickedtothePublisher,thePublisherhasthefreedomtochoosejustoneoftheseclickcommandsfromthesuppliedtag(wherethePublishermaybesite-servingtheactualad).Aseachoftheclickcommandsmightpointtoadifferentlandingpagethe Publisher can use the power of the Publisher ad server to rotate the clickcommands across the site-served creative for the purposes of optimization andlandingpagetesting.

Sincealloftheclickcommandsareassignedtothesameplacement,DFA recordsthese“click”metricsunder the sameplacement inad server reporting.ThisgivesthePublishermorefreedomindeterminingwhichad todisplaywhilesiteservingwhile still counting clicks in the correct place in theDFA platform. Ad serverswhich limit thecreationofoneclickcommandtoone trackingpixel ina trackingplacement format do not force the Publisher to use both image (pixel) and clickportionsof the tag.The tagwillworkand themetricswillcounteven ifonly theclickportionorjustthepixelportionisused.

With the combination of these more advanced tactics in the trafficker’s arsenal,mostdisplaycampaignscanbesetupandmanaged.Beforeclosingoutthissectionlet’sfocusbrieflyonthemechanismsatworkherewhichareusefulindeterminingwhyanadthatmayhavebeentraffickedalreadyandimplementedontoaPublishersiteiseithernotreportingordisplayingasexpected.

7. AdReplicationAdDeliveryandReplication

To get an ad to deliver all over theworld requires several technologies working

together.Theadserveritselfwillnormallydothecomplexcalculationsanddecisionmakingaboutwhichadtodisplayaswellasstoringtheimagesandflashreadyfordistribution,butgetting theads from theadserverand itshardware tobillionsofpages (and eyeballs) requires the support of a CDN or content delivery network(SeeChapter1).The traffickerhas thecorrect access to thead server interface to swap inandoutnewadsevenwhenthecampaignislivebutthisactiondoesnothaveanimmediateresult.Thetimeittakestoswapacreativeover(orreplacinganoldadwithanewone)andhaveitappearallovertheworldinalivecampaignisreferredtoas“adreplication”, and for far reaching global distribution campaigns should take nolonger than around fifteen minutes (after saving the changes in the campaignmanagementscreensoftheadserver).

Forthoselookingtocheckthatsuchchangesarenow“live”itispossibletogotoaPublisherpageandrefreshituntiltheyseethechangelive,butitisrecommendedtoclearbrowsercacheandcookies(andwaitingfifteenminutes)beforerefreshingthepage. This ensures the old ad has not been cached locally and removes therestrictionsplacedbyfrequencycappingattheAdvertiserorPublisheradserver.

Finally if the ad is still not displayed, it is wise to check the settings for IP(geographical) targeting sincea tester sitting in thewronggeographywillnot seethecorrectcreativeifthecampaignisworkingasexpected(moreonthisinChapter4-Audiences).

Server-SideCaching

Inadditiontothisthereisacachingmechanismontheadserveritself,meaningthatifatagthatlooksthesameisservedfrequently,theadserverwillmakeandstoreitsowntemporarycopyforfasterdelivery.Toeradicatethetemporarycopyintheadserver,mosttagsincludeastringortokeninthepixelportionofthetagcodeitself.Thisstringissimplyarandomnumberthatchangeseverytimetheadisservedout,and since the random number is never repeated, the cache on the ad server iseradicated.Thestringisnormallyrepresentedwithsomethingthatlookslike:

”?rand=RANDOMNUMBER”

When the trafficker is piggybacking a 4th party tag, the instructions in the thirdpartyadserverwillexplainthatRANDOMNUMBERincapitalletterscanbeover-writtenwith a system-friendlymacrowhich,when thepixel is fired,will insert arandomnumberinautomatically.InsideDGMediamindthismacrois:%RAND%

Whatexactlyisatag?

WhenthetagcodeissentontothePublishertheimplementedtagsareconduitsforinformationbeingsenttotheadserverforreportingandcreativedeliverypurposesbut the tag itself is avery small a lightpieceof codeand stand-alone itdoesnotcontainallthecodeitneedstoperformcertainfunctions.Assuch,thetagcallsupononeormoreJavaScriptlibrarieswhichcontainmorecomplexcommands(andareatouchheavierintermsoffilesize).Theconstructoftheselibrariesisuniquetoeachadserverandsomemakebetteruseofvariouscodingmethodsthanotherstoensure

aslickerandsmootherexperiencewithsay;videoads,forinstance.Understandinghowthesepartswork,detractsfromthemoreimmediatefocusofthistextbutthirdparty ad serving providerswill be able to supply this information.The importantthing to note is that more complex capabilities are only possible through theenabling of these libraries, therefore any tags implemented in amore simple tagformatsuchasasimpletrackingplacementasopposedtoafully-fledgedJavaScripttagwillhavecertainfunctionsmissing(includingviewability).

9. MakingabiggerimpactwithDisplayCampaignsThebest advice to improveknowledgeand skills further in standarddisplay is toundertake some trafficking in the ad serving interface itself.Most third party adservers provide awealth of trainingmaterials and support to guide the traffickerthrough the campaign management process in their specific system and will behappytoassistinthesetupofafewtestcampaignsfortrainingpurposes.Thegoodnews is that all the concepts covered are common to all the big third party adservers, those absent features in one systemwill soon crop in the others, play agameofcatch-upwithoneanothertoprovidemoreandmorestandardisedfeatures.

Thenextsectionwill lookat theRichMediachannel.RichMediaisanadvancedformofdisplayadvertising,designedtoengageandattracttheuser.ItisespeciallypopularwithbigbrandAdvertisersandisthegatewaytoVideoadserving.DuetoitspowerfulimpactRichMediaislikelytoseea“come-back”inthenextfewyears,emerging in innovative ways through video players, mobile and through moreadvancedformsof realityaugmentation,but its technical setup is insomewaysalittlemorechallengingthanthesetupforstandarddisplayads.

Chapter3-Section3RichMedia• WhatisRichMedia?

• WhyuseRichMedia?

• CoreFormats

• Expandableads

• Figure3.3.1a

• Floatingads

• Figure3.3.1b

• Videoads

• Figure3.3.1c

• PushdownandSidekick

• Figure3.3.1d

• DataCapture

• HomePageTakeovers,WallpaperandPageSkins

• Figure3.3.1e

• Figure3.3.1f

• Polling

• StateBasedads

• LocalConnectionadsorSyncAds

• Figure3.3.1g

• Printads

• StandardisationandPublishercertification

• Fromthetop:BuildingRichMediaads

• Trainingthecreativedeveloper

• TwoschoolsofRichMediaspecs:PublisherandAdserver

• ParentandChildfiles-thepoliteload

• ExpansionsandFloating-thedeclarationofstate

• Interactiontracking:DefaultandCustomEvents

• Interactionsasadditionalclicks

• BuildingCustomRichMediaads

• QaingandTestingRichMediaads

• TraffickingRichMediaads

• PublisherTroubleshooting

• RichMediaforMobile&RichMediaforInstream

Thefollowingsection looks indepthat the fullcampaignsetupprocess forRichMediaadswhichareservedandtrackedbythethirdpartyadserverandtraffickedtoPublisherwebsites.

Strictly speaking Rich Media is not a channel but a subset or evolution of StandardDisplay ads. This means that a large margin of the trafficking process covered in thesection on StandardDisplay also applies toRichMedia.With this new territory comesnewterminologyandconceptsthatbeginrightatthemomentofcreativeconceptionandtrickledowntothedeliveryoftheadintheuser’sbrowser.

WhatisRichMedia?

TechnicallyRichMediareferstoanadthatcontainsoneormoreofthefollowingfeatures:

• Video

• Expandablepanels

• InteractionTracking

• PoliteLoading

• FloatingEffect*

(*Floatingadsaresometimesreferredtoas“OutofBanner“ads,althoughthisphrasehasalsobeencoinedtodescribealltypesofRichMediaadsinsomeadservers)

WhyuseRichMedia?

Asoutlinedatthetailendofthelastsection,RichMediaoffersAdvertisersthechancetomakemoreof an impactwith creative that ismore engaging for the user than standardDisplayadsorSearchads.RichMediaisreallycollaborationbetweenthethirdpartyadserverandthePublisher.ImplantingaRichMediaadontoaPublisherwebsiteisgoingtodrawuser attention towards the addue to thenatureof its animations, indoing so it ispossibleforadstobecreatedthatcompletelyhidethePublisher’scontent.Thisiswherethephrase“outofbanner”comesfromassomeRichMediaadsliterallyspilloutofthespacetheyweredesignedtofitinside.

SinceRichMediacausesa longer loading timeonapageandpotentiallycouldconflictwiththePublisher’sowncontentandcode,thePublishermustbeon-boardtoacceptRichMediabefore it is inserted into theMediaplan for anAdvertiser campaign.TypicallyaPublisherwillofferRichMediaatahigherCPMtoAdvertisersandAgenciesforspecificadspots.Adsoccupyingthesespotswillneedtoadheretosetofspecificationsdefinedbythe Publisher which will have been created to ensure that within the confines of thespecifications,thepageandtheuserwillnotbeovertlyaffectedbytheengagementwiththeseads.

AtahighercostofCPMandabiggerdrainonthemarketingbudgetforthecampaign,onemightquestionhowtoprovethatRichMediaisofferingmorevaluethanstandarddisplay.

Theanswercomesfromthepossibilityoftrackinginteractions.AsdetailedinChapter5-Reporting, Rich Media goes beyond impressions, clicks and conversions to providemetricsthatexplainbothwhattheuserisengagingwithandforhowlong.Thesetriggersand timers allow the Advertiser to determine how a user engages with the creative(allowing for improved creative optimization) and provides a new window oneffectiveness by looking at “engagement” on interaction as opposed to clicks to assessperformance.ItissafetosaythatwhenlookingintotheworkingpartsofaRichMediaad,thereismorecomplexitythanthereisforastandardadwhichmeansagreateropportunityfor creativity. Greater creative paves the way for awards for creative developers anddesignerswhich in turnattractgoodPR forAdvertisersand forCreativeAgencies.ThemostimpressivedesignseventuallybecomeadoptedatanindustrylevelanditisthankstothestandardisationofformatsbytheIABandcollaboratingpublishersthatwenowhaveformats suchas the“RisingStar formats”whichcircumvent thestandardisationofRichMediafrombecomingtoodiverseandunmanageablefrombothadevelopmentandtestingperspective.

CoreFormats

FormatsarethecoreofwhatmakesRichmediasoscalable.ByfocusingonthefeaturesofthemostcommonrichmediaformatsitispossibletoidentifywhatkindofformataRichMediaadisjustbylookingatitorplayingwithsomeofthecontrols.Formatsarereallyjust descriptions of advanced functionalitywithin aRichMedia ad but the presence ofestablishedformatnamesmovesallowsplanners,buyers,marketersandtraffickerstohavemeaningfulconversationsaboutrichmediacampaignswithlittleornoknowledgeabouthowtheadworks.

Withanestablishedsetofformatnames,itispossibletousethenamingconventionusedby the creative developer in the creative file name, or the name on themedia plan orPublisherIO,toensurethatallpartieshaveamutualunderstandingabouttheformatsusedinacampaign.Formatsarenotuniversallypriced,moreadvancedfunctionalitytypicallymeansahigherendcosttotheAdvertiser.

ThefollowingisasetofthemostcommonformatsfoundinRichMedia:

Expandableads

Figure 3.3.1a an expandable ad grows in size over the publisher content,overlapping it briefly while the ad content is delivered to the user. This figuredepictsRichMediaAd states of expansion and contractionwhich are covered inmoredetaillaterinthissection.

One of the most commonly used formats in Rich Media is the expandable(hopefully in time a word that will find itself a place in the Oxford Englishdictionaryfollowingtenplusyearsofindustryacceptanceanduse).Thisisaformatthatgrowsinsize(“theexpandedstate”)andthencontracts(“thecontractedstate”)downtoafixedadposition(calledthe“adlocation”)onthepage.Expandablescan“autoexpand”(ifPublisherspecificationsapprove)meaningthattheystartplayingintheexpandedstateratherthanrequiringauseractiontocausetheexpansion.Anexpansioncanalsobetriggeredbyaclickormouserollover.Theycanalsoexpandinanyof the fourdirectionsor expand inalldirectionsatonce.Expandablescanhaveatransparentbackgroundsothatintheexpandedstatetheanimationappearstointeractwiththecontentofthepageitself.

Floatingads

Figure 3.3.1b The floating ad format loads over the Publisher page content,sometimestintingthepagebehinditgreyuntilitisclosedwiththeXshowninthecorner.Once closed, floating ads do not strictly return to any fixed point on thepage.

Adiminishing format inboth theUKandUSmarkets; the floatingadappearsontopofthepagecontent,nottiedoranchoredtoafixedadspot.Theaditselfdoesnot open in a newbrowserwindowbut often comeswith the samequalities as anewbrowserwindow(itcanbedraggedaroundandclosed).ThefloatingadgoesthroughvariousstatesofplayduringwhichausercanchoosetoclosetheadviaanXinthecorner.ComingacrossafloatingadasauserwithoutanXdenotesafailureonboththepartofthecreativedeveloperandtheadserverforfailingtonoticeitsabsence (this helps explain why the Rich Media QA process is so important).FloatingadsaretypicallyfrequencycappedatthePublisheradserverlevel,sothatif theuser loads the samepage several times theywill eventually stop seeing thefloating ad so as to sustain a reasonable user experience and not risk “creativefatigue”.

Videoads

Figure 3.3.1cVideo ads appear in a standard ad spot on a Publisher web pagetypicallyplayingashortvideoclipwithcontrolsfortheuser.Mostcommonlyusedbyfilmstudiostodisplaytrailers.

Video ads are one of the most commonly used RichMedia formats and can becombinedwiththeexpandableorfloatingfeatures.Videoadslooklikestandardadunitsbutplayvideo(usuallywithsomecontrols)frominsidetheplacement.Videoads shouldnotbeconfusedwith Instreamads (whichare typically traditionalTVads purposed for online video broadcasters and purpose built video players - SeeChapter 3 - Instream and Video). In Rich Media Video, videos can startautomaticallybutthemutingfunctionalityisdictatedbythePublisherspecificationsforwhichtherequirementismostcommononmute.HDvideo,greenscreenvideo,full screenvideoandmultiplevideoplay (wheremanyvideosappear in theadatonce- likeinthecaseof theformatcalled“thevideocube”)areallsupportedbymost third party ad servers.Video can also be streamed into the ad live, and thestream itself can be loaded in real time or pre-loaded into the ad by a methodreferred to as “progressive streaming”. Specialist technologies and players haveemerged to support uniqueAdvertisers needs such as live video feeds: one suchexampleisBrightroll.

PushdownandSidekick

Figure3.3.1dASidekickformatbeforeandaftertheuserrollsovertheadwiththemouse.Therollovercausesthesidekicktoexpand,seeminglypushingthecontentofthePublisherpagetoonesideratherthanrollingoveritasitwouldnormallydoinastandardrichmediaexpansion.

InnovativeRichMediaformatsareconstantlybeingdevelopedbutfindingonesthatare engaging and not obstructive to the user experience can be a challenge. ThepushdownandsidekickIABapprovedformatsallowtheadtophysicallypushthepagecontentdownortothesideinahybridofthetraditionalrichmediaexpandablead.Theuserisexposedtothelargercanvasoftheadtemporarilyuntiltheyrequestthecontentbereturnedbyabuttonorslider.Suchformatsareevenmoreeffectiveon deviceswithout amousewhere a sweeping action on a touch screen ismoreintrinsic to the user experience with the ad. Ad Serving providers and specialistRichMedia technologies such asDGMediamind and US-centric Pointroll workwithmajorPublisherstodevelopnewformatswiththeaimofbettermonetisingthePublishers inventory. Formats chosen by the Publisher can often bemadewith acontractofexclusivitywiththetechnologycompanythatdevelopedtheformatforafixedperiod.InsuchinstancesanyAdvertiserorAgencyrunningtheiradinanewand exclusive format might have to ad serve their ads through the advisedtechnologytohaveitaccepted.

Datacapture

Owingtotheirlargerfilesize,RichMediaadscandomorethanstandardflashads,evengoingsofarastoreplicatethefullAdvertiserwebsiteintheconfinesofanad.AgoodexampleiswithaDataCaptureformwhichpassesorcollectsdataaboutauser via a form in the ad and sends it off to anAdvertiser database in real time,potentially returning information like a voucher code.A use casewould be for amobileproviderliketheEuropeanmobilegiantTelefonicaofferingfreesimcardstousers.

Since no payment details are required, the user completes a form in the ad andsubmitsitbeforetheyaregivenaconfirmationmessageinthead.Theconfirmationmessagecontainsaconversiontag(seeChapter4)whichlogsaconversionwithoutneedingtodirecttheusertotheAdvertisersiteandbyreturningtheconfirmation,thedatacaptureformhasspokeninreal timewiththeOrangedatabaseclarifyingthatthefreeitemswillbesenttotheaddressdetailssupplied.DatacaptureformsofthisnaturearenotmeanttostorePersonallyIdentifiableInformation(PII)ontheadserver, but instead encrypt the supplied information and send it to theAdvertiser

databasedirectlyinlinewithlocaldataprotectionlaws.

HomePageTakeovers,WallpaperandPageSkins

Figure3.3.1eawallpaperexecutionsitsbehindthePublisherwebpagestretchingfromthelefttotherightgutter.WhenfurtherRichMediafunctionalityisaddedtothedefaultexecutiontheeffectisengrossingincludinganimationswhichappeartopullthewholeofthePublisherpageapartinitsentirety.Lesssophisticatedversionssimplyplaceverylargeimagesinthebackground.TypicallythePublisherpageisnottransparentasshowninthisdiagram.

ModernPublishersitedesignscentrethecontentontheusersscreenproducinglarge“gutters”of empty spacedown the sides.These canbe filledwith agiant image,images or a playing video (a very nice effect on a big screen). This format isreferredtoasaWallpaper.IncomparisonaPageSkintendstomeanadvertisinginevery possible ad space and even sponsoring Publisher content working in closecollaborationlongbeforethecampaignisduetobegin(sponsoredcontentisnotyetservedbythirdpartyadserversbutisreferredtoasan“Advertorial”).ThepresenceofaPageskin isalso referred toas“skinning”apagealthoughwhen itcomes toskinningahomepage, it is referred toasa“homepage takeover” (HPTO).Homepage takeovers are a big deal - they are usually more expensive than any otherformatsbecauseoftheenormousreachtheyhave(thinkMSNorYahoo!Homepage)andrarelylastlongerthanasingleday.HPTO’salongsideInstreamadunitsarethecream of what is referred to as premium inventory (the most valuable ad spacemoneycanbuy).

HPTO’s bring a wealth of requirements and needs to the third party ad servingtechnology(suchasrealtimereportingandoptimization,aswellastheneedforthead server to have strong relationships with big Publishers to ensure thattechnologiesworkinharmonywithoneanother,evenwhenPublisherpagesarere-designed).

HPTO’sarenotassimpletoformulateastheysound.ForinstancetheHomepagesofamajorPublisherlikeYahoo!differineachcountry,thereforetheHPTOformatfor an international campaign, may have to undergo an extended period ofdevelopmenttosuitthehomepageofeachcountry.

Figure 3.3.1f The Homepage Takeover is a complete unification of all of theavailableadspacesonasinglePublisherwebpageusedinasinglecampaignbyasingleAdvertiser.Homepagetakeoversareusuallyhighcostexecutionsthatrequirestrong collaboration between the ad server, the Publisher and the Creativedevelopmentteam.

Polling

Apolladcantakeuserinformationenteredintoasingleadanddisplaytheresultsof that collective input in real time inside the ad to all other users. Behind thescenes, theadserver isoftenpoweringasmalldatabasewhichcollects theresultspurelyforthepoll.

This collaboration effect combinedwith a high reach campaign, can achieve pollresultsthatarecompletelyunattainablebyapollonanAdvertisersitealone.Someclevermanipulationof this technologycanconnect thousandsofusers together inreal-time collaboration. So instead of collecting simple survey results fromindividual users and displaying a graph to the collective, users can takewebcamphotosofthemselvesandcollectivelybuildacollage.

State-basedads

NotstrictlyaformatbutratherafeatureofRichMediaadsistheconceptof“State-basedads”whichareamoreflexiblewayofperformingcreativesequencing thanusing the in ad-server sequencing options alone (see Chapter 3 - Display adserving).Statebasedadssimplyrememberwhat“state”theuserlastputtheadinto(suchasleavinganadhalfwaythroughtheplayingofavideo)sothatshouldtheyencountertheadorcampaignorevenAdvertiseragain,thenextadwhichfollowscouldpickupfromwheretheuserleftofforperformanotheraction.

Statebasedadsworkbydroppinganadditionalcookieontheusermachinesothatthe“state”canberecalledlaterorutiliseareservedareaoflocalmemoryavailableforwhichbothFlash(FlashCookies)andHTML5cantakeadvantage.

LocalConnectionadsorSyncAds

Figure 3.3.1g Sync ads appear to communicate with each other on the pagecreating an affect that the two separate units are really just one ad. A clevermanipulationof theFlashActionscriptcodecombinedwith thecapabilitiesof theadserverallowstheadstoappeartobecompletelysynchronised.

Local connection ads, also called “Sync ads”, create the wonderful effect of theleader-boardormastheadandtheskyscraperunitspeakingtoeachotherintandem.The surprise dawns upon the user when the seemingly separate ad units bothdisplayed on the Publisher page at the same time, suddenly interact with oneanothersuchaspassinganobjectbetweenthemorsyncingintimewitheachother.ForthePublisheradoperationsteamitisjustacaseofimplementingtwoseparatead tags on the page at the same time.RichMedia templates created by the thirdpartyadserverallowtheCreativeAgencytoswiftlyandeasilyreproducethiseffectforthosePublishersthatsupportamatchinglayout.SyncAdsaremorecommonlyused in theprocessofskinningawholepage,since twoof themajoradunitsarealreadytakencareofinauniqueway.

Print-ads

As explained later in this section rich media ad units can contain a number ofhiddenawayfilesas“childfiles”whichare initiallydisplayed to theuser.Oneofthese so called “child files” could be a full size print version of a retail voucher,which inside theconfinesofa smallorevenstandard richmediaunithas troublebeingdisplayedforprintingpurposesonasmallscreen.Therichmediaunitcanbecodedtoallowtheusertotakeadvantageofthevoucherofferfrominsidetherichmediacreativeandhitaprintbutton,toprintoutacopyofthehiddenchildfileandthus the voucher. Again this is more of a Rich Media feature but gives a goodindicationas towhatelsecanbestashedawayutilising themuch larger file sizesgrantedwiththeuseofRichMediacapabilitiesandspecifications.WholebrochuresinPDF format arenot uncommonlyused to stash awaymore information for theusertoexposelateron,insideachildfile.

Arunthroughofpopularformatsisratherlikebrowsingamenuinarestaurant.DifferentdishessuitdifferentAdvertisersatdifferenttimes.Thereisnohardandfastruleforwhattousefortherightimpactormessage.Oftentheskillofthecreativedeveloperoragencyhasthebiggestinfluenceinthisarea.Itisallwellandgoodknowingwhatisavailablebut

it is important to ask two key questions before moving forward with a Rich Mediacampaign:

a. Canmychosenformatberunonanysiteorpage?(SpecificationsandCertification)

b. Howaretheseculinarydelightscreated?(HowareRichMediaadsbuilt?)

Theseconsiderationswillbebrokendowninsequenceinthefollowingparagraphs:

StandardisationandPublishercertification

In theearlydaysofRichMedia,evenwith the IAB trying to standardise formats,manyhighvolume trafficsiteson the internethad justnotencounteredRichMediaads in thepast and although theywere keen to run new formats, these had to be testedwith testcreativeandtagsbeforethecampaigncouldgolive.RichMediatechnologyprovidersandthirdpartyadservingprovidersemployedteamsofpeopletocontactPublishersiteswellahead of a campaign launch date on their very first engagement regarding a specificformattotestthatitwouldworkonthePublishersitebeforethecampaignwentlive.Withrigorous testing the formatscouldbe runand thesameformatwouldnotneed tobe re-testedforfutureAdvertisersandcampaigns.ThePublishersitewastherefore“certified”toruntheformat.

Today big Publishers have gotten far more efficient at this process and Rich MediacertificationforformatsdeemedasstandardisedbytheIABnolongerrequirecertificationin this way. What this means it that as long as Publishers and technology providers(includingthirdpartyadservers)servetheIABRichMediaformatsinthewaylaidoutbytheIAB,thestandardisedmethodmeansthatnofurthertestingisneeded.

This industrydevelopmentmadeRichMediascalableandtheprocessofstandardisationandcertificationmeansthatanyPublishersubscribingtothestandardisedpracticesoftheIABcanrun IABRichMedia formats.The samegoes for thead server.Thepracticeofcertificationstillgoesonwherestandardisationandadoptionhasnotreachedfullindustrybackingbecausethechannel inquestionis toonew: todaythis is trueofcertainformatsfor Instream and certain formats ofMobile (see Chapter 3 - Instream and Chapter 3 -Mobile). These newer formats are the living evolution of traditional RichMedia so itmakessensethattheindustrywillreturntothecertificationprocessesthatworkedsowellinthepast.

Fromthetop:BuildingRichMediaads

ThebuildingofRichMediaadsfollowsthesamedevelopmentcycleasbuildingstandarddisplay ads generally speaking. The differences are building in and testing the morecomplex elements of the functionality. The final part of this section will cover off theadditionalconsiderationsthatneedtobemadewhenbuildingRichMediaads.

1. Trainingthecreativedeveloper

2. TwoschoolsofRichMediaspecs:PublisherandAdserver

3. ParentandChildfiles-thepoliteload

4. ExpansionsandFloating-thedeclarationofstate

5. Interactiontracking:DefaultandCustomEvents

6. Interactionsasadditionalclicks

7. BuildingCustomRichMediaads

1. Trainingthecreativedeveloper

Creative developerswhohave built ads for ad serverswithin theFlash authoringenvironmentinthepastwillstillneedtobeintroducedtotherealmofRichMediaad creation since the various Flash elements (Expansions, Floating, Video,InteractionsandPoliteloading)utilisedifferentaspectsofthecomponentssuppliedtothecreativedevelopmentteambytheadservingtechnology.WhenitisfirstclearthatRichMediawillbeusedinamediaplan,thetechnologyproviderhostswhatisreferred toasa“kickoffcall”between theMediaAgencyand theCreative team.ThismeetingisusedtoallowtheMediaAgencytoexplaintotheCreativeAgencywhat execution or format had been agreed to runwith a given Publisher and thecreative developer is subsequently asked if they need additional training on theavailable tools in order to build the required format. Creative developers shouldusually be trained an absolute minimum of a week before the ad needs to becompleted, toallow thedeveloper timefollowing the training toget familiarwiththe new aspects of the supplied components and toworkout how to get creativewiththecode.ThekickoffcallisusedtoallowtheMediaAgencytoexplainwhattheirexpectationsarefromtheexecutionintermsofadserverreportingduringandafter thecampaign. Insome instances thiswill require thedeveloper toundertakeadditional work implementing what are referred to as “custom interactions” or“customevents”.Beforetheworkcanstarthoweverthecreativeteamalwaysneedstoknowone thingaboveallelse:whatare the limitations to thecreative?This isvitalsinceRichMediaopensthedoorwaytomoreadvancedcreativeexecutions,soit isup to theMediaAgencyand thirdpartyad servingprovider to laydown thegroundrulesormoreadvancedspecificationsfromtheoutset.

2. TwoschoolsofRichMediaspecs:PublisherandAdserver

AdServerRichMediaSpecifications

Creativedevelopersanddesignersapproachingrichmediacommonlyaskfortheadserver“specs”.EssentiallytherequestistodeterminewhatneedstobedonetotheFlash file creative so that is accepted by both the third party ad server and thePublisher. Third party ad servers and rich media technology providers make theworkofadevelopereasierbysupplyingasetofFlashComponentswhichcanbeinstalled in thedevelopers localversionofFlash (just as forStandardDisplay adserving).Thecomponentsalongwithasupplyoftemplates,allowsthedeveloperto“build”thead,containingthecoderequiredtouploadthecreativetotheadserver.This code is the majority of the “specification”. Casting one’s mind back toStandardDisplay,theonlycoderequiredbymostadserversisthepresenceofthe“clickTag”.WithRichMedia this isoftensucha lengthychunkofcodesiphonedintomultiple locations inside the flash file that it is simpler to supply a templateratherthancodeitallinfromscratch.

Other requirements are usually around the supported formats (the restrictions forwhichcanbegleanedfromtheComponentpackage:andbylookinginsidethead

server interface). Sometimes the ad server specifications restrict the number ofsupportedchildfiles,andthefileweightoftheoveralladunit,sorarelywillarichmediacreativeexceedtenchildfiles,andfileweightsofover30mbarestillararity(although most ad server can handle more if pushed to do so). In fact the bestresponse to a query about specifications is to look at the Publisher specificationsfirst, as these contain much tighter restrictions than ad server specificationsthemselves.

PublisherRichMediaSpecifications

User experience onPublisherwebsites is greatly impacted by the speed atwhichuserscannavigateandloadthesite.Forthisreason,Publishersspendtimecarefullychoosingwhatthefilesizelimitsshouldbeofobjectsandcontentpulledintoapage(as this directly impacts the page load time). Standard flash ads andRichMediaalike need to be limited to certain available ad dimensions and file sizes andPublishers release “Publisher specifications” to media and Creative Agencies tokeeptheuserexperienceatitsoptimum(mostPublisherspecificationsarepubliclyavailable).

ItisimperativetohavethecreativedeveloperbuildthecreativetotheselimitationssothatadsandadtagsarenotrejectedwhentestedbythePublishersadoperationsteam, before the campaign goes live. Publishers across the board have varyingPublisher specificationsand thismeanscheckingbeforeacampaign toensure thedevelopment team have received specifications from all Publishers. Invariably tosavetime,CreativeAgenciesbuildtothe“lowestcommondenominator”ofspecs,meaning that themost restrictingspecificationsdictate thebuild, so that thesameads can be distributed to multiple sites without issue. This is all the same forstandarddisplayasitisforRichMediabutRichMediaspecificationswillbemorecomplex and flexible than those for standard display; particularlywith regards toadsutilisingmethodsthatdonotdisrupttheuserexperience.Themostcommonoftheseistheparentandchildfilemethod;knownasthe“polite”load.

3. ParentandChildfiles-thepoliteload

ParentandChildfiles

Unlikethevastmajorityofstandardbanners(whichconsistofasingle“.swffile”),Richmediaadscanbemadeupofmanydifferent files.Theparentormaster fileloads first on thePublisher page (containing the core code required for the ad topasstherichmediaspecificationsofthethirdpartyadserver).Arichmediaadcanbemadeupofasinglemasterfileandanumberofchildfiles,wherethemasterfilemakesareferencetothechildfilesbynamesothattheircontentcanbepulledintotheadasitplays.

For example amaster file containing the controls for avideoplayer, references achildfilecalled“video.flv”(wherethechildfilejustcontainsavideo)andcouldnotbeplayedinthebrowserwithoutthecontrolsandcodefromthemasterfile.Childfilescome inmanydifferentvarieties; fromdifferent file typesofvideo toPDFs,worddocuments,soundfilesandXMLfiles.WhentheCreativeAgencyuploadsthemultiplefilestotheadserver,theadserverusesthenamesofthechildfilestolink

themtothemasterfile;commonlydictatedbyanintelligentmethodofpairing,suchasbyuploading the files togetheror into the same storage folder in the “creativelibrary”sectionoftheadserver.MoreadvancedmembersoftheMediaAgencycantake a look inside the parent file to determine which order the child files arereferenced in (this can be done either in the Flash authoring tool or in variousfreewareprogramsdesignedtopeer into thecodedirectly).Theorderof thechildfiles, just likethenamingofthem,isimperativetohavetheadworkingcorrectly.Sometimesatraffickeruploadingarichmediaadintotheadserverwillseeanerrorwhen a child file is missing and opening the parent file up to see which file ismissingintheorderoffiles,willrevealthefile’snameandallowthetraffickertochaseupthecreativedeveloperfortheabsentfile.

PoliteDownload

Why have one parent file and many child files? Why not simply have a singleparentfileandbedonewithit?LookingbacktothePublisherspecificationsandthediscussion about the user experience; polite downloadmeans that themaster fileloads first, usually with the rest of the Publisher page and then the child file“politely”loadsinafterwards(oftenasabackgroundprocess),sothattheuserdoesnotexperienceaslowpageloadtime.Acommonpolitedownloadsituationwouldseeamasterfileof40kbandachildfileofaround2mb.Thepoliteloadpreventsadisruptiontotheuserexperienceandtheeffectisoftenundetectablebytheuser.Anexamplewouldbe instanceofan in-pagevideoadwhere theparent file issimplydisplaying theholder imageof the first frameof thevideofilewithaplaybuttonwatermarkoverthetop.Whentheuserclicksthebuttonthevideobeginstoplaybutthebufferinghasalreadystartedandcompletedbeforetheuserhitsplaysothereisnodisruption to thevideo.VideoandpolitedownloadRichMediaadscommonlyutilisethisfunctionalitybuttheparentfilesofexpandablesandfloatingadswouldbefoundtocontainreferencesinthecodetothead’srichmedia“state”;anessentialconceptinRichMediareporting.

4. ExpansionsandFloating-thedeclarationofstate

Forbothexpandablesandfloatingads themoreadvancedfunctionality isenabledowingtothepresenceofwhatisreferencedtointhedevelopmentandbuildcycleasthe “state” of the ad. (This is not to be confused with “State-based ads”, whichallowstheuser’sinteractionwithanadtoberememberedforsubsequentencounterswith thecampaign)Hiddenaway in thecodeofaRichMediaexecution forboththesetwoformatsisthead’sstate.Whentheadchangesitsstate,furtheractionscanbeundertakenby the code.For examplewith expanding ads, the ad has twokeystates: Expanded state and Contracted State. The creative developer can insertcommands into the core codeof the ad, to runas the it entersor returns to thesestates,suchasthecommandtobeingplayingavideoorsoundfile.Thisfreedomisalsograntedtothetraffickerwhocan(aswithstandarddisplayads)piggybackextracode toexecuteonachangeof state suchas the firingofapiggybackedpixelorconversiontag(seeChapter4).Statesareimportantforreportingaswell,sincethechange from one state to another fires off default interaction tracking such ascountingthenumberoftimestheadhasbeenexpanded.

5. Interactiontracking:DefaultandCustomEvents

RichMediacansupportthetrackingofinteractionswhichmeansthatapartfromtheusualtrackingofimpressions,clicksandconversions;therichmediafilecansendinformationtotheadserverwhentheadregistersanyactionatallandtheseappearinreportingclassedas“interactions”.

Interactionscanbetimers,rolloverswiththemouse,changesinthead’sRichMediastate, clicks, user gestures and the pressing of keyboard buttons. Rich MediatemplatessuppliedbythethirdpartyadservertotheCreativeAgencywillpre-codeanumberofdefaultinteractions.Agoodexamplemightbethatforavideoad,theadserverrecordswhentheuseris25%throughtheplayingoftheadandthen50%and 100%. Each of these interactions will have its own column in ad serverreporting to aggregate the number of times they occurred; which assists inunderstanding creative performance. For example with the trailer for a film, acreative that isnever recording“100%plays”maycontainvideo that is too long.Other default interactions include total number of expansions and a timer forinteraction duration. It’s worth noting that default interactions are not industrystandardisedmeaningthatthereisnolistofalldefaultinteractionsthatalladserverandRichMediatechproviderswillsupply.

Itisthereforeworthcheckingwiththethirdpartyadservingproviderexactlywhatcomesasadefault (anexampleRichMediareportwilldisplay this).Furthermorethemethodology behind default interactions is different. One towatch out for isknowing when the interaction timer begins: some ad servers do not track“accidental” rollovers as part of the ad engagement while others do. This canexplain quite varied differences between interaction rates between differenttechnologysuppliers.AusefuldefaultinteractionforAdvertiserutilisingTVaswellas online deliverymechanisms is the “brand exposure duration”metric. This canmakearichmediaadcomparabletoanadforTVcampaignforthemarketerontheAdvertiserside,wherethetotalviewinghoursnationwidearecapturedandtallied.All suchmetrics are simplydefault interactions andappear in ad server reporting(seeChapter5).

ThankfullyAdvertisers are not left justwith the default interactions but have theabilitytoinformthecreativedevelopmentteamtoaddcustominteractions.Custominteractionscanbeaddedbythecreativedevelopertoanypartoftheadduringthead build process. This could mean the pressing of a custom button, the userhoveringoverthebrandlogo,oreventhemeasurementoftimebetweeninteractionsofdifferenttypes.

Custominteractionsalthoughusefulhavelimitedapplicationsoitisimportantnottohavethedevelopergooverboardwhenimplementingthem.Duringthekickoffcall,as theAgencyorAdvertiseradvises thecreativedeveloperof the interactionrequirementsitisworthconsideringwhatactionscanbetakenasaresultofhighorlowfigurescollectedfortheseinteractions.Forinstances,trackingandreportingonthenumberoftimesauserscoresagoalinaninteractiveRichMediafootballgameadwill not inform either theAdvertiser orAgency about user brand engagementandmayonlyservetoappendthegoalscoringfunctionalityinfutureiterationsof

theupdatedcreative.

6. Interactionsasadditionalclicks

Custom interactions add almost unlimited flexibility when it comes to tracking.Another key component for RichMedia ads is that they support multiple click-redirects.So,differentelementsoftheadcanclickthroughtocompletelydifferentAdvertiser web pages. The ad server allows all clicks taken by the user in thismannertoberecordedasclicksbutreportingcanbreakouttheincrementofclicksby their click throughURL todiscernwhichpage received themost trafficwhilestillaccuratelyreportingonthe“totalnumberofclicks”.MoreonthisinChapter5-Reporting.

7. BuildingCustomRichMediaads

It isallverywellandgoodstandardising themostpopularandwellused formatsbutthewholeconceptofRichMediaisbasedoncreativeimpact.TostandoutinacrowdonemustlookdifferentandinRichMedia,suchanachievementsometimesliterally means ‘thinking outside of the box’. To achieve executions that lookincredibly impressive, such as those that shake the whole page and seem todismantle elements that remain functional (such as search boxes and navigation),the creative developer must work alongside the Publisher. Often the ad servingprovider will liaise with the Publisher directly to cast aside the Publisherspecifications for a single high impact execution that, like a firework, is verydramatic but shorty lived. Most home page takeovers utilise an element ofcustomisationwhichadslongertothedevelopmenttimeoftheadandwillcomeoutpricierthanstickingtostandardformats.

Once the creative developer has completed thework of adding functions to the designteam’screative,thead(justlikewithstandarddisplay)isreadytouploadtotheadserver.ThiscanbeundertakeninmostinstancesfromwithintheadserversuppliedcomponentsinsidetheFlashauthoringenvironment.Unlikestandarddisplayads,RichMediaadstendtobemadeupofseveralfiles(atleastoneparentfileandonechildfile).Someadserversallowdeveloperstozipallthefilesuptogetherina*.zipfileorfolderbeforeuploadingthemtotheadserverfor thetrafficker toworkonthem.Beforethetraffickercompletesthe trafficking process, the complexity of the Rich Media ad requires it be morethoroughlytestedandQA’d.

QAingandtestingRichMediaads

Building richmedia ads to specifications takes some time and someCreativeAgencieshavenotworkedwithsomeadserversbefore,andsowilltaketimetoinstallandgetusedto their components. The kick off call, as described earlier in this section helps ensurecreativedevelopersarefullytrainedbeforeundertakingthework.However,aswithusinganythingforthefirsttime,progresswillbeslowerthanallowingtheCreativeAgencytoworkwithtoolsthattheyarealreadyfamiliarwith.

This is an interestingdiscovery for ad servingprovidersbecause lastminute campaignsrequirefastcreativeprocessingandMediaAgenciesandAdvertiserswillsometimesdeferto theCreativeAgency’spreferenceswhenselectinganad serving technologyprovider.Thisdecisioncanbesolelybasedon theusabilityof the toolsandhowcomfortable the

developeriswithsupportreceivedfromtheadserverdirectlyintimesoftrouble.

Development timevariesdramatically,butasingleadconceptwithrichmediaelementsfrom scratch never takes less time than a single working day (if you find a CreativeAgencythatcandothis;checkthequalityoftheirworkandthenstickwiththem!)Infactthe longest part of the process for any rich media campaign setup is thedevelopment/codingstage.

TheCreativeAgencyisexpectedtotesttheirowncreativelocallyandpotentiallyeveninasimulatedPublisherenvironmentonthethirdpartyadserver(somethingthattheMediaAgency can also do) before the ad is invariably tested manually by the third party adservingsupportteamaswell.

Oncethecreativepassesthesechecks(aprocesscalledQA-forQualityAssurance),ThePublisherwillalsotesttheadviathereceivedadtagsbeforeputtingtheadlive.Theadcan fail to pass QA at any one of these checkpoints either for specification failure, orfunctional failure and the creative ispushedback to thedeveloper to amendor rebuild;sometimeswithassistancefromthesupportteamatthethirdpartyadserver.Adscanalsofail evenoncea campaigngoes livedespite rigorous checks soQAbecomesevermoreimportant.

TraffickingRichMediaads

ApartfromtheextendedtraffickingtimerequiredforQA,richmediaadsaretreatedthesamewayasstandarddisplayads,sothatoncetheadiscreatedintheadserver,itcanbeassignedtoaplacementandthetagscanbesenttothePublisher.

As RichMedia ads can be more expensive than standard ads to serve through the adserver,thetemptationisforsomeMediaAgenciesistogodirectlytothePublisher.Thedangerhereis that theAdvertiser losestheabilitytode-duplicatetheuser tounderstandtheirlifetimeinteractionwithabrandwhenthewiderpictureofapathwaytoconversionis considered. In addition independent third party reporting is unavailable and complexinteraction reporting is not commonly available in publisher side rich media servingtechnologies.

Creative Agency and cross Publisher support would also vanish meaning that smallproblems can quickly escalate beyond repair. Third party ad servers conveniently offertools to make sure that RichMedia ads are not wasted such as through on-interactionretargeting(Chapter7)orwithadcontrolsthatensureRichMediaisreserved,deeperinusersequencingormanipulatingfrequencycappingtoensureRichMediaadsareshownfrugallytousersordisplayedatmoreeffective timesof theday.Noone isdenyingthatRichMedia is a cheap alternative to other channels but the value of the impact can beharnessedthrougheffective traffickingbymanipulatingcleveradcontrols.Asexplainedin the previous section, the experience of the trafficker would add enormous value tomediabudgetingbybeingawareofalloftheadservercapabilities.

PublisherTroubleshooting

AsmentionedevenafterQAandtrafficking,theRichMediaadcouldfailonthePublisherwebsiteowingtoanoversightinthetechnologyorhumanerrorinsetup.Insuchinstancestechnical support arebought invia the thirdpartyad servingprovider toworkwith the

Publisher to fix theproblem.This is anadditionaloperational cost thatPublishersmustbearinmindbeforeofferingtoallowRichMediaadstobeimplementedonsitestoutilisegreaterincomeontheinventory.

But how after so many checks can a rich media ad fail? The most commonlyreported types of failure on a Publisher site are to do with how the differentelementsofthepagearelayered,suchthattheadappearsunderneathorontopofthewebpagecontent.Generallyspeaking,thecauseoftheproblemisawonderfullyexoticCSSelementcalledthe“Z-index”.Z-indexreferstothelayerthateachpartof the web page lives on and there are thousands of such layers that couldpotentiallyexist.

Althoughtheadisdesignedtoappearontheupmostlayer,sometimesthisdoesnotoccur and the ad servermayhave to use a few JavaScript tricks (where able), toinfluencethelayeringoftheelementsaroundtheadsoastopushthemdownontheZ-index and raise the ad upwards. This is made available by piggybackingadditional code onto the existing ad or even inserting it into the creative itself.Publisher troubleshooting, although time sensitive for live campaigns, can be aprocess of trial an error; especiallywith limited Publisher support, so patience ismandatorywhileaconflictisbeingresolved.

Oftenthesupportteamhandlingtherequestwillrequirethattheadtagbeloadedinthe livepageor an exact replica inside apublisher “stagingenvironment” so thatthrough trialanderrorwithcode, the issuecanbe resolvedand thecampaignputlive again. Rich Media ads put live on ” blind networks” face the troublesomeconflictofhaving tosupplyaURL to thesupport team to resolve theerrorwhilecruciallynotexposingthesitenametotheAdvertiserorAgency.ForthisreasonitisnotrecommendedtorunRichMediathroughsuchagreements.

Due to the open ended nature of Rich Media (to express creative flare throughtechnology) there can sometimes be unforeseen technology conflicts which mayresultfromupdates tobrowsers,updates to thecoreFlashplayer technologyfromAdobeandotheranomaliesinJavaScriptandFlashelementsonthesamepage.Allof which can take a thorough investigation to resolve. Thankfully Publishertroubleshooting issues rarely occur and the Advertiser’s return, for high impactformats, are both rewarding to theuser visually and facilitate campaigngoals foreffectivereachandgreaterinteractionforthecampaign.

RichMediaforMobile&RichMediaforInstream

Thankfully once one has mastered Rich media, most of the other capabilities of theDisplay channel seem like child’s play. Beyond Display, Rich Media has begun topenetrate the channels of Mobile and Instream too. These nuances are covered in theremaining sections in this chapter. But before leaving Display advertising behindcompletely there remainsanelephant in theroom,onewhichhasallowedgiantsamongwebtechnologycompaniestothrive.ThiswouldbetheabilitytocreatedisplayadswhichareDynamic in nature, circumnavigating the need to have the creative developer buildquitesomanyads.Thenextsectionfocussesonwhat isreferredtoasDCOChannelor“DynamicCreativeOptimization”anditunlocksauniverseoftruecreativeoptimization.

Chapter3-Section4DCO

• StandardXMLDisplay

• Figure3.4.1

• Figure3.4.2

• DCODesignPhase

• Creatingthedynamictemplate

• Figure3.4.3

• DynamicCreativePlanning

• Figure3.4.4

• DCO-AdCreationPhase

• SingleVersioningorManualCreation

• ExcelMassVersioning

• MassVersioningfromaFeed

• Figure3.4.5

• Using a web service to restrict load on the Advertiserhostedproductimages

• Figure3.4.6

• DCO-AssignmenttoPlacementPhase

• Placementvs.audiences

• Figure3.4.7

• TargetingdifferentVersionstodifferentAudiences

• Figure3.4.8

• ConnectingDCOwithAudienceBuying

• DCO-OptimizationPhase

• Auto-OptimizeVs.ManualControl

• Figure3.4.9

• Auto-optimizationasa“BlackBoxTechnology”

• Multi-varianttesting

StandardDisplayandRichMediaadscouldbedescribedas“Static”adssothatoncethecreativedeveloperhascompletedbuildingtheadsandhasuploadedthemtotheadserver,the content of that creative does not change. The elements or assets that make up thecreative are rigid and cannot be moved without republishing the master file. If an

Advertiserortraffickerwantedtheadtolookdifferent,forwhateverreason,thecreativedeveloper would have to adjust the creative on their side and then upload it to the adserversothatthetraffickercouldreplacetheexistingwiththenewad.ThisprocesscanbeacostlyundertakingforAdvertiserswithregardstocreativecostsandforalmosttenyearsadscreatedforadservingfordisplayexistedonlyinthisconstruct.

Creativethatcanbevisuallyadjustedwithcontentcontrolledbythetraffickerandnotbythe creative developer means that costs all remain on the media side (Agency andAdvertiser)asopposedtoonthecreativeside(creativedeveloperandCreativeAgency).Controlledcostsor lowercostsaremoreattractiveindustrywideandmeanmorecanbespentonbuyingtheadspace.Thecreativecanbedesignedsothatitselementsaremorefluidandnot “static”and this is called“DynamicCreative”.Theutilisationofdynamiccreativeforthepurposesofimprovedcampaignperformanceinthethirdpartyadserveriscalled“DCO”or“DynamicCreativeOptimization”.

DCO began life when creative developers realised that if they used an XML file typewhen they created ads that it would save them a lot of time if they wanted to makechangesremotelylaterwithouthavingtorepublishtheflashfiles.ThissectionstartsoffbylookingattheseearlyXMLdrivendynamicadsbeforefocussingonDCO,lookingatthecapabilities of this ‘channel’ and the bright future which lies ahead for its applicationacrossDisplay,MobileandInstream.

StandardXMLDisplay

Asa.swffileisnearimpossibletobreakintowithoutheadingbacktotheoriginal.flafileandrepublishingthead,creativedevelopersfoundit toofrustratingandtimeconsumingandresolved theaccessproblemusinganXMLfile instead.Unlike.swffiles, theXMLfilecaneasilybeeditedinaprogramlikeNotepadorWord,soacreativedeveloper’sflashskills and premiumaccess to the flash authoring tool are not requisites to changing thecontent of anXML filewhich in turn is used tomake adjustments to the. swf file.Bycodingthe.swffiletolookforitsinstructionsinaneighbouringXMLfile,theXMLfileissuddenlyaneasy-accessgatewayintotheadtochangethecontent;beitaprice-point,ataglineorevenabackgroundcolour.Usingthismethodacreativedevelopercouldbuildonepieceofcreativewithanemptytextfield;thecreativecouldreferenceanfilecalled“content.xml” which could contain the English-written text to populate the text field,writtenoutinfull.

Providingthesetwofilestothetrafficker(advert.swfandcontent.xml)theycoulduploadbothfilestotheadserverforonecampaignandthenreplicatethetwofilesandreplacethecontent insidethecontent.xmlwiththesamecontentwritteninFrenchbeforeuploadingthetworeplicatedfilesintothecorrectplacefortheFrenchcampaign.

BeforeXMLfilescameintousethecreativedevelopercouldhavebeenpaidtoproducetwo pieces of. swf creative instead of the one. swf produced today, thus creative costscomedown.

Figure 3.4.1 the developer can update the ad indirectly with an Ad purpose built toreceiveinstructionsviaanadditionalXMLfile.Whentheuser’sbrowserrequeststheadfromtheserver(AdA)theadlooksforitsXMLfilewhichprovidesthelatestinstructionstoupdatetheAd.TheuserthengetsdeliveredtheupdatedAd.

ThisexamplesuggeststhatanXMLfilecanbepopulatedwithcontentbutitcanalsobepopulatedwithURLsor references to the locationof images or additional content.An.swffilecouldappeartocontainaseriesofimagesbutbehindthescenes,the.swffileisreferencingafilethatmightbecalledpictures.xmlandpictures.xmlcouldcontainalistofthe location of all the images set to appear the series in the ad. To change the imagesdisplayed in the ad, the traffickers could edit pictures.xml to introduce new images orremoveoldones.Inadditiona.swffilecouldbecodedtoreferenceanadditionalsectionwithinpictures.xmlthatcontainsanumberusedtocontrolthespeedatwhichtheimagesin theseries rotate (perhaps thehigherup thenumber, the faster the rotation).AnXMLcouldcontainawholeseriesofremotecontrolstoadjustjustaboutanythinggoingoninthe ad.Each set of controls exists in separate section of theXML file called a “node”,referencingtherightnodeintheaccompanyingXMLfilemapsthecontroltothefunction.

EventheXMLdoesnotneedtositalongsidethecreativeitisupdating.TheXMLcanbecompletelyremotesothatinanidealworldanXMLfilecanintheorysitonthedesktopofaplanningmanagerfromaMediaAgencyandtheycanupdatetheadfromtheirlocalmachine.(Toperformthisfeat thecreativeneeds tohavetheURLof theXMLsetuptolookfortheuser’slocalmachineasauniqueIPonanetwork).MorecommonlytheXMLcansitontheAdvertiser’sownsitebutineithercasetherearesecuritysettingsinplacetoensurethattheXMLcannotberemotelyaccessedoradjusted.A“Cross-Domain”fileisaverysmallfilethatsitsontheAdvertiser’stopleveldomainandgrantsaccesstothethirdpartyadservertoallowittograbtheXMLfiletomakechangestothecreative.Withthisdirect connect to the XML file, creative can be changed in real time (and at scale thechangewillbeliveeverywherewithin15minutesthankstoadserveradreplication).

Figure3.4.2With a cross domain file in place, an ad can securely call anXMLon anexternal server (such as on an Advertiser site) and as long as the ad server has beengrantedaccess,theXMLisfreetodelivertheinformationbacktotheAdwithoutopeningupanunsecurecommunicationchanneltotheexternalserver.

Asgreatasthissounds,creativedevelopersveryquicklyencounteredaproblem;changesmadetotheXMLfilecouldnotbepreviewedintheflashadbeforetheamendmentwaspermanentlyadded.Withoutapreview,newtextcouldbleedintotherestofthecreativeunintentionallyor imagescouldbe referenced thatdon’t exist and incorrect settings, setremotely, could cause the ad to becomeunstable.Furthermore allowing theXML to bechangedsoeasilymeantthatitwasnearimpossibletotrackchangesmadetothead(whenrepublishingfromaflashfile,developerstypicallyappendthenameofthefile,creatinganewversionofthesamead).WholetoolswererequiredfordeveloperstoreallymakethemostoftheXMLremotecapabilitiesandtheseweretoarriveonthesceneinaninterfaceasDCO.

DCO-DesignPhaseCreatingthedynamictemplate

ThemodernisedworkflowforDynamiccreativerequiresthatthecreativedeveloperonly produces one creative for each ad which can be better off considered a‘template’ creative (containing all of the ad possible designs and functions). Aseparate template would be constructed for each of the required placementdimensions for the campaign. The template essentially a blank canvas whichcontainselementsorassetswhicharedynamicinnature,meaningthattheyhavethecapabilityofchangingthewaytheylookandbehave.

Tobuild suchcreative, thecreativedeveloperutilises the thirdparty’s componentpackinsidetheflashauthoringenvironmenttocreatedynamictextbox’sandemptyimageslotsinwhichtextofafixedlengthorimagesofafixedsizecanbedroppedintoata laterdate.ThedynamicelementscantechnicallybeanythingthatcanbebuiltintheFlashenvironmentwhichincludestext,images,videos,sounds,specialvariables,files,stringsandintegers.TherichabilityofFlashtooffertheflexibilitytoworkwithsomanymultimedia formatsmeansCreativeAgenciesstillhave theopportunity to get very creative despite the fact that theymayonly be producingonetemplate.

Figure3.4.3AnAdcanbere-designedasatemplatebymakingtheareasthatwereprevious static assets into dynamic areas which can be populated with similarcontent.Thetemplate,whenpluggedintotheXMLdescribingthelocationofstaticassetsorstaticcontentcanbeusedinconjunctionwiththetemplatetoreproduceanewAd.ThisiscalledanAdVersionratherthananAdsinceitisaversionoftheoriginal.

DynamicCreativePlanning

At this early stage planning is very important; for instance, howmany dynamicelements should the creative developer put into the template? If one element isreservedforataglineastextbasedcopy,oneelementforthepricepointandonefortheimagetherearethreeelementswhichcanbechangedutilisingtheXML,butisthat enough?Somecreativedevelopersgo the extramilebymakingdynamic theimage placeholder for the Advertiser’s logo and even the small print terms andconditions.

Planning ahead aboutwhich elementswill be created dynamic is essential, sinceoncethetemplateiscreated,itcouldbedynamicenoughtobeusedforanalmostinfiniteamountof time,evenreusedforseveralcampaigns,withouteverreachingcreative fatigue by the browsing user. Planning and identifying the dynamicplaceholder is the first step, the second is to consider thepossibility of text copyleakingintotherestofthecreative.Ifcopyisreplacedwithalongerpieceofcopylateron,orifimagesneedtobeusedthatdonotfitthedimensionsofthead,thenthe creative planning process must make such considerations and the dynamiccapabilities of DCO creation will accommodate such needs. For instance, thedevelopercouldaddcodetothetemplatesothatthecreativeautomaticallyresizestextorimagestofitthedimensionsofthetemplate.

Figure3.4.4DynamicTemplates shouldbedesigned for the long termso thatanXMLcanbemappedtoanyoftheDynamicfeaturesandthuscontrolthem.Shouldthe Advertiser wish to control anything that was not made specifically Dynamiclateron,anewtemplateneedstobepublishedbythecreativedevelopersobuildingformost eventualitiesmakes sense. Each of these dynamic elements are: (1) Thebackgroundcolouroftheversion(2)Theplayercontrols;perhapsthereareseveralsetsofcontrolstochoosefromloadedupasanadditionaldynamicSWFfile(3)Adynamic spot for the logo for those particularly long term templates that mightoutlive the brand logo. (4)Particularly innovative developers thatmight animatepartoftheadwillconsideraddingadditionaltweensorpathsthatcanbeactivatedby the XML later on (5) A dynamic video (6) The call to action itself could beadjustedfortheactualclick-throughURLorthetext(7)AdynamicfieldforthefonttypesothatitcandynamicallyadjustedviatheXMLlateron.

Oncethetemplateisfinalisedthecreativedevelopercanuploaditlikeanyotheradto the third party ad server. The ad server will detect the use of the dynamiccomponentusedbythedevelopertocreatethedynamicelementsandthetraffickerworkinginthethirdpartyadserverinterfacewillbegiventheoptiontodecidehowmany “versions” of the creative to create. Third party ad servers and specificdynamic creative technologies can support the creation and management ofhundredsandthousandsofversions:truescalingforstandarddisplay.Subsequentlytheseversions canbe assigned toplacementsdirectly (just as standard adswouldnormally be assigned) or for scaled personalisation and targeting versions can beassignedtospecificaudiencetypes.

DCO-AdCreationPhaseVersioning

Versionsofadsaddstwonewlevelstothehierarchyinthethirdpartyadserver.Undertheadlevelexists theversionlevelandbeneaththis, theasset level.(At timeofwriting,adservers could not offer reporting or tracking capabilities below the version level to theassetlevelbuttheindustryisexpectedtoadaptquicklytoaccommodatetheassetlevelaswell). To focus now on the version level, there are three common ways of creating aversion (also called “variant”) ad from the template creative supplied by the creativedeveloper:

1. SingleVersioningorManualCreation

Some ad servers which support the creation of Dynamic Creative, provide aninterfaceontheplatformwhichallowsforthebasicdynamictemplatesuppliedbythe creative developer to be adapted into Dynamic versions, version by version.Suchinterfacesareessentiallycreativetoolswhichcanbeusedatraffickertoallowfortheinputofcopyintothenewadversion,theabilitytouploadorreferenceanimagefromwithintheversion,insertapricepointandalsotochangethecolourofthe text, theadbackgroundcolour, thesizeof the text (insome instances toevenspell-checkthecopy),beforetheversioncanbepreviewedandsaved.

The preview option is popular with Advertisers who like to proof read or vieweveryversionofeachadthatpotentialcustomerswillsee,sothattheycanprovidetheir approval before it goes live on the web. As with the preview option forstandarddisplayadsintraffickingstandarddisplaycampaigns,theapprovaloptionwas removed from many ad servers as the Advertiser approval chain became abottlenecktocampaigndeliverytimelines.

Despite a variety of supporting features, the essential point of manual versioncreationor“singleversioning”istoundertakethedesignandbuildofamanageablenumberofversions,typicallyforaverysmallcampaignoratleastacampaignwithveryfewadversions.SingleVersioningisnotascalablemethodandhasitsmeritswhen an Advertiser is playing with DCO for the first time. More seasonedAdvertisersandAgenciesmovethingsupanotchto“Massversioning”.

2. ExcelMassVersioning

Adservingprovidershave listened tobigAdvertisersandAgencies running large

scale campaigns with many placements and sites by providing bulk tools whichutilisetheeasyreplicationofrowsofdatatoscale.Thesamemethodologieshavebeen extended to Mass Versioning in the absence of a better designed UI forcreatinga largenumberofuniqueadversions.Usinga spread-sheet layout insidetheadserverinterfacethetraffickercanbegintomakenewadversions,rowbyrowfroma single supplied creative template.Thecolumnsare alreadypre-definedbythe ad server interfacewhich identifies the dynamic elements of the creative andconvertsthemintocolumns.Eachnewrowrepresentsanewversionoftheadandeachcellcanthereforebecustomisedwithavalue(suchasawholenumberforapricepoint,atextforthecopyandURLstopointatexternalcontentlikeimages).Aspreadsheetviewquicklyallowsfortheupdateofmanyexistingversionsandadds‘copyandpaste’capabilitiestocreatenewversions.AswithSingleversioningtherearechecksinplaceforspellingaswellaspreviewfunctionsforeachoftherowssothat onmass, preview links can be generated and sent over to the advertiser forthemtoapprove.

3. MassVersioningfromaFeed

The creation of hundreds of ads is a common demand for Advertisers and theirAgencies where a vast product line is being supported. To build separate adsfeatures different products in each one is completely unattainable for the likes ofretail Advertisers which is where DCO can come to the rescue. Such productcataloguesareusuallyalreadyaggregatedontheAdvertiserwebsiteandthereforeaproduct listor“feed”exists (behind thescenesof theAdvertiserwebsite).Suchalistmightevenbeascomplexasadatabasewithinformationthatmightbeupdatedvery frequently such as the colour of the product in question and the number ofitemsleftinstock.

InturnthisfeedallowstheCMStodynamicallygeneratewebpagesandthesamemethodcanbeused tobuildoutversionsofads in their thousands.Theadservercontains controls in the interface which allows the trafficker to connect to anexternal database or feed and generate ads from it. The controls also permit thetraffickertomapvariablessuchasthestockcontroltotheversionandcreatearulethatremovestheversionfromadservingwhentheproductrunsoutofstock.Ifthepricepointorstocknumberismanuallychangedinthefeeditshouldreflectintheservedversion.IngeneraltheavailableDCOsolutionscoupledwithadserverinthemarket range in complexity. Flashtalking for example allows the trafficker tobrowsetotheAdvertisersitefromwithintheadserverinterfacetoselectaproductfeedthatmightalreadybedisplayedontheAdvertisersite.DoubleClickallowsthetrafficker amuchmore integrated setupwhen connecting toAdvertiser feeds thathavebeensetupusingGoogle’sownstockcontroldatabasetools.

Figure3.4.5afeed(suchasanAdvertiser’sproductfeed)canbeusedtogenerateamassnumberofversions.VersionsdeliveredtoPublishersitesarethosegeneratedfromthefeed.

Sometimesproductfeedsanddatabasesarecomplexorpartitionedoutandrequireadservingprovidersupporttobuildoutanewfeedfromwhichtocreateversions.Inadditionifproductfeedsarelikelytochangefrequently,theDCOtoolneedstobeconfiguredto‘check’thefeedwiththesamefrequencysothatdynamicvariableslikepricechangesarereflectedasclosetoreal-timeaspossible.

UsingaWebServicetorestrictloadonAdvertiserhostedProductimages

Amassversioningcampaignwithaproductfeedwill typicallypull the imagesofthe product into the ad version from the Advertiser website directly and notnecessarilythethirdpartyadserver.ThisisdonebyhavingtheURLlocationoftheimageincludedintheproductfeedandmappedtoacorrespondingdynamicimageplaceholderinsidethemasteradtemplate.

Hosting imagesandcontent isacommonquery thatcomesup in theDCOsetup.After all, with so many images to reference thousands of products; can theAdvertiser not simply direct the ad version to retrieve the image direct from theAdvertisersiteeachtimetheadiscalled?

Acoupleofconsiderationsneedtobetakenaheadoftime.Thefirstconsiderationisthatanadserverisapieceoftechnologythatispurposebuiltformasstraffic.IftheversionsandimagesthemselvesareexpectingtobereferencedbyeveryenduseronthesamescalethentheAdvertisersiteneedstobesetuptohandlejustasmanyadcallsor riskbeing takendownbyusers,demandingcontent fromtheAdvertiser’sownservers.

Togetaroundthis,theimagesthemselvescanbepulledfromanAdvertisersiteandcached bywhat is referred to as a “web service” technology sitting between theAdvertisersiteandtheadserver.ImagesarecachedbythewebservicewhichtakestheriskofhighvolumetrafficawayfromtheoriginalhostURL.Cachingimagesin

thiswaycomeswithitsownriskssoitworthfullyexploringthistopicwiththeadserver in question before setting up and running such a campaign. In such ascenario,anadaptedfeedwouldneedtobeappendedtotheexistingproductfeedsothatnewURLsfortheproductimagesarereferencedbythedynamicfieldintheadtemplate.

Figure3.4.6Hitting product feeds frequently to keep the versions up to date can be aheavystrainonsomeAdvertisersitesorservers.TogetaroundthisproblemaWebServicecanbesetuptoreplicatetheproductfeedontheadserver.ThewebservicewouldmakearequestthecoreXMLonalimitedfrequency(1)forexampleonceperhourinsteadofforeveryadcall.Therequestreturnsaresponse fromthe feedvia thecrossdomain(2) forsecuritypurposes.ThewebservicethenupdatesthevariouspartofacopyoftheXML(3).WhenadversionsarerequestedbytheusertheversionsareuptodatewhichwhatisinthecopiedXMLontheadserver(4).

FeedsthemselvesarealsorarelyreadytoplugdirectlyintoaDCOsetupontheadserversidesoWebservicescanbeusedtoscrapetheoriginalproductfeedatdifferenttimesandactasanadaptedfeedforthecreationoftheDCOversions.FeedScrapingisnotusuallyafreeservicesoitisworthanAdvertiserorAgencyfullyinvestigatingfeedadaptationandscrapingwhenworkingouthowmuchitwillcost tocreateandsetupamassversioningcampaign.AswithlegacyXMLcreativesetup,ifafeedsitsinsideanAdvertiser’sdomainor firewall itwill require thepresenceofacross-domainwhichaccepts thecalls toandfromtheadservertoworkcorrectly(adheringtotherequiredsecurityprotocolstoaccepttheadservertraffic).Theuseofawebserviceseradicatesthereal-timeelementthatwasso attractive about XML dependent creative, the amendments to the version will onlychangewhen thewebserviceupdatesacopyof the feedor recaches the stored images.Thisprocessofscrapingatthetimethisbookwaswrittenwasaroundonehourmaximum,soachangetothefeed,shouldItchangeaversion(suchaswithanupdatetotheamountof available stock on a product) would only be reflected up to an hour later. Some adservershaveevenlongertimelagsonthis.

Withthefreedomtocreatesomanyversions,comestheresponsibilityofidentifyingeachof those versions individually. Each version is assigned an ID in the ad server forreference purposes, since the only other real unique identifying marker is literally theuniquecombinationofdynamicvariablesthatmaketheversionuniqueinthefirstplace.Ifaversionischangedoradjusted,itsuniquecombinationsofdynamicvariablesnolongermatches the original configuration and this is a problem for reporting. In truth anyadjustmentof aversion should create awhollynewversion so that theversions canbediscernedinadserverreporting.Inturnthecreationofanewversionaswiththecreationofanewadrequiresacertainamountofre-assignmenttoplacementsoraudienceswhichona largescale(suchasusingaversionacrossa largenumberofplacementsandsites)canprove tobea traffickingheadache.Themostrecentupgrades toDCOinterfacesforthirdpartyadserversarebuildinginthefunctionalityandlogictocopewithassignmenton this scale such as by associating versions that have been adapted from one another,together.

DCO-AssignmenttoPlacementPhaseJust aswith standard ads, adversions canbe assigneddirectly to placements.Once theversionshavebeencreated in the ad serving interface the trafficker ensures there is thecorrect configuration of placements, sites and delivery groups in order to attach the adversions.

Placementvs.Audiences

Beingabletoscalecreativemeansthatverydeeppersonalisationbecomespossible,sinceaversionofanadcanbecreatedwhichhasmorerelevancetoauser, thanarun-of-the-mill standard display ad (designed for mass reach) would have. Relevance generallybringsanimprovementincampaignperformanceespeciallywhenDCOiscombinedwithasourceofaudiencedata,whichbringsabouthighlycustomisedtargeting(seeChapter4-AudiencesandTargeting).

Asexplained,insidetheadserverthetraffickerhastheopportunitytoassignadversionstoplacements(as ispossiblewithStandarddisplayandRichMedia)butwithDCO, thecapability for assignment extends so that across placements the same version can bedisplayedtomultipleaudiences.

Byassigningversionstoaudiencesthetraffickerexposesasingleadversionpotentiallytoa much larger group of users that might be exposed to a single placement. This isparticularlytruewhenamediaplanfeaturesalargenumberofsitesandwithinthatavastnumberofplacements.

AllusersarrivingatasinglePublishersiteandseeingasingleplacementwillbeasmallernumberofusersthananaudienceofusersencounteringadsacrossseveralplacementsonseveralsites.Sinceanadservercanidentifyusersinanaudiencegroup,whenevertheadserver serves to the audiencemember, their presence in the audience can over-ride thedefaultcapabilityofservingtheaddesignatedonlyforthatsiteandplacement.

Thiswillonlyworkifthedefinitionforanaudienceisfairlybroad.Iftheaudienceistoogranular(suchasusersthatliveinacertainZipCodewhohavevisitedanobscurepageontheAdvertisersite)thentherulessetupbythetraffickerontheplacementwillnotfindthe

matchingaudienceandwillthereforeserveanon-targetedadoradversion.

Settingupthousandsofveryspecificaudiencegroupswithcorrespondingadversionsisverytimeconsumingandifonlyacoupleofuserseverencounteraversion,therewillnotbe enough data to suggest that the versioning of the ad is having an influence on theirdecisiontoclickforinstance.Ifacoupleofusersalsoclickonaversionwhichisalmostidentical,howwouldoneworkoutwhichcreativeversionisthebestperformer?

Figure 3.4.7 if versions are assigned to placements (where one placement is on eachPublishersiteforinstance)thenthenumberofindividualsseeingeachversionwillbelowanddecidingwhatthebestversionisbecomesmorechallengingasthesampleissmaller.Applyaversiontoanentireaudienceandaversionisshowntoawidergroup.Thisisnotalways the case (sometimes audiences are smaller than the traffic going to a singleplacement)buttogetthemostmeaningfuldataset,theversionshouldbeappliedsothatitisseenbyasmanyusersaspossiblewithasfewversionscreatedaspossible.

TargetingdifferentVersionstodifferentAudiences

AsMassversioningallowsforavastcombinationofdynamicelementstobeturnedintoadsoftheirown,thenextstepistoinvolvethemediaplanningteamfromtheAgencytodecide which of those ad versions are suitable for which audience type (since theassignment to placements alone is circumvented). This means defining the effectiveaudiencesintheadserverandthenassigningthecorrectversionstothecorrectaudience.Themedia planning team are important in this step since themedia is often bought tomatchtheaudiencetargetingrequirementsprovidedbythemarketingteam.

Audiencescanbeassembledordefinedinavarietyofwaysforbothdynamiccreativeandforstandardcreative.Chapter4deepdivesfurtherintotheavailableaudiencegroupsfortargetinginsideathirdpartyadserver(somethingthatdiffersfromtheaudiencetargetingcapabilities offered through a Publisher ad server). Of these types, the “retargeting”audienceisthemostcommonforwhichasingleadversionmightbehighlyrelevanttoauser.

Aretargetingtag(aswillbefurtherexplainedinChapter7)can,forexamplebeplacedonanAdvertiser’sproductpage:invisibletotheuserandallowtheusertobetargetedagain

lateriftheyleavetheAdvertisersitewithoutpurchasing.Theresultisthatanadversion(created using mass versioning) matching the last product the user saw, can appear toeffectivelychasethemaroundPublishersitesuntiltheymakethepurchase.ToenablethechasethemediaplannersprocuretheinventoryrequiredacrossPublishersites.PureDCOsolutions like Criteo do all this legwork for the Advertiser in question, buying theinventory on their behalf with the intention to chase down such a sale, using massversioningtokeepthecreativerelevant.

Figure3.4.8AudiencesvisittheAdvertisersite,inthisexampletheproductpagewherearetargetingpixeladdstheircookietoacookieretargetingpool.WhentheaudiencelaterencountersanadspotboughtbytheAdvertiseronaPublishersite,theadcheckstoseeiftheuserbelongstoaretargetingpoolandifsoshowsthemtherelevantversion.ThisisallpossiblewithstandardadsbutDCOallowsforthisatscale.

ConnectingDCOwithAudienceBuying

Theabilitytodeliveracustomisedversionofanadbasedpurelyonauser’sactionsmeansthatDCOcanbeusedinconjunctionwithmachine-basedaudiencebuyingsystemsintheRTBspace.Thisisasignificantlymoreadvancedconcept,whichiscoveredoffinmoredetailinChapter8.

PinpointingauserandshowingarelevantpieceofcreativeintheformofanadversionhasspunoffintotwocoretechnologyfeaturesDynamicRetargetingandCarouselAds.Todigest both of these two technologies requires a deeper understanding about retargetingtags (and though these are features of DCO ads) the detail about their setup andimplementationaredetailedinChapter7-Retargeting.

DCO-OptimizationPhaseAuto-Optimizevs.Manualcontrol

Oncetheversionsarecreatedandassignedtoplacementsandaudiencestherecomesthequestionofhowtodecidewhichversionstoshowtheenduser.Althoughpersonalisationcomesintothegameforversionscreatedusingcorrespondingproductfeeds;inthecaseofbuilding out hundreds of versions of ads that have different creative nuances (such asdifferent font andbackgroundcolours) it is lessclearwhatversion to show towhoandwhen.Sophisticated rules are requiredbecauseoptimizingmanyadversions is abiggertask thanoptimizing standard ads.Withversions en-mass itmakesmore sense to allowmachine-learning to “optimize” the versions so that those that are performingwell areexposed to a wider effective audience and those that are not doing so well simplydisappear.Havingsaidthisitisworthcheckingthatthethirdpartyadserverthatsupportsdynamiccreativeallowforbothmanualandautomaticoptimizationssincethesefeaturesarenotuniversallyavailable.

Thecontrolofweightingorsequencingads,orrunningdifferentadversionsatdifferenttimeswithvarious frequency caps, areoptions that remainopen inDCOas they are instandard display and are effective tools in refining a campaign (with a small enoughamount of versions for the target audiences seeing those versions to not become toogranular).

OnceaDCOcampaign is setupmanuallyby the trafficker; theautomatedapproachcanprovetobeapowerfultoolifwellmanagedatarms-length.Autooptimizationhassomecorefacets.

1. Firstofallthesetupneedstoconsiderhowtooptimize

a. Will the machine learning adapt according to an A/B style test (seeChapter4)?

b. Orusemulti-varianttesting(seefurtheroninthissection)

2. Then a consideration needs to bemade on the core successmetricwhichmay forexample:

a. preferaleantowardsclicksoverconversions

b. orfavourahigherbasketvalueinconversions

c. Orconvertersfromaparticulargeography.

3. Thelastconsiderationistheeffectivepoolfortheoptimization,sothiswouldbetodisplaytheversionorcollectionofversionstobeshownjusttousersarrivingatagivenplacementoroptingforawidergroupbytakinglearning’sacrossacompleteaudience.

Thetraffickerselectstheplacementfromthesiteinquestionintheadserverinterfaceandattaches the versions in a group, after selecting the preferred success criteria from theabove. When the campaign goes live, those more successful ads will be shown morefrequently or receivemore valuable user engagements and this can all be displayed inreporting.Themachinedoesneedtimehowevertolearnbasedontheuseractivity,whichcreativeisthebesttodisplay.Thiscantakeacoupleofweeksormoredependingonthecomplexityofthecampaign.

Figure3.4.9Auto-optimizationatwork.Thesuccessmetricsisthetotalnumberofclicks.Highnumberofclicksinthisexampletellstheoptimizerwhat%oftheaudienceshouldsee the specific ad version in the following week. In the first week the weights arebalancedbutbytheendoftheweekAdAisgettingthemostclickssoinWeek2mostuserssee Ad A. However by the end of week 2, the small portion of users seeing Ad B areclickingonitatahigherratethantheusersseeingtheotherAds.Thereforeinweek3theup-weightissettoAdB.Bytheendofweek3AdAseemstobegettingthehigherrate.The learning from this auto-optimization that Ad C can probably be dropped from thecampaignasaversion.

Auto-optimizationasa“BlackBoxTechnology”

Some Advertisers ask to see the output of the auto-optimization engine making itsdecisionsbutnotalladserversofferthisasanoutput.Thisisacontroversialpointsincethelearning’smadebytheauto-optimizerarenotsomethingthattheAdvertiserorAgencycan retain and apply to other brands (in some instances the learning’s cannot even betransportedtonewcampaignssincethecreativeversionsusedmaybedifferent).Thereismerit in being able to export this data although as yetDCO ad serving technology hasbeen slow to visualise the learning’s and give the trafficker the control to over-ride theauto-optimizationprocesstoimproveitbasedonmanualoptimizationknowledgeonceithasbegunoptimizing.

Multi-varianttesting

GetreallygranularwithDCOandthefocusisnotonthecreativeasacompleteentitybutrather the tiny fragments inside the creative otherwise known as Assets. Multi-VarianttestingallowstheAdvertisertomakedecisionsabouttheeffectivenessofindividualassets(for example; if a blue background on any version is alwaysmore effective than a redbackground).ThisgranularityallowstheCreativeAgencytobuildamuchmoreeffectiveadfromthegroundupwithoutnecessarilyoptingforDCObutrequirescontinuoustestingacrossalargecontrolgrouptoremainrelevant.Multi-varianttestingispopularwherethecreative development team ismuch closer to the trafficking and reporting process. Theeffortrequiredtomaintainsomethingsosubjectiveinnatureistoohighapricetopayforavery small improvement in conversionuplift.Adirect responseAdvertisermayclaimthateveryadditionalsalemattersbutif theworkrequiredtobringthatsalehikesuptheoperationalcostsofdevelopingtheDCOtemplateortheworkofthetraffickerpastthatofthegaininacoupleofsalesthenitisnotworththeeffort.

WehavedivedintothesetupofDCOcampaignsthatbothuseafewversionsandthosethat operate across mass versions. Trafficking and optimizing these campaigns is alsofairly simpleprovidingplentyof testingandpracticegoesonbefore thecampaigngoeslivewithDCOactivity for the first timeand thecampaigngoalsandaudience targetingrequirementsareclear.DCOadaptswelltobothbrandAdvertisers(toimproveandscalecreativecopyandcontent)andfordirectresponseAdvertisers(forpersonalisedaudience

targeting,retargeting,andcreativeoptimizationforthepurposeofdrivingmorevaluableuserengagements). In thenextsectionwebegin todrifta little furthera-fieldstill fromStandardDisplayadservingandevenbeyondRichMediaintothecampaignmanagementfor Video otherwise known as Instream and Pre-Roll; a certain future for an internetinfrastructurethatprovidesuserswithanevermoreimmersiveadvertisingexperience.

Chapter3-Section5InstreamVideo• WhatisInstream?

• Figure3.5.1

• PreRollMidRollPostRoll

• Figure3.5.2

• SiteServedVideo

• Figure3.5.3

• WheredoVideoAdNetworksaddvalue?

• VAST-thevideotagstandard

• CompanionAds

• Figure3.5.4

• AdSkippingandTrueView

• VPAID-StandardisedInteractiveVideo

• Figure3.5.5

• VPAIDAuthoring

• Figure3.5.6

• DCOforInstream

• Figure3.5.7

• BeyondVASTandVPAID

• PublisherCertification

• Figure3.5.8

• TraffickingInstreamads

ThirdPartyadservingprovidersconsiderInstreamvideotobethefastestgrowingchannelindigital. InstreamVideoisperhaps thedistantcousinofStandardDisplayandacloserrelative to the Rich Media channel. The trafficker actually has a much easier jobtrafficking Instream Video ads than the job required to undertake DCO or Search forexample,themorecomplexlegworksitswiththePublisher.

ThissectionwillstartoutbyexplainingwhatInstreamisbeforeexplaininghowitisstillinitsinfancybeforeprovidinganunderstandingaboutthetagstructure,thecapabilitiesoftheads in termsof trackingand reportingaswell as thecompletepicture forcampaignmanagementfromtheperspectiveofthetrafficker.

WhatisInstream?

Instream ads can be thought of as the evolution of TV advertising in many instances,

Instream ads are literally an Advertiser’s TV ad embedded into the video player of aPublisherwebsitedesigned toplayoutvideocontent.TheadsareusuallyshownbeforethefeaturedvideocontentshownbythePublisher.

ReferringtoInstreamasVOD(VideoonDemand)isalittleinaccuratesinceVODencapsulates the ability to stream video whenever required and would includeplayingvideo content fromset-topboxes andpaid for streaming services such asAmazon’sLovefilmorNetflix.None-the-lessitisstillcommontohearthoseworkinginonlineadvertisingtorefertotheInstreamchannelas“VOD”.

Figure3.5.1anexampleofanInstreamVideoad.TheInstreamadplaysasaTVadwouldplay, before the featured video content in the same player. As with most Instream ads,someofthemainvideocontrolsaredisabledtopreventtheuserforwardingthroughtheadtothecontentfasterthanitisplaying.

Publishers offering Instream content are on the increase but the globally availableinventoryisstillsomewhatlimited.ThebiggestsitesforVideoattimeofwritingincludeGoogle’sYouTubeandintheUSsiteslikeHulu.Allofthemajorbroadcastersalsoofferin-browservideoplayersandgrantadspotstoAdvertisersofteninconjunctionwithdealsdesignedtobuyTVadspace.

TocreatetheInstreamads,AdvertiserswilltypicallyusetheverysamecontentwhichwasfilmedforTVbutsomeareconvincingtheirproductionteamstocreatevideospecificallydesignedfortheweb.Theadswhichareproducedtodaytendtobefifteenseconds,thirtysecondsoraminutelonginplaytimeandareconvertedbyspecialisedproductionhouses(oftenthefunctionofmoreadvancedCreativeAgencies)intodigitalvideosuitabletobeusedonline.PureTVvideoisusuallycreatedataveryhighdefinitionwhichneedstobeconvertedbefore it canbemoreeasily transportedaround.Thereare several companiesoperatingintheTVassetmanagement,distributionanddeliveryspace;bigplayersincludeExtremeReachandAdstream.

OncethevideoisdigitiseditneedstobefullyadaptedtoPublisherspecificationsasonewould expect from ads in both the display and rich media channels. Then it can beuploadedtotheadserverandtraffickedoutbeforeappearinginfrontofauserasapre-roll,mid-rollorpost-roll.

PreRollMidRollPostRoll

Despite the fact thatvideoadsservedbefore the featuredvideoare themostcommonlyusedInstreamformat(otherwiseknownasthePre-Roll), therearealsoinuseMidRolls(middle of the featured video) and Post Rolls (shown after the featured video). Theseformatsallexistinthe“stream”ofthevideoitselfandarereferredtoas“Linear”Instreamadsmeaningthattheadandthevideocontentallseemtoplayinonecontinuousstreamfrombeginningtoend.

“Non-Linear”Instreamadsallowtheusertoperformadditionalactionssoastopotentiallyextendthepossibleengagementperiodwithinthecreativebeyondthefifteensecondsofthecorevideo.Providing that thePublisher specificationspermitnon-linear, thiscanbeachievedbyintroducinginteractiveelementsintothead.

Figure3.5.2LinearVs.non-linearads.Linearadsplay through frombeginning toend,only thevideo itselfwhenclickedwill take theuser through toa landingpage inanewwindow.Withnon-lineartherecanbeadditionalclickableandinteractiveregionswhich,whenrolledoverorclickedmightpausethevideoopeningpanelstoamini-websitefortheAdvertiserinsidethevideoplayeritself.Thearealabelledas‘F’mightclickthroughtotheAdvertiser’sFacebookfanpageforinstance.

Toguaranteethatvideosaresuitablefor thePublisherorBroadcaster,mediaplannersatAgenciesandsomeAdvertisermarketershavebeencircumventingtheuseofthirdpartyadservingandpassingthevideoassetsdirectlytothePublishersiteforimplementation,distributionandserving.ThisbehaviourgiveslittleregardforthedamageitisdoingtotheunderstandingthetruevalueoftheinventoryinquestionandidentifyingtheadcampaignvaluebeyondsinglesitesortheInstreamchannelinisolation.

SiteServedVideo

Video heavyPublishers and digital broadcasters have been notably resistant in grantingpermission to have video ads served by an independent third party. This thinking andpractice is reminiscent of the internet in the early years when many Publishers. SomebroadcastersmaynotevenbeawareofthevaluethatthirdpartyadservingofferstotheAdvertiser’s funding their valuable content and with a well a more well establishedindustryinTVadtradingthaninonlineitisnosurprisetofindsomeresistancetochange.

To remind the broadcasters, third party serving allows for the de-duplication of cross-Publisher reach, frequency and the de-duplication of conversions (see Chapter 4),independent tracking and auditing as well as standardised IAB approved reportingmethodology across Publishers and video players. The Publisher’s excuse to site servesaysmoreaboutwhatthatPublisheristhinkingandpracticingthantheymightrealise.

Figure3.5.3ServingdirecttothePublisherorbroadcaster’svideoplayerwithoutathirdpartycreatesblindnessinthedataandanoperationalnightmaretocollatethegenerateddataattheendofthecampaign.Theflawsarelistedfollowingthisdiagram:1)therearedifferentcountingmethodologiesforthevideobecausedifferentPublisheradservershavedifferentcountingcapabilities,2)thereisnoindependentauditingofthecollecteddatasoitcouldbetamperedwithbythePublisherintheirfavourwithouttheAdvertiserknowingand3)theprocessgeneratesmultiplereportsinvariousformatsthatmakelifedifficultfortheplanner/marketer/analystreviewingtheresultingreports.

ThetrendamongMediaAgenciesforsometimewastoserveInstreamfromthirdpartyadserversbutcutcostswhilestill retainingsomedegreeof transparency;by implementingtrackingpixelsandclick trackerswhere thevideo itself is site served.Thismethodstillcreatesadegreeofblindnessanddoesnotpreservetheindependentverificationoftheaddelivery in theadserver.Atrackingpixelfor instancecannotconfirmifavideoadwaseverseen,everplayedoreverskipped.Thedemandfromtheindustryhasbeenforbetterstandardisationand the IAB arrived insupportof two technologystandards for Instreamwhichensures:

• ConsistenttagimplementationamongdiversePublishersandplayers

• TheabilitytostandardiseInstreammetricsandcorefunctionalityamongcompliantvideoadservers

WheredoVideoAdNetworksaddvalue?

IfanAdvertiserchoosestorunacampaignacrossasinglenetworkthenthereisnoqueryof de-duplication because users are confined to seeing ads served from one place andtrackedbyasinglesystem.ForthisreasonNetworkstrytohaveasmuchreachaspossibleandintheVideoworldthisisentirelypossible.Videology isanexampleofahighreachVideoNetwork.Thirdpartyad-servingcontinuestobeusefulonjustasinglenetworktoindependently verify the delivery and tracking of these ads. Until very the very recentstandardisationofInstreamadtagsandtechnologies,thirdpartyadservingadswasnearimpossiblebecauseof themanydifferencesbetweenPublisherplayersandthewaytheyhandle,process,playandtrackvideo.ThestandardpassedbytheIABwasVASTandifaPublisher video player is not certified to run VAST or VPAID ads it is best to avoidplanningandbuyingtheirvideoinventorybecauseitcannotbeeffectivelymanagedwithathirdpartysystem.

VAST-Thevideotagstandard

VASTstandsforVideoAdServingTemplate, thirdpartyadserverswilloutputaVASTtag (which looks like any other ad tag) telling the player it is implemented in,what to

display.Thebroadcaster’sdigitalplayerwillsendanadrequesttothethirdpartyadserverwhichwillrespondtotheplayerbysupplyinganXMLfilethatitcanread.Reportingisthenaggregatedintheadserverasusual.VASTsupportsLinearads,Non-Linearadsand“companionads”.

CompanionAds

Instreamadsintheplayercanbeaccompaniedbyadsonthesamepagewhichtendto appear near thevideo content.LikeRichMediaSyncAds, companion ads areusedtocomplementthevideocreativeortomakeitappeartobemoreinteractive.ThePublishercanusedifferentpartsof thead tagcodefor the traffickedInsteamtagtoassemblethecompanionadnexttothevideocontentwhenitplays.

Figure3.5.4AcompanionadisservedontothesamepageastheInstreamvideo,commonlyinareservedadspotnexttotheplayer.

AdSkippingandTrueView

Someoftheadsproducedcanbeskippedafterafewsecondswhileotheradsplayallthewaythrough,disablingthevideoplayercontrolsthatwouldusuallyallowtheusers to fast forward. In recent years Youtube has presented users with a buttonallowingtheusertoskipthewholead.ThistechnologyisrolledupintoaproductofferingcalledTrueviewsothatAdvertisersonlypayuponanentireviewingofthead.TheinnovationofGoogle’sskipfunctionalityhasbeenincludedinthestandardsforVAST3.0whichmeansthatincreasingly,deeperinteraction,startingwithaskipwillbeavailableforAdvertiserstoimplementonallads,evenoutsideofYoutube.

In addition to VAST tag standard there is an advanced player communication standardcalledVPAID.

VPAID-StandardisedInteractiveVideo

VPAIDstandsforVideoPlayerAdInterfaceDefinitionand isastandardisedmethodofcommunicationbetweenthevideoadandthePublisher’splayer.Oncethetagisembeddedon thePublisher side, thevideoplayer calls the ad server to request thedeliveryof the

video ad itselfwhich is then streamed in.TheVideo player than performs an on-goingcommunicationbetweentheadandtheplayeruntiltheadplaysallthewaythrough.Ifleftuntouchedbyanyuserengagement,theVPAIDadwilltypicallyplaythroughasthoughitwereanyotherlinearadbutengagewithitandthevideomightsloworpausetopresentnewinformationtotheuser.Throughengagementsandvideoplays,theadreturnsmetricinformationtotheadserver.UnlikeVASTwhichcontainsaroundfifteenstandardmetrics,VPAID(likeRichMedia)addsalmostlimitlesscustomisationascustommetricsontopofadditionalstandardmetricsaswell.

Figure 3.5.5 A example of a VPAID ad. Collected metrics are stated in grey italics,additional functionality beyond standard linear VAST shown in black text.With VPAIDthereisalotofdatathatcanbegleanedaboutauser’sinteractionwiththevideoadasitplays,manyofsuchactionscandisruptthelinearplaytopausethevideo.Functionssuchas imagegalleriesmightopennew images in thevideobuton topof theplayingvideo.Share toFacebookmayopenanewminiwindowwithin theplayer toallow theuser topushaposttoFacebook.TrackingforVPAIDopensupthesortofoptionspreviouslyonlyavailableinRichMediasuchasinteractiontrackingandmoregranularvideocompletionmetrics.

VPAIDAuthoring

The development of VPAID ads sounds complicated to some Media Agencies andAdvertiserswhoareunawareofpreciselyhowstandardised these formatshavebecome.ElementsamongVPAIDadsaresocommon(suchasaddingacall-to-action,addingaliketoaFacebookFanpage,YoutubepageorTwitterfeed)thatacreativedeveloperdoesnotevenneed to be involved in the process of converting aTVad into a fully functioninginteractivecreative.Thisisdowntoawholeheapofavailabletoolswhichareinfactsmallauthoringenvironments(mimicsFlash-likeauthoringcapabilities insomeinstances).AdserversarenowbeginningtomarketsuchtoolstoAgenciesandAdvertiserstocutdownon creative costs but the same sort of tools are extended to specialist video Publishernetworksaswell,theseincludeVideologyandInnovid.

Figure3.5.6AVPAIDAuthoringtool;allowsthetraffickertotransformanexistingvideocreative into an interactive one. No coding experience is required since the interfaceallowselementstobedraggedontothevideostagewheretheycanbecustomisedandtheresultingdataoutputscanbetrackedviaaninteractionfeedpanel.Thecompletedprojectcanbemadeintoanadattheclickofabutton.

DCOforInstream

Aswithother formsofDisplayadvertising(StandardDisplay,DCOandMobile) theadservercanbeusedtomakeadecisiontoserveaparticularcreativetoaparticularaudiencetype (seeChapter 4). If a flurry of audiences can be deduced, how does one go aboutamendingafixedpieceofvideotocreatedifferingversionsofthecreative?

TheeasiestsolutionistouseVPAIDauthoringtoolstoproducestaticframesinthevideothat appear tocarryadifferent creativemessage suchasdifferentpricepointorwrittenreference to a location. Some tools have emerged however, that allow the trafficker toappendthevideocreativebylayeringtext,imagesandadditionalvideofootagebothonalayerwhichappearstositabovetherestofthevideoordeeplynestedinthevideoitself.Thelatteroftheseoptionsproducesanastoundingeffect,mosteasilyimplementedwhenlayering text or images onto a blank canvas such as a wall that might appear in thebackgroundofthevideo.

The addition to the video appears to display inmultiple frames and inmultiple anglesbecausethetextorimageiscarefullymappedtoothermovingobjects.

Figure3.5.7DCOisavailableforInstream.Evenwithvideothathasalreadybeenfilmedormade,thefootagecanbeadaptedbyinsertingadynamicareaintothevideothatevenchangeswith the lightandmovementof the footage.Thisareacanbemapped tootherimages or 3d objects as assets; the outcome is new versions which can be applied toaudiencesorplacementsjustaswithDCOforstandardDisplay.

Technology permits scalability by allowing the user to create many versions, insertingdifferent,relevantmessagesintothesesubtlelocationsandexportingdifferentversionsofthe video in doing so. Such technologies are in their infancy and therefore are notaltogether cost effective at this advent. On example in particular: Impossible Software;have the ability toplug such a feature into ad serving technology. In thisworkflow thecreativedeveloperwouldpassthesinglevideo(whichwouldbeusedasthetemplate)intotheuserinterfaceofthevideoversioningtoolandthetraffickerwouldcontroltheoutputoftheversionsintothethirdpartyadservinginterface.Subsequentlytheseversionswouldbeassignedtoplacementsandaudiences(seeChapter4).

BeyondVASTandVPAID

Digital Advertising is overly ripe with acronyms and confusingly a VPAID ad can beservedthroughaVASTtag.Theimportant thingtoremember is thatVASTisa taggingstandard whereas VPAID is more like a format or a standardised way to describe thefunctionality of an interactive video ad. As of publication the so-called standardisationefforthasgivenbirthtoabroodofnumbersandlettersthatwithoutexplanationhavelittlelogicallabelling:

• VAST1.0

• VAST2.0

• VAST3.0

• VPAID1.0

• VPAID2.0

Without getting too technical each is a significant improvement on the other but sincethereareseveralstandardsforvideoincirculation, thirdpartyadserversneedtocertifythat adswill display and functionwhen runwith premiumPublishers and broadcasters.ThisisbothtocheckthattheclaimsthataPublishercansupportbothVASTandVPAIDandifsotoseewhichversiontheyaresupporting.Inturntheadserverneedstoensureitsupportstheoutputtothecorrespondingstandards.

JustasRichMediaPublishersusedtorequirecertificationtheprocessremainsthesame.

When amedia plan becomes available from theAdvertiser direct or theAgencymediaplanner,thelistofPublishersshouldbesuppliedtothethirdpartyadservingprovidersothat, if required, a certifications team can perform certifications.Certifications are onlyrequired on new sites and players butwhere these exist, the timeline for the campaignmanagementandsetupprocesswillincreasetoaccommodateafullcertification.

PublisherCertification

InInstreamaddelivery,thevideoplayerenvironmentonthePublishersitethatrendersthead tag is not standardised, not everyone is using the same kind of player technology.Therefore each Publisher site needs to be certified to ensure that third party servedinstreamadsarefullyfunctionalintermsofdelivery,verificationandreporting.

Effortshavebeenput inplaceby the IAB andother interestedparties to standardise thedeliveryofInstreamadsindustrywidewithlimitedsuccess,aclaimofusageoftheVASTstandard by a Publisher is no guarantee the practice is being undertaken.Unfortunatelyduetothesheervolumeofvideoplayertechnologiesinuse,eventhethirdpartyadserverhas extended processes around running video campaigns and media plans must bechecked.Ifnewsitesandnewvideoplayersappearonthescene,testingandcertificationaresensibleundertakings.

These processes are common place when a channel is in its infancy (and the mobilechannelrequiressimilarcertificationtoensurethatthethirdpartyadservertrackingwillwork. The key is to test ad tags so that there is no discrepancy between ad server andPublisher reportingbefore thecampaigngoes live.ThiswayPublisherscanoptimizeaddeliveryso that there isnooverorunderdeliverywithalmost identical reportingas thethird party ad server (where the Advertiser or Agency will bemonitoring the deliveryindependently).

Thecertificationprocessisactuallyverystraightforwardandwithresponsiveparties,theprocess need not takemore than a few days. Having said this, in practice, up to threeweeksisrecommendedbetweentheannouncementofanewvideoPublisheronthemediaplanandthecompletionofthelongtermcertificationofthatPublisher.

• Thirdparty ad servingprovidershaveaPublisher certification team inplacewho,oncealertedtothepresenceofanewsiteonaplan,makecontactwiththe Publisher verbally and send over test tags in an email to the Publishercontact.

• ThePublishertakesthesetest tagsandloadsthemupontothePublisher-sideadserver,displayingtheresultingadsinatestenvironment.

• The location of the test environment is supplied to the third party ad servercertificationsteamwhorunsometeststocheckthatthemetricsarecollectingcorrectlyandthattheadservercallsarealltakingplaceattherighttime.

• WhenbothPublisherandadserversatisfiedwith themetrics theyareseeingthePublisher is labelledascertifiedand theAdvertisercangoahead torunactivityonthisPublisherinthischannel.

• Certificationusuallymeans that anyAdvertiser from thatpoint forwardscanskipthisprocesstorunvideoonthePublishersiteinthetestedplayer.

Figure3.5.8CertificationforInstreamremainsarequirementindigitaladvertising.Thecertification teamat the thirdpartyadservingprovidersends test tags to thePublisherdirectorthenetworkforthesiteoperationsteamtoimplementthetesttagsinthecorrectplayerenvironment.Thepreviewof this implementation is sentback to theCertificationteamwhoassessthepreview,awardingcertificationifthepreviewpassesvarioustests.

Certifications can be lengthy processes, especially when the certification revealsdiscrepancies and errors. For this reason it is recommended for the media planner tosupply the listofsitesandpublishersfor thevideochannel to the thirdpartyadservingproviderlongbeforethecampaignlivedate.Owingtothelastminutebuildingofmediaplans,MediaAgenciescansometimesfindthemselvesintroubletryingtopushvideotouncertified sites. It is therefore very important to be aware of the full scope of thecertificationprocessforInstreambeforeeverbuyingthevideoadspace.

Publishercertificationasaprocess isnothingnew.RichMediawasalsooncean infantchannelandusedtofollowthesameprocess.Eventuallyalltheestablishedpublishersandplayerssupportingvideocertifiedinfullandalthoughre-certificationsneedtooccurwhentherearebigsitechanges, thecertificationsallgoonquietly in thebackgroundwithoutdisruptingmediaplantimelinestoomuch.RichMediaasachannelhasnowmaturedandInstreamwill follow suit and standardisation improves; ridding this extended campaignprocessinthecomingyears.

TraffickingInstreamAds

Asmentionedearlierinthissection;InstreamvideoadstendtobesimplertotrafficfromthethirdpartyadservinginterfacethanDisplayorRichmediaadswhichwillcomeasarelief to all traffickers. Linear VAST ads mean the simple vetting of and placementassignmentofcoreflashfiles(*.mov,*.flv,*.aviareexamplesandallofwhichareverycommonvideofiletypes).WithVPAIDtheprocessismorecomplexbutasanimmaturechannelVPAIDinInstreamisstillrarelymadeupofcomplexinteractiveexecutions.

Traffickers are encouraged to obtain the highest video quality available from the videoproducerorCreativeAgency.Withahigherdefinitionvideoasset,theadservinginterfacecan transcode thevideo into lowerquality videos for sites andPublishers that prefer toworkwithalowervideobit-rate.Adsarecreated,assignedtoplacementsandplacementstositesinthesamewayaswithDisplaytraffickingbeforethetagsaregeneratedandsentto the Publishers. With VPAID, the process of testing interactions in a previewenvironmentisalsoadopted.

With some ad servers the trafficked ad tags look a little different thanDisplay ad tagswhenthecodepackagesareopenedonthePublisheradopsside.Theadtagscontainlinks

tovariousqualitiesofthevideotoallowthePublisherteamtoselectwhichversiontheywould prefer to serve to their users. More advanced Publisher ad servers assess thesuitabilitytoserveacertainbitrate(usuallybasedonthespeedoftheinternetconnection)andservethevideowiththebitratedeemedmoresuitablebythePublisheradoperationsteam.

WithInstreamadstrafficked,thetrackingfeedsdownintothethirdpartyadserverwhentheadisservedandthemetricsarerecorded(seeChapter5-ChannelSpecificReportingfordetails regarding thesemetrics).Aswithallotherchannels, theunified trackingandreportingopensupthepossibilitytounderstandhowInstreamadsaddvaluetocampaignsrunning in other channels.Hopefully such insights justify the higher cost to serve.Theconsensusremainsthatthischannelprovidesamoreengagingimpactupontheuserwhichwillencouragebetterperformanceofthecampaignasawhole.

Channelswhichcandelivervisualcreativearehotlysoughtbybigbrandmarketerswhousetheimpactassociatedwithimagerytodrawuserattention.Withsuchcreativitythereislimitedcall-to-actionwithinthecreativeexceptforactivitiesthatmightencouragebettereducation or awareness about the product or service. Rich Media and Video digitaladvertising are very well suited to this need and on the periphery sit Display and itsmaturingcousin;Mobile.ThenextsectiondivulgesthecampaignmanagementprocessforMobile,disclosingthekeyconceptsbehinditsgradualrefiningandwhyasachannelitissointangible.UnlikeInsteamwhichmightbearelativelysimplechoreforthetrafficker,mobilecontinuestoproduceheadachesforAgencyoperationsteamstheworldoverinthecontextofthirdpartyadserving.

Chapter3-Section6Mobile• TheConnectivityConsideration

• InternetEnabledDeviceDivergence

• PublisherCertificationforMobile

• Figure3.6.1

• Challengeswithinthebrowser

• ResponsiveDesign

• WhateverhappenedtoFlash?

• Thefall-backMechanism

• TraffickingadsforMobileBrowsers

• QAformobilebrowser-basedads

• Figure3.6.2

• Challengeswithintheapp

• Figure3.6.3

• TraffickingadsforIn-appadcampaigns

• WheredoMobileAdNetworksaddvalue?

The Mobile channel presents the greatest upcoming opportunity for Advertisers. Thegrowingvolumeofinternettrafficcomingfrommobiledevicesisstaggeringprovidingachannel of enhanced reach. Since the ratio of devices to humans is reaching a levelplayingfield,deviceshavebecomeuniquetoeachhumanwhichmeanstheeradicationofthe assumption that a whole household is accessing the internet from the same place.Furtherinthissectionwewillseehowthisresolvesindatacollection.Beyondimprovedreach,thefeaturesofwhatwereoriginallycalled“smartphones”arestandardisedenoughthatcreativedeveloperscanassumethatadscanbedevelopedtomakethemostofthesefeatures.Examplesinclude:

• Gestures

• Geo-location

• Accelerometer

Thiswould suggest that themobile channel should be an exciting area for growth andcreativitybutitisfraughtwithchallengesthatmanytechnologyproviders,AgenciesandPublishers have been accused of not implementing consistent standards as have beencreatedintheotherdigitalchannels.

Thissectionbeginsbylookingcarefullyatwhatitmeanstoadserveto“mobile”devices,thechallengeswithservingtobothbrowserandin-appenvironmentsandtheprocessoftrafficking tocertifiedmobileproperties.Finallywesettleon thepoliticsofmobileasa

narrowlycontrolledmediachannelandtheimportanceofsupportingindependenttrackingmythologiestodisclosereliablereportingmetrics.

TheConnectivityConsideration

EvenwiththeintroductionofSmartPhones,internetconnectivityandbandwidthremainan issue for delivering high quality creative to mobile devices. For instance, mostAdvertisers andMediaplannerswill rule out thepossibilityof running richmediawithvideoformatsonmobilePublishersbecauseitputsaheavydrainontheuser:theneedtodownload a lot of information to display the ad. To avoid the user suffering a badexperience on a Publisher site, high file size ads have been discouraged, which limitscreativity.Where technology is still catchingup is findingaway to identify thoseusersusingamobiledeviceonthemoveversusthosethatarestationaryandaredownloadingdata through aWi-Fi connection. If connected theWi-Fi, suddenly bandwidth and datadownloading does not need to be restricted but ad servers are not currently providingdevice level targeting that distinguishes between Wi-Fi and connection via a mobileprovider. This feature may never be needed in the ad server itself owing to theimprovements inwireless data transmission. 4Gpromises internet speeds thatwill rivalWi-Fiandifdatacostscomedownoverthenextfewyearstherewillnotneedtobeanadserverfeaturetoseparateouttheusersbytheirbandwidthspeeds.ForthetimebeingitisadvisedforcreativedeveloperstobuildthefrugalPublisherspecificationsavailable,sincethismakesfullconsiderationforuserswhomightnotbeabletodownloadlargeads.

InternetEnabledDeviceDivergence

Since there is awhole plethora of internet-enabled devices and the distinction betweenwhatisamobiledeviceandwhatisnot, isaconundrum.Inordertoseparateoutmediainventory,Publishersbeganbycreatingdifferentpropertiesinthehopethatthebrowsinguserwouldheadtothepropertymostsuitedtotheirdevice.Withthismethodologyauserarrivesatthestandardsitefromtheir“desktopdevice”,choosestodownloadtheappfromtheirmobiledeviceorusestheirin-builtmobilewebbrowsertoheadtothepurposebuiltmobilewebsite.LargerPublisherssuchasnewspaperpublicationshavetakenthisastepfurtheranddevelopedapplicationsdesignedforuseontabletstogiveabrighter,largerandmore engaging experience. Here the distinction is clear, ads must be created for thebrowserenvironmentandforthe“in-app”environment:

• Desktopbrowser

• MobileBrowser

• In-appforMobile

• In-appforTablet

Making the distinction between the user’s arrival is an easy task for the Publisher, thebrowser type can be detected by the JavaScript design of the Publisher site such thatmobile users are re-directed to a mobile site while others land on the usual desktopbrowser page. For apps, the distinction is made when the user first downloads thePublisher’sapp;choosingtheappmeantfortabletsversustheonemeantforothermobiledevices.

PublisherCertificationforMobile

As with Instream Video and originally rich media, both apps and browser Publisherpropertiesneedtobefullytestedandcertifiedbeforeanadcampaigncangolive.Withoutcertification, theAdvertiser runs the risk that the data collection and actual ad deliverymight fail.As a best practice,when amobile campaign is being booked and inventorypurchased,thethirdpartyadservershouldbealertedsothatifrequired,thecertificationprocesscanbegin.Justaswiththeotherchannels,thecertificationsteamatthethirdpartyadservingprovidercontactthePublisherdirectlyandsendovertestversionsofmobileadtags.ThePublisherimplementstheseonthecorrectpropertiesandthecertificationsteamchecktheadsdeliverandfunctionandthatdataiscollectedinreporting.

Just as with Display inventory, Mobile inventory is partly held by large PublisherNetworksthattendtobemobile-specificandcertifyingthroughtheNetworksmeansthatmuchofthemobilewebisPublisher-certifiedtorunthirdpartytags.AswithDisplayitisrecommendedthatshouldanynewPublishersitesorNetworkscropuponamediaplan,that these details be supplied to the certifications team around three weeks before acampaignstartdate.RichMediaandDCOadformatsareenabledwithinmostthirdpartyadservinginterfacesandopenuptheoptionsforwhatkindsofadscanbedeliveredandtrackedatscalebuttheuseofthesewillextendtheperiodofcertification.ThankfullytheIAB has started to standardise some core Rich Media formats specifically for mobilewhich,aswith theRichMediachannel fornon-mobiledeviceshasbeenahugehelp ineradicatingtheneedforcertificationalmostaltogether.

Figure 3.6.1 Examples of Rich Media units for Mobile campaigns which includeexpandableads (simplyadsmadeupof twocorepanels), Interstitialads (which fill thescreentocoverthecontentmuchlikeafloatingunitfromRichMediaforDisplay)andtheClick-to-callfunctionality(uniquetothemobilespacedrivingtheuserthroughtoaphonenumberwhichconnectsfromthedeviceinstantly).

Beyond certification the mobile landscape is diverse and technically speaking,cumbersome and complex. The distinctions and challenges faced in delivering ads tomobile browsers and in-app environments is great enough to have slowed the spend inmobilechannelsformediaspace.

Challengeswithinthebrowser

Withinthesameadcampaignthecreativedevelopedfortheseenvironmentswillbeconsistentindesignsoastokeepthecostsofcreativedevelopmentdown.Iftherewereonlyfourpossibleenvironmentsthingswouldberelativelysimplebutmobiledevices differ in screen size significantly. This wasn’t always the case, sodeveloping a standard 300x250meant the same ad could be used on the desktop

PublishersiteandinthePublishermobilesite.Thatisuntiltheauserarrivestothemobile site with a small screen and sees just half a 300x250; not an effectiveexecutionandessentiallyawasteddelivery.WhentheuserarrivesatthePublishersite theJavaScriptcanbeuseddetect thescreen-size to requestanadof themostsuitabledimensionsfromthethirdpartyadserverbyfiringoffthecorrespondingadtag.ThisstillmeansthattheCreativeAgencyneedstodevelopmanyadsizesandwill use the Publisher specifications to determine which ones to build out. ThePublisheradoperationsteamandthetraffickerwouldhaveabigtaskontheirhandsif they had to create hundreds of tags to do this job. Thankfully technology hascometotheaid.

ResponsiveDesign

Responsive Design is a new concept but essentially means that with asingleadtagimplementedonaPublisherpagethatthescreensizecanbeadjusted and as the screen enlarges or shrinks, the delivered ad unitchanges.Theeffectisthatthesameplacementappearstodeliveraleader-boardonatabletbutonamobiledevicethatspotshowsabannerformat.The Publisher ad operations team can celebrate, receiving just one tagfromtheAgencytraffickerandyetallusersseeingonlytheaddimensionsmost suited to their device. Discouragingly the Creative Agency stillneedstodesignadstoeachofthosedimensionsandalthoughthirdpartyad servers are now offering creative templates in Flash to assist withbuilding responsive ads, there is still likely to be a higher cost to theAdvertiser than there would be for developing a single format for adesktop-onlycampaign.

WhateverhappenedtoFlash?

Flash, the cornerstone of the creative development process is being phased outacrossmobiledevices.Butitisnotthebrowserswhicharenotsupportingthepluginrequired to play theFlash animations, it isAdobe themselveswho have opted tostopofferingthepluginformobilebrowsers.Thechallengetoadapttheplugintoanever divergent technology landscape without profit put the nail in the coffin forFlash. InsteadHTML5 presents an opportunity to have the same (if not a better)visualexperiencethanFlashonamobiledevice.

Thefall-backMechanism

Within standard display a fall-back mechanism has long been thestandard: serve the flash animation, and in instanceswhereFlash is notsupported, show a backup image. This method does not serve mobilecampaignsverywellatall.Sincethevastmajorityofmobiledevicesdonot support flash, users are seeing static images as ads instead of thecostlyanimatedcreativethatwasdevelopedforabetteruser\experience.Thedemandtodaylookslikethis:

• Iftheuser’sbrowsercansupportFlash,showtheFlashad

• Iftheuser’sbrowsercannotsupportFlash,showtheHTML5ad

• If the user’s browser cannot support HTML5, show the backupimage

AsmobileinternettrafficgrowsfurtheritislikelythatthedemandfromAdvertiserswillbe that theHTML5version is theoneat the topof thefoodchain.Attimeofpublishing,HTML5isbeingtreatedasaninferiorcodingenvironment(whereinfactitscapabilitiesarebettersupportedandsuperiortothoseofFlash).Althoughfall-backmechanismshavebecomeintegrated into the delivery of ads by third party ad servers, formobilecampaigns,itisstillabestpracticetolookatdevicedeliveryreportsandtoseehowmanybackupimagesarebeingdeliveredbyacampaign.

TraffickingadsforMobileBrowsers

TheAgencybuyerormarketerprocuresthemobileinventoryfromtheNetworksorPublishersandsuppliestheMediaplanto:

a. The third party ad serving provider to check the sites for the purposes ofcertification

b. Thecreativedeveloperwhobuildsouttherequiredcreativebasedonthemediaplan,takingintoaccountthePublisherspecifications.

c. The trafficker, who sets up the campaign in the same way as a displaycampaign

Oncethetraffickeridentifiesthatthesitesbeingservedtoaremobile,theadservergrantstheabilitytoattachcreativetotheadstosupportbothresponsivedesignanda fall-backmechanism.The trafficker then traffics thead tags to thePublisheradoperationsteamwhoimplementthemonthemobilesite.Thetotalnumberoftagsbeingkepttoaminimumthankstoresponsivedesign.

QAformobilebrowser-basedads

ToQAtheads the traffickerneeds touseadifferentmethodthan isused inotherDisplaychannelssincetheactualadneeds tobe testedonasmanymobiledevicetypesaspossible.Inordertochecktheadthoroughlyandmonitortheadcalls,thetraffickercanuseanHTTPtracerlikeFiddlerandadjusttheproxysettingsonthetest mobile device so that all the traffic to and from themobile device runs viaFiddlerrunningonalocaldesktopbasedcomputer.ThetraffickercanthencreateapreviewURL of the ad inside the third party ad server interface and send to themobiledevicefromwheretheURLshouldbeaccessedviathemobilebrowser.

Figure 3.6.2Owing to device divergence, a thorough QA process would meantestingapreviewlinkgeneratedbythetraffickeronallavailablemobiledevicestocheck that the ad loads and is at least clickable before the mobile campaignlaunches.Thisgoesformobilephonesandtabletdevices.

There are some web-based mobile browser emulators available to check ad tagsacross a range of devices without performing a QA across every device typeavailablebutwherepossible it is recommendedfor the trafficker toQAtheadoneachphysicaldevicetype.

Challengeswithintheapp

The app space has a growing number of ad formats cropping up and these arepredominantlyownedbythedominatingMobileAdNetworks.Theseadnetworkscontrolthedeliveryofadstotheappsmakingiteasyforappdeveloperstomonetisetheirapplicationsbysigningup to thenetworks,deployingapieceofcode in theapp andhaving the network serve up adswhenever the apppages load.This hasassisted in supporting wholly free apps, providing developers with a revenuestream.

Tothirdpartyadserveanadintoanappmeanschannellingthroughthenetworkbutalsodeliveringtheunitandgatheringthedataintoanon-browserenvironment,sohereiswherethingsgetalittletricky.Appsarerarelydesignedtoallowformultiplecallsintoandoutofauniquelydevelopedcodestructure,soappsmuchoriginallybedevelopedusingatemplatethatproducesastandardisedcommunicationchannelin and out of the app environment. This template is referred to as an “SDK”(Software Development Kit) which, when supplied by the ad server ensures thebasic ad callswill occur sufficiently for accurate counting in ad server reporting.EncouragingPublisherstobuildapplicationswithsuchanSDKforthepurposesofthird party ad serving (and thus greater and fairer monetisation) means workingwith Publisher-side development teams to have them understand the benefits ofputtingmoreworkinthedevelopmentoftheapplication.

Figure3.6.3TheSDKcircumventstheproblemoftheblack-boxenvironmentoftheApp, facilitating calls into and out of the ad to the ad server or measurementtechnology.

TraffickingadsforIn-appadcampaigns

WhetheraPublisherchoosestoworkwithanSDKornotwillstillrequirethethirdparty ad servingprovider’s certification team to test ads in the applicationbeforecertifying theenvironmentandapproving theapp for thirdpartyaddeliveryonamediaplan.EvenwithouttheSDKinplace,othertechnologieshaveemergedonthePublisher network side to facilitate ad tracking called “Ad Enablers”. Thetraffickingprocess for in-appadcampaigns is the sameas in-browser formobile,where either the direct app or the mobile network are listed as the site in thecampaignmanagementworkflowinthethirdpartyadservinginterface.

ToQAhowanadwillbehaveinanappenvironmentismorecomplexthatQAformobileadsinthebrowser.ThetraffickermustworkwiththePublishertoprovideatest page when the actual ad tag is implemented. Very few mobile apps offeremulatorswhichcanbeusedbythetraffickeronamobiledevicetoemulatethein-appenvironmentforthepurposesoftesting.Withoutthetesttheriskistheadwillnotfunctionasexpected.

WheredoMobileAdNetworksaddvalue?

AsmentionedintheInstreamVideosection,ifanAdvertiserchoosestorunacampaignacrossasinglenetworkthenthereisnoqueryofde-duplicationbecauseusersareconfinedto seeing ads served from one place and tracked by a single system. For this reasonNetworkstrytohaveasmuchreachaspossibleandasinthevideoworld,inthemobileworld this is entirely possible. InMobi is an example of a high reach mobile network.Thirdparty ad-serving continues tobeuseful on just a singlenetwork to independentlyverifythedeliveryandtrackingoftheseads.

It is clear that thedeliveryof thirdparty servedads toboth thebrowserand the in-appenvironments,offertheirfairshareofchallenges.Onlycertificationat itsmostthoroughcancomeclose to theguarantee that theadwilldeliverwith the supportofa fall-back.Data collection for these ads is not quite as standardised and straightforward as for theotherdisplaychannelsandthiscomesbackthedeviceitself(moredetailonthisinChapter5-ChannelSpecificReporting).

AsMobilematuresitwilltakeprideofplaceamongchannelsthataremorepurposebuiltto drive specific actions like sales.Channelswhich drive specific actions like sales arereferredtoasDirectResponsechannels.Themostsignificantamongtheseis theSearchchannel.ThenextsectioninthischapterunwrapsPaidSearch(alsoknownasPPC)which

isamuchmorematuredigitalchannelthatisfullythirdpartyadserversupportedandisseenasthegatewaytoperformancemarketing.

Chapter3-Section7PaidSearchandNaturalSearch• SEMandPPC

• CreatingCampaignsandKeywordsintheEngine

• Figure3.7

• Figure3.7.1

• ConnectingaBidManagementSystemtotheEngine

• Figure3.7.2

• Whytrackpaidsearchthroughathirdpartyadserver?

• HowtotrackPaidSearchthroughathirdpartyadserver?

• ManualSearchTracking

• Figure3.7.3

• AutomatedSearchTracking

• Figure3.7.4

• AdServertrackingintegratedwithBidManagement

• Figure3.7.5

• ExtendingtrackedPaidSearchtothelongtail

• Figure3.7.6

• ExtendingtrackedPaidSearchtoMobile

• SearchandotherChannels

• WhatisSEO?

• Figure3.7.7

• ReferralURLtotriggeraclick

• Figure3.7.8

• AdServingcostsforSearch

The next areas of this chapter provide reference information for what are called the“PerformanceChannels”.PerformanceChannelsarehighlymeasurablechannelsmarryingreachwitheffectiveness;specificallydesignedtodriveaudiencesthroughtoconversions(seeChapter4).PerformanceChannelsarelocatedlowerdowninthestrategicmarketingfunnelbecause theyareexploredandusedbyconsumers thathavepredominantlycomecloser to finalising thepurchasedecision.Thiscanbecompared to thedisplaychannelswhichoftenbetterservetheneedsoftheAdvertiserfurtherupthefunneltocreatebrandandproductawareness.

Searchiseffectivewhenusershavealreadybuiltupanawarenessofaproductorbrand

throughmorevisualchannelssuchasdisplay,astheyusethesearchenginetoresearchtheinternetformoreinformation.InthissenseSearchenginesarethenavigationhubofthewebexperienceandassucharevaluableplacesfortheAdvertisertoattractattention.

Search can be divided into two channels Paid Search and Natural Search. The twochannelsarealmostnevermanagedbythesameteambutbothchannelscantechnicallybetrackedbyathirdpartyadserver.It isworthnotingthatfor theSearchchannels, theadserverisnotconsideredto“serve”anythingsincethereisnoimagedisplayedbutrathercanattendtoaportionoftheuserclickpathfortracking.Theadserver’spositionisstillwhollyrelevantbecauseinternetusersthatmighttakeactionswithanyotherchannel,mayalso perform an action in a Search channel and capturing this pathway is essential forconversionandattributionreporting(seeChapter4).

ThissectionbeginsbylookingathowpaidsearchisusedasachannelbytheAdvertiserandhowthethirdpartyadserverconnectstothesearchenginestoseamlesslyintegratefortrackingpurposes.Thisprocesshasevolvedovertheyearsandhastakenabigworkloadoffofthetrafficker’sshoulders.Wewilltouchonhowtheshadypastshapestheprocesstodayandwhythefutureofcampaignmanagementforpaidsearchlookssostrong.Thissection will also touch on bid management solutions; which assist the core ad searchprogramstoscaletheimplementationofkeywords.Morecontroversiallythissectionwillendbyexplaininghownaturalsearchcanbetrackedasachannelinitsownrightandhowthesettingupofthischannelfromatrackingperspectivemustbecarefullyconsideredsothatmoredirectlypaid-forchannelsdonotappeartolosetheireffectiveness.

SEMandPPC

Paid Search is often given the acronym “SEM” which stands for Search EngineMarketing,however,Paidsearchchannel isalsocommonly referred to industry-wideasPPC, which has the potential to confuse those new to the world of search since PPCactually means Pay Per Click. It is true to say that the paid search channel ispredominantly funded throughaPPCorclick-basedpaymodel,asopposed to theCPMdominated model for the Display channels, but it might not be the best name for thischannelasawhole.

DefiningPaidSearchisfairlystraightforwardbutthissectionwillnotdwelltoomuchonitsintricaciesbeyondtheintermediateeducationalrequirementsoftheAdvertisertothirdpartytrackitsactivity,sincethereareagreathordeofhow-totextbooksalreadywrittenonthetopic.

Paid search is essentially the sponsored search listings that appear on the search resultspageinthesearchengine.Thesetraditionallyappearalongthetop(justunderthesearchbar) and down the right side of the page. They are usually presented on an off-colourbackgroundandtheadspresentedusuallymatchthekeywordtypedin,insomeway.ThesearchenginealgorithmforthepaidsearchlistingsismeanttoensurethatAdvertisersthathavepaidmorefortheirkeywordstodisplaywillappearhigheruptheresultslistormorefrequentlyacrossmoreuserssearchresultspages.

WhenreferringtothepayingofkeywordsitisusefulforustoloopbacktothetermPPC.Thesearchkeywordsthemselvescost theAdvertisermoneyonlywhenthereareclickedon(thedeliveryofthesponsoredlistingitself isessentiallyfree).Advertisersarriveinto

anauctiontoofferupthemaximumpricetheywouldpayshouldauserclickontheirad.WinningtheauctionwilltriggerachargeintheAdvertiser’saccount.

Initsmostbasicform,theAdvertiserispermittedtoenterintothisauctionbycreatinganaccount for buying keywords with the Search Engine. For instance Google offers aprogramcalledAdwords,Microsoft’sBingoffersAdcenter.

CreatingCampaignsandKeywordsintheEngine

PaidSearchactivityisusuallymanagedbyaspecialistsearchteamwhichbelongstothemarketing team on theAdvertiser side directly or sits in theMediaAgency (separatedfrom the Display operations team). Some Media Agencies have carved off a decentportion of their staff to create new search specialist Agencies (a good example isPerformics;anSEMAgencyunderthePublicisGroupumbrella).TheseteamsaccesstheSEMinterfacebelongingtotheenginebynavigatingtotheirsiteandloggingin.

Thefirststeptosettingupasearchcampaignforthefirsttimeisaddingbankdetailsandfunding the search account. Once funds have been deposited, the team can begin bycreatingoruploadingkeywords.Keywordsarebasicallyanythingthatausermight typeinto the search engine to find something. Search engine marketers are encouraged tocreateasmanykeywordsastheycanthinkofandmanyuserspecialisttoolsandprogramstogeneratekeywordsintheirthousandsandmillions.

OncethekeywordsaredeterminedtheyneedtobecoupledwithashortdescriptionandtheURLwhere theusershouldbedirected toonce theyclickon thead,essentially likewithad serving;aclick-throughURL.Keywordscoupledwith thedescription, title andclickthroughURLmakeupthesearchadsthemselves.TohelpmanagetheseadstheyaresortedintobucketsintheSEMinterface:called“AdGroups”whichinturnareclumpedtogether toformcampaigns.Themostcommonwaytomassproducemillionsofsearchads involves using spread-sheets on a localmachine to couple the generated keywordswiththeircomponentpartsandthenameoftheadgrouptowhichtheybelong.Thesheetisthenloadedupintotheinterfaceunderneaththecorrectcampaign.

Figure3.7Componentsof thenowstandardisedSearchAd include theKeywordas theheader underlined in bold followed by aURL (which does not have to reflect the truelandingpageortheclickthroughURL).Bothareaccompaniedbyashortdescription.

Campaigns in the SEM interface can be defined at the marketer’s discretion, manymarketersarefollowingaconventionofhavinganewcampaignforeachbrand,productandmarketing campaign. The campaigns listed in the interfacemay not need to all beactive at the same time, so it is simple to activate and deactivate the activity. Also

alongside eachkeyword themarketer needs todeterminehowmuch they arewilling topayforthekeywordssothemaximumbidpriceisentered.Thespread-sheetasawholeisthenuploadedtotheinterfaceandthecampaigncanbeputlive.

Inside thesameSEMinterface in theengine, it ispossible toget reports toshowwhichkeywordswereclickedonandhowmuchwaseventuallypaid.Despiteitssimplicitythereisawholerangeofoptionsavailabletoenhancetheperformanceofthesearchcampaigns,fromadjustedmessaging at different timesof day to utilising retargeting strategies (seeChapter7).Whatisactuallyhappeningtotheuserwhentheyclickonasponsoredsearchresultinthesearchengine,isthattheirclickactivitysendsacounttotheSEMreportingserver so that the information about the click such as the keyword clicked on and thetimestampoftheactivityisstoredjustastheuserarrivesatthelandingpagedestination.

Figure3.7.1Theclick-paththattheuseristakenonasaresultoftheactionofsearching.Inthisexampletheusernavigatesintheirbrowsertothesearchenginehomepage(inthisinstanceGoogle)wheretheirsearchtermiscarriedtothesearchresultspage.HeretheSEM Search Ad is displayed. The user clicks the ad (which is labelled as a sponsoredlisting) and appears to land directly on the homepage. The user is in fact redirectedthrough the Google Adwords reporting server so that the click is registered and theAdvertiserischargedfortheactivity.

ConnectingaBidManagementSystemtotheEngine

ThesearchengineshavedevelopedtheirownAPIstoallowothersystemstoconnectandcontrol thiskeywordandcampaignmanagementenvironment,atadistance.What thesesystems have that the search engine SEM interfaces do not, is more complex ways ofdecidingwhatkeywordstopaymoreforoverothers.InthecontextoftheauctionmodelinwhichPPCisdeterminedrequiresthatatscale,pricesbesetaccordingtothevaluethateach keyword is bringing. Since a multitude of factors determine this, these external

systemsarecommonlyreferredtoas“BidManagementSystems”andpullonvariousdatasourcestomakedecisions.

Figure3.7.2TheuserwouldtypicallyinputasearchtermintothesearchenginesuchasGoogle and the ad clicked on (represented here by a keyword)would take them to thelanding page (redirected through the Adwords reporting server). With a bid managerinvolvedtheuser’sclick-pathwillstillrouteviatheAdwordsreportingserverbutwillalsorouteviathereportingserverfortheBidManagementSystem.Thisisrepresentedby(1)and (2). Information about the user’s journey is also collected by the Bid Manager’stechnology (alsoembeddedon theAdvertiser’s landingpage)and this isalso fed in theBidManagementSystem’sreportingserver.Withinformationaboutclicksandconversions(seeChapter4)theBidManagercanupdatethebidpriceonthekeywordviatheSearchEngine’sSEMAPI.

BidManagerstracktheuserastheyclickonthekeywordtodeterminehowfarthatusergoesdownthepurchasefunnelandadjustedbidsforkeywordsbasedonthisasastartingpoint.Tothisextent,whenthebidmanagerissetupontheAdvertiser’sSEMaccount,thebidmanagerconnectstothesearchengineviatheAPIandaddsa“codewrapper”aroundtheclick-throughURLofthekeywordslistedthere.Thewrapperisportionofcodethatsits either side of the click-through URL and when the user clicks the link, the bidmanagerisnowtrackingtheuserindependentofthesearchengine.

DespitehavingtheirownSEMinterface;Adwords,GooglehasitsownBidManagementtoolwhichmakesuppartoftheDoubleClicksuite;thelatestiterationattimeofwritingiscalledDS3whichevolvedoutofthelegacytoolDARTSearchwhichwasoneofthefirstad-server integrated bid management tools. Other popular and powerful players in thespaceareMarin,KenshooandIgnitionOne.

BidmanagementtoolshavebegunevolvinginrecentyearssothattheytakeonnewdatafromtheAdvertisersideandbeginactingalittlemorelikeaDSPoraDMPbutleadingwith bidding on the search channel (seeChapter 8 - Programmatic). Thedifferencehere is thatBidManagersarebuilt forenormousscalefor thehandlingand processing of millions of keywords where bidders in the Display space areworkingwithfarfewerads.

Having looked intohowcampaignmanagementworks for searchmarketingcampaigns;bothsmallerscaleinsidetheSE(SearchEngine)interfacesandatalargerscalewithbidmanagementsystems there remainsan importantspacefor the independent trackingandcountingofthedatabythethirdpartyadserver.

Why track paid search through a third party ad server? Owing to the scale anddynamicsofPaidSearch, it isconsidered tobeawhollyseparatechannel fromDisplayadvertising.Thismeansthatinordertotrackusersthat:

1)clickedonpaidsearch

2)wereinfluencedtosearchbydisplayads

..theremustbeacommontechnologytrackingbothchannelsatasinglepoint.Theregularuse cases for using a third party ad server still stand; such as the case to have anindependent tracking technology, kept separate from the Publisher/Search Engineprovider.Italsobecomespossibletocompareandcontrastthecostsofpaidsearchacrossmultipleproducts,brandsandcountriestootherchannelsinasinglesystemofrecordwiththesamecountingmethodology.

HowtotrackPaidSearchthroughathirdpartyadserver?

Tothetraffickerworkinginsidethethirdpartyadserverinterface,thesearchenginescanbe classed as Publishers and the keywords as the ads or placements. The traffickernavigatestotheAdvertiseraccountinquestionintheadserverhierarchyandtherearetwomethodstothenconnectthethirdpartyadserverfromthere.

ManualSearchTracking

ManualSearchtrackingishowthingsusedtobedoneintheworldofadservingforpaid search. This method is now reserved for very rare instances in which theSearchEngineAPIfailsorforconnectingtosmallermoreobscuresearchengines.Thetraffickerundertakesthefollowingsteps:

1. Setsupanewcampaigninthethirdpartyadserverinterfacewiththecorrectnaming convention (that might mean the advertiser is naming differentcampaignsaccordingtowhichchanneltheybelongto)oraccessesanexistingcampaignthatcontainsactivityforallchannels

2. Generates a tracking pixelwith single click tracker for each keyword in theengineinsidethecampaign

3. Inserts the same landing page URL as can be found alongside the existingkeywordsintheengineintotheclicktrackers.

4. Trafficsalloftheclicktrackers,sothattheadtagcodefortheclicksbecomesavailable

5. Downloads the keyword list from the Search Engine SEM interface for thecampaigninquestion

6. Replacestheclick-throughURLcolumninthekeywordlistspreadsheetwiththematchingcodefortheclickfromtheadserver

7. Uploadsthecompletedspread-sheetbacktotheSearchEngineSEMinterfaceandputsthecampaignlivefromthere

Figure3.7.3ManualSearchTrackinginvolvesaheavy]operationalburdenonthetrafficker.Atrackingplacementiscreatedforeachkeywordtomatchthekeywordsin the engine. The click redirects are then trafficked to the Media Agency whoreplacetheclick-throughsoneachofthekeywordswiththosegeneratedbytheadserver.Whentheuserclickstheadforthekeywordinquestion,theuserclickpathisroutedthroughtheSearchEngineSEMreporting(suchasAdwords)andthethirdpartyadserverreportingbeforelandingtheuseronthelandingpage.

This manual process is enormously time consuming and problems occur whentryingtodifferentiatetheclicktrackersinadserverreportingfromotherkindsofadservers if a naming convention has not been set up to divulge the channel type.ThankfullytheSearchEnginesthemselvescametotheaidofmarketerswhosetotalnumberofkeywordswasbecomingunmanageablewiththissystem.

AutomatedSearchTracking

The search engineAPI allows the third party ad server to connect directly to theengineandwritetheclick-throughURLstothekeywordsthemselvesdirectly.Thetraffickerthen:

1. Logsintothesearchenginefromaslightlydifferentscreeninsidethethirdpartyadserver but still sittingwithin the sameAdvertiser account. This loginwill be thesameas theoneusedby the searchmarketers to access theSEM interface in thesearchenginedirectly.

2. Sees thecampaignsas theywerecreated in the searchengineand selects the rightone to track all keywords in that campaign so that the tracking is fed into the adserverunderthisaccount.

Thethirdpartyadserverwillthenautomatically‘wrap’thelandingpageURLsintheenginewithadserverclicktrackingwithnoadditionaleffortrequiredfromthetrafficker.

Figure3.7.4Theautomaticsearchtrackingoptionpresentedbysomethirdpartyadservers ismuchmore favourable operationally speaking.Here the trafficker onlyhasclick“track”byselectingtherightcampaignnameinthethirdpartyadserverand the corresponding name in the Search Engine SEM interface has all itskeywordswrapped.

Once thekeywordsarebeing trackedtheadserverconnects to theengineat leastonceperdayviatheAPItorequestreportinginformationwhichispulleddownintoad server analytics so that the trafficker can pull reports about the cost of thekeywords and their performance. Thismeans the Advertiser or Agency does notneed to log into the various engines to extract the results but can get them allcombinedinadserveranalytics.MoreinformationonthiscanbefoundinChapter5-ChannelSpecificReporting.

AdServertrackingintegratedwithBidManagementBidmanagementsystemscanbeused in conjunction with third party ad server reporting and at a deeper level the twotechnologies can share data independent of the search engine itself. Operating bidmanagementsystemsandad-servingiseasyforthoseadserversthatclaimtohaveabasiclevel of integration. Essentially both bidder and ad server work separately butharmoniously in that their codewrappers bothwrap the keyword in the engine via theSearchengineAPIandthe twocodewrappersdonotdisrupt thedifferentsystemsfromtrackingandreportingseparately.Toachievethisstatetheadserverneedstohaveworkedwith the bid management provider in a process of certification. In some instances thewrappersdisruptoneanotherunlessonewrapperisaddedaftertheother,thistendstobethe case with lesser known bid management providers and less user-utilised searchengines.

Sharingdatabetweenadserversandbidmanagersisarelativelynewideacomparedtothelonger history of search marketing. In this instance the ad server is collecting a vastamountofdataabouttheAdvertiser’sadcampaignsoutsideofthesearchkeywordbutcantieall this informationbackto it.With theAdvertiser’sapproval, theadservercanpassdatathatmightshowifsomekeywordsaremoreorlesseffectivedependingwhentheusersaworacteduponthem.Inturnthebiddercanimproveitsbiddingalgorithmsthroughasystemoflearning.

An advanced example of such an integration would include the passing of E2C(exposuretoconversion)orP2C(pathwaytoconversion)datafromtheadservertothe bidder using a daily feed (See Chapter 5 - Cookie Level data). Using thismethodallowsthebidmanagertoapplyan“AttributionModel”tousersfeaturedin

thedatatoadjustthebidpriceforcertainkeywords(moreonAttributionModellinginChapter4).

Figure3.7.5Introducingintegratedbidmanagementintotheequationallowsthebidsonthekeywordstobemuchmoreintelligentandthepricestopaytobefarbetterinformed.Theuserclick-pathnowtriggerscountsinSEMreporting(suchasAdwords).AdServerreportingandBidManagementreportingatpoint(A).TheuserlandsonthelandingpagetriggeringtheBidmanager’slandingpagetag(B)demonstratingthatthekeywordcanbegiven higher value since it drove a site-side action. The landing page also triggers aconversiontag(C)-moreinfoonthisinChapter4.Theuserdataabouttheclickandtheconversion aswell as user data about other engagements across othermedia on otherchannelsfortheclickingusersispassedinto(D)“P2CReporting”(seeChapter5)andthedata is collected in theBidManagement systemat (E). Inside theBidManagementSystem,themarketerisfreetoapplyanattributionmodel(SeeChapter4)tothekeywordinformationtoestablishitsvaluebeforeadjustingthebidpriceat(F).

ExtendingtrackedPaidSearchtothelongtail

GoogleAdwordsproducthasasistertoolsetcalledGoogleAdsense.AdsenseallowssmallPublisherstoattractCPC(Costperclick)revenuesthroughAdvertising,whichotherwisewouldbe completelyunattainable since the amountofuser traffic to such sites remainsverysmall.NichePublishersiteshaveveryspecificcontent that,despitehavingasmallamountoftrafficmighthaveaperfectlysuitedaudiencefora“contextuallytargeted”ad.

AdsenseworksbyhavingthePublisherdownloadAdsensecodefromGoogle’ssite(aftersigning up for an account and inserting bank details) and sees thePublisher embed thecodeonanypageoftheirsite.UsersthenarriveatthesiteandtheAdsensecodereadsthecontentofthepage.ThiscontentisthencheckedagainsttheAdwordsdatabasestoseeifthereareanymatchingkeywordswhich it thendisplays to theuser.TheseareAdwordspaidsearchads,appearingintheAdsensespotonthepage.SinceAdsensedrawsfromtheAdwordsdatabaseitself,Adsenseusesthesameclick-throughURLsasadwords,ifthesekeywordsaretrackedinthethirdpartyadservertheywillalsobetrackedbyAdsense.

Advertiser’s should have the option of separating out the traffic coming fromAdsense supported Publisher sites versus those coming from the Search Engineitself by only applying certain SEM campaigns in the SEM interface to either

Adwords or Adsense since the traffic quality might be different. It is worth theAdvertiser checking with the Search engine directly to see if campaigns beseparatedout.Iftheycan,thetraffickershouldalsoseparateoutthesecampaignsintheadserversothatitisclearintheadserverreporting.

Figure3.7.6GoogleAdsense ispoweredby theGoogleAdwordsengine.When theuserloads the Publisher page, the Adsense code collects the contextual information on thepageandpasses it into theAdwords engine before the relevant ads are displayed.ThisdiagramissimplifiedhoweverAdsensewillrarelyscrapethepageinrealtime.RathertheGoogle spiders will have cached the page in the Adwords engine in the past week toassociate theURLwith therelevantcontext, thenAdsense just sends theURL toenginewhichchecksadsassociatedwiththatcontextualinformationwhicharethenservedout.Theclick-through-URLswillstillcontainthepathwayasusedinthethirdpartyadservertocounttheclicksthere.

ExtendingtrackedPaidSearchtoMobile

Just as with separating out Adwords campaigns from Adsense campaigns, the SearchEngine should discern those campaigns delivered to Mobile from those delivered todesktop devices.Mobile devices offer the geo location facility and as such, the searchkeywords are likely to be more relevant to user and therefore more valuable to theAdvertiser.WhenbiddingforkeywordsintheSEMinterfaceorthroughthebidmanager,thegeolocationrelevantadsarelikelytocosttheAdvertisermoreiftheresultingactivityis likelytobringinmorerevenuefor theAdvertiser.If themobilesearchactivitydrawsfromthesamedatabaseofkeywordsandclickthroughURLsthatthemarketersetsupinthe SEM interface, in the bidmanager and in the third party ad server then it will betrackedinthirdpartyadserving.Again,tobreakthisoutintothecorrectkindofreporting,the campaigns should be sufficiently labelled in the Search Engine SEM interface andcreatedasseparatecampaignsintheadserverunderthesameAdvertiseraccount.

GooglehaveutilisedtheirtechnicalabilitytoallowAdvertisersofallsizestotakeadvantageofadvertisingtoabulknumberofPublishersthrough“GDN”theGoogle

Display Network. This is not search activity but is more closely aligned withAffiliateadvertisingcoveredoffinthenextsection.

SearchandotherChannels

Tracking and counting paid search provides a wealth of data within the third party adserver that can be used to create inventive strategies to improve the optimization andservingofcreativeacrossotherchannels.Forinstance,ifusersaresearchingonspecifickeywords, thosecanbepaidforthroughSEMandthetrafficcanbedriventoaspecificAdvertiserlandingpagewhichmayhousearetargetingpixel(SeeChapter7)whichcouldbeusedtopowerafollowupDCOcampaigniftheuserdoesnotgoontobuytheproductinquestion.

Paid Search is often cited as being one of the most effective marketing channels fordrivingperformanceactivities.ButfromtheAdvertiser’sperspective,anychannelthatiseffectivelydrivinguserwebtrafficthroughtotheAdvertisersiteshouldbefundedbythemarketing team so as to make those channels even more effective. Owing to the waySearchEngineswork, paying for an expensivewebsite that isSearchEngineoptimizedcan be an effective use of this budget and SEO (or “Natural Search”) therefore istransformedintoachannelthatneedstobetrackedalongsidetheotherchannels.Itisclearto see how this channel differs from the others, there is no procurement of advertisingspace,yettheAdvertiser’ssiteappearsonthesearchresultspage.Itisimportanttotouchbrieflyonhowthisworks.

WhatisSEO?

SEO(SearchEngineOptimization)describestheon-goingdevelopmentofawebsiteforitto be picked up naturally by the way a search engines work such that the pages aredisplayed in the standard search listings. These “natural search” listings appear in thecentreof thesearchresultspageand traildownthepageontomultiplepages.Thecopyandlinkswhichmakeuptheseresultsaregeneratedbythesearchengine.Astheresultspageloadsinformationcomesfromthesearchenginedatabasewhichiskeptupdatedbythe search engine’s web crawling technology which are called “Spiders”. Spiders arecontinuously crawling the web in accordance with the natural search algorithm whichrankswhichpagesarethemostrelevanttouser’ssearchtermandreturnstothedatabasetherightinformation.

Figure3.7.7ThespidersorwebcrawlersfromtheSearchEngine(1),crawlthroughtheweb’sURLstoreachwebsites(2).Thecrawlersreadthesourcecodeofthepage(3)andpass the required informationabout thatURLback to theSearchEngineDatabase (4).TheinformationisstoredasacacheintheSearchEnginedatabase.CrawlersorSpidersrepeattheprocessofcrawlingpagesmoreoftendependingonhowpopulartheyarewhichcanbedeterminedby thenumberofotherdomainspointingsendingusers traffic to theURL.

Publishersandwebsiteownersworkveryhardtotrytoget theirpageseenat thetopofthese results and because these results are unpaid, the cost to the advertising effort isindirect.Asstated,itwouldnotbetruetosaytheseresultsarefreesinceitisthesweatandtearsof thewebsitedevelopment team thatensureahigh“page rank” in theengine.TojustifythecostofthemanhoursputintodevelopingagreatSEOstrategy,theSEOresultsneedtobetrackedtoestablishtheeffectivenessofthestrategy.

SitesideanalyticstoolssuchasAdobe’sOmnitureorGoogle’sGoogleAnalyticsareveryeffective at detecting the incoming site “traffic” from these engines and understandingwhere it came from toquantify the effectof theSEOefforts.Having said this site sideanalyticstoolscanrarelytietogetherauser’sbehaviourwiththenaturalsearchkeywordsalongsideotherpaidforchannelssuchasDisplayandInstreamVideo.Alsosuchresultsare better verified by an independent third party so as not to allow for favourwith thePublisherowningthemediathatisdrivingthetrafficinthefirstplace.Thethirdpartyadserverswere never originally designed to track this kind of activity but this need is inincreasingdemandfromAdvertisersandMediaAgencies.

ReferralURLtotriggeraclick

As a reminder fromChapter 2 -Advertisers andMarketers, the user that arrives at theAdvertisersiteleavesatrailbehindthemforthelastaction,givingawayhowtheyarrivedatthesitecalleda“referralURL”.ThereferralURLgivestheSiteSideAnalyticstoolanidentifier fromwhich thesearch term(that theuserentered into theengine toget to thesite) can be obtained. Site SideAnalytics tools do this by using JavaScript. This sameJavaScriptmethodcanbewrittenoutbythewebsitedeveloper,separate to theanalyticstoolandbeusedtotriggeraclicktracker.Suddenlytheuserreceivesanadservercookiewheretheywouldnothavegotonefromthenaturalsearchresultitself.Thismakesitlookinadserverreportingthattheuserclickedonthisclicktrackertogettothesiteratherin

thatthesitearrivaltriggeredtheclicktracker.TheresultisthattheSEOchannelistrackedbythissimpleuseraction.SuchJavaScriptcodeshoulddeployedonallpossiblelandingpagesoftheAdvertisersitethatausermightbeabletoaccessfromasearchengine.

Behindthescenesinthethirdpartyadserverinterfacethesetupisalittlemorecomplexthancreatinga trackingplacementandusing theclick trackerportion tomount into thescriptonthepage.Thetraffickerhastwooptions:

1. Either the script can be designed so that each possible keyword triggers adifferent click tracker.Thiswouldmean the trafficker is given a list of thekeywords by the Advertiser website management team (this list can beexported from the site sideanalytics tool).The traffickerwould thencreateonetrackingplacementforeachkeywordandtheclickportionofthetagsbetraffickedandsentback to thewebsitemanagement team. In turn this teamwouldtweaktheJavaScriptonthepagesothatwhentheuserarrivesonthepage, thesearch term that informed theirentrypoint to thesitebecapturedand the corresponding click-through URL be triggered. Reporting for thisoptionisverysimplesincethetraffickerjustpullsastandardcampaignreportand the keywords are already pre-defined in the report. The difficultywiththis method is that a JavaScript would not be able to house too manykeywordsnorwouldthetraffickerwanttocreateormanagetoomanytagstosupportit.

2. ..orthescriptcanbedesignedtopasstheexactkeywordsearchtermfromthereferrerURLintoadynamictokeninthethirdpartyclicktracker.Thiswouldmean the trafficker only creates one ad in the ad server and having thewebsitemanagement teamimplementingtheclickportionontheAdvertisersiteinsidethepurpose-builtJavaScript.TheJavaScriptwouldbetweakedsothatastheuserarrivestothesitetheclickisfiredandthekeywordportioninthereferrerURLisdynamicallyinsertedintotheclickcallbackintoadserverreporting.Thetraffickerwouldthenneedtoensurethatwhenreportingontheclicksfromthisadinthecampaignreportsthatthecontentsofthedynamictoken is displayed and aggregated so that themost popular keyword entrypointscanbeascertained.Thiskindofreportingcanbehardertosetup.

Figure3.7.8Theusertypesasearchtermintothesearchengineandistakentoasearchresults page.On this page the natural search results are listed running down the pagewhichmatchtorelevantsites.TheuserclicksononeoftheseresultsandistakentotheAdvertiser site. One the landing page is embedded a script that either fires off thecorrespondingadserverclick tracker for thesearch term(1)or inserts thesearch termintoadynamictokeninthesingleclicktrackeroptionavailabletotheAdvertiser(2).Ineitherinstancetheinformationissuppliedtoadserverreporting.

IfthesesetupsarefollowedtheseclickswillbetreatedlikeclicksintheDisplaychannelorclicksinthePaidSearchchannelandthiswillcreateasituationwhereconversions(seeChapter4)onSEOunderthelastclickattributionmodel(seeChapter4)willtakethefullcredit for the conversion and actual paid-for media will be eclipsed by this activity.Although this may sound complex: what it really means is that natural search willsuddenlyappeartobemoreeffectivethanothermediachannelstrackedintheadserver:somuchso that itmightconvince themarketer tostoppaying for thepaid-forchannelsaltogether.Thisissomethingofanillusionbecauseifthemarketerstopspayingformediainpaid for channels, the site side analytics toolswill report a completedrop off in sitetraffic altogether and conversions will also fall short (see Chapter 4). Natural searchresultsareverycommonlyusedbuttheyshouldbethoughtofasoneoftheinfluencersintheuser’spathwaytotheirtransactionorjourneyonortotheAdvertisersiteratherthanthe only influence. For this reason the natural search tracking activity needs to bedownplayedinadserverreporting(seeChapter5-ChannelSpecificReporting).

AdServingcostsforSearch

third party ad servers usually charge Advertisers and Agencies based on impressionvolumes for ads served.Sinceno image is beingdelivered for theSearch channels, theproviderwillnormallychargebasedonclicksonlyfortheclicktrackers.

Thisintroductionintotheworldofpaidsearchandnaturalsearchexposesingreaterdetailtheimportanceandrelevanceofthirdpartyadserverreportingandconversioncalculation.

The next section continues with the theme of performance marketing, looking at whathappenswhenPublishersbegin to realise that thebestway to attainuser attention is toutilise the SearchEngines themselves. Itmay be true to say that such Publishers valuetheiradspaceovertheircontentandassuchtransformtheirsitesintoconduitstochannelusertrafficthroughtoAdvertisersites.ThesePublisherwebsitesarecalled“Affiliates”.

Chapter3-Section8AffiliatesandAffiliateNetworks• AffiliateSitesandNetworks

• HowtheAffiliateEarns

• AuditingthequalityoftheLongTail

• Figure3.8.1

• TrackingAffiliateActivityinthirdpartyadserving

• Figure3.8.2

• Theabsenceofimpressiontracking

• Affiliatesthatgrow

Thissectionwill lookatAffiliateadvertisingasachannelandput it into thecontextofthirdpartyadserving.LikeSearch,adservershaveagreaterroletoplayintrackingratherthan serving thephysical ads and so thebenefits of amarketer,Advertiser andAgencyusingathirdpartyadserverinthisspacearethesameasapplytotheworldofpaidandnaturalsearch.WewillbeginbydescribingwhatAffiliateadvertisingisbeforedivingintothecomplexitybehindthetrackingmethodologiesrequiredandthereportingcapabilitiesthat this provides. Being a performance marketing channel, Advertisers which areconsidered to be Direct response Advertisers are more common users of the AffiliateChannelthanthoseseekingtodobigbrandingexercises.ThisisbecauseAffiliateactivityatscalecanbeacosteffectivemethodtodrivesalesandotherconversionactivities(seeChapter4).

AffiliateSitesandNetworks

The landscapeofAffiliateadvertisingasadigitalmarketingchannel ismadeupof twocore entities: the Affiliates websites and the Affiliate networks which house them.Typically speaking the Affiliate network supplies the digital creative (core images and*.swffiles)andclicktrackerstoamassofAffiliatewebsites.

LargeAdvertisers,particularlyinthegamblingandonlineretailverticals,havetheirownAffiliateprogramssodonotalwaysutilizeAffiliatenetworks. Insteadactasthenetworkthemselvesbystoringthecreativeanddistributingtheclicktrackers.Indoingso,theAdvertisertakesontheresponsibilitytoprovidetheAffiliatewebsiteswiththeirreportingdata.

AffiliatesitesgoontoearnacommissionwhenusersclicktheaffiliatelinksandpurchaseontheAdvertisersiterightawayorwithinagiventimeframe.Itisinthebestinterestofthe network to drive traffic that will end with a purchase, so often the Network willencourageaffiliatestousethecontextoftheirdigitalpublicationstoincentivizetheusertoclickonthead.TypicallyspeakingeightypercentofalltrafficisgeneratedbyAffiliatesthat offer an incentive scheme. That is that they offer a discount, a general offer withmoneyoff,orsomeothertypeofmonetaryincentivesuchasvirtualpoints.TheremainingtwentypercentofAffiliatestendtokeepthemselvescontentfocusedsuchthatthealready

nichenatureof theirwebsitesutilizes theSEOframeworkof thesitedesign,and this initself attracts users. The user searches a relevant term in the search engine, finds theAffiliate site perhaps halfway down the page of the search results and clicks to theAffiliate site. Here they find content of interest and a contextually relevant ad, whichdrivesaclickandpotentiallyasale.

HowtheAffiliateEarns

NetworkssuchasTradeDoublerandCommissionJunction havegradually attracted theattention of smaller Publisher sites that sign up to the network to become Affiliates.Withintheinterfacesuppliedbythenetwork,thewebsitemanager(forthenowAffiliatewebsite) browses the ads available to the Affiliates on the network and enrolls in thecampaign.Enrollment releases the ability for theAffiliate to implant the adon the site.Clicks on the ad are usually counted by theAffiliate network and are displayed in thatinterfacealongsideattributedconversions(seeChapter4)sothatAffiliatescandeterminetheir daily revenues. This method does not work for all small Publishers, since thePublisher is essentially displaying the ads for free and only getting paidwhen the userclicks or performs a conversion activity after clicking. This can result in the Publishergivingawaymillionsofimpressionsforfreewithnoreturnfortheuseoftheadspace.

AuditingthequalityoftheLongTail

ItcanbeveryhardforaNetworkandAdvertisertoauditthesesmallPublishersitesonacontinual basis, which can result in some negative consequences for brand image.Generally speaking an Affiliate site will be tailored towards the purchase activity andtherefore will tend to abide by an Advertiser’s branding regulations which should besupplied by the Network in the enrollment phase. This lack of control has put someAdvertisers off of using affiliates but newdevelopments in verification reporting in thethirdpartyadserverarelikelytoremedythisoverthenextfewyears(seeChapter5-AdVerification).

WithoutusingtechnologysolutionstherearechecksthatcanbedoneontheAffiliatesitessimplyusingtheoperationalheadcountavailableeitherattheMediaAgency,theAffiliatenetwork or on the Advertiser’s marketing team. There are two processes for policing:“Affiliate Auditing” & “Affiliate Screening”- both of which would be outlined to theAffiliateintheenrollmentphase.

AuditingisusuallyaquarterlyprocessthatinvolveslookingovereveryindividualadthatisbeingshownacrossallsitesandcheckingcopyandadpositiontoseethatitalignstotheAdvertiser’sbrandingregulations.

Screeningisaprocessofacceptingordecliningaffiliateswhoapplytojointheprogram.Theadvertiseroraffiliatenetworkcandecidewhichaffiliate isgoodenough to join thecampaignandthuscanstoplowqualitysitesfromenrolling.

Figure3.8.1Affiliation extends the distance between the planners and buyers, and theinventorythatisbeingprocured.Toensurethatthefinallocationoftheadissafeforthebrand to sit inside there are several safety nets both operationally and technologically.TheauditingandscreeningprocessiscoveredindetailhereandVerificationiscoveredindepthinChapter5.Itisworthnotingthatverificationworksoffoftheimpressionratherthantheclickmetriconwhichmostaffiliationisbased,sothoroughverificationreportingisnotavailableforcampaignsconductedacrossaclick-basedbuyingmodel.

TrackingAffiliateActivityinthirdpartyadserving

InthethirdpartyadserverthetraffickerwillseethenameofawholeAffiliatenetworkonthemediaplaninplaceofaPublishersitename.Inthethirdpartyadserverinterface,inthe usual trafficking workflow for Display activity, the affiliate network name can beselected under the ad server site names.Once the site is added to the campaign in theinterface, the traffickergenerates a single trackingplacement for eachAffiliatenetworklisted.Onlytheclickportionoftheplacementisrequiredandwhenthetagistraffickedtotheaffiliatenetwork,itisthisportionthattheAffiliatenetworkoperationsteamloadontothe enrollment page of their network; representing the Advertiser’s campaign. TheAffiliate network then adds information to the page about brand regulations and in itscompletion thepackage isknownasan“affiliateprogram”.When theaffiliate logs intothe interfaceof theaffiliatenetwork, theyenroll in theprogram.Oncetheadisontheirsite,usersthatclicktheactivitywilltriggeracountforaclickinthethirdpartyadserver.

Figure3.8.2Affiliate tracking consists of a single tag or click tracker provided by thetraffickerintheadserverinterface:suppliedtotheAgencyoraffiliatenetworkoperationsteamdirectlywhowillsetupthetagtoloadtheaffiliateidentifierintotheclicktrackerasatokenwhenthetagfires.WhentheuserarrivesontheAffiliatesite,theidentifierforthatsiteispassedintothetokeninthesinglenetworktagastheclickdataispassedintoadserverreporting.Theuserthencontinuesontothelandingpageandtheconversiondata(seeChapter 4) is also captured in the ad server reporting.Within reporting a feed iscompiled that showswhatactivity (clicksandconversions) shouldbeattributed toeachaffiliatewhichcanincludethemonetaryvalueoftheconversionortotalconversions.ThisinformationisthenusedbytheAffiliateNetworktopaytheaffiliates.

SincetheAffiliatenetworkmostlypaysoutonaCPAandnotaCPC,thetraffickerinthethirdpartyadserverinterfacealsoneedstoensurethataconversiontagisgeneratedandisimplementedontheAdvertisersiteintherightplace(moreonthisinChapter4).

Theabsenceofimpressiontracking

Advertisers do not use the impression portion of the ad tag in their Affiliate activitybecause theaffiliateprogram isusuallyopenedup toanunknownnumberofAffiliates,whowilldeliveranunknownquantityof impressionsof thead inorder togetclicksorconversionevents.Affiliatesarecommonlyaskedtohousethecreativeorimagesontheirown servers so that even theAffiliate network is not burdenedwith serving the actualcreative.If theimpressionportionwasusedortheadserved, theadservingcoststotheAdvertiser could be astronomical since it is a very broad base of long tail PublishersdrivingthetraffictotheAdvertisersite.Itwouldalsobeverydifficult tocontactall theAffiliatesitestoaskthemtotakedowntheirimpressionpixelstoceaseadservingentirely.ForthisreasonisnoteffectiveforbrandadvertisertousetheAffiliatechannelforbrandactivitysinceclicksandconversionsalonearenotagoodrepresentationoftruereachandfrequency(seeChapter5-Reporting).

Affiliatesthatgrow

AffiliateactivitydoesnotjustconsistofsmallPublishersites.AffiliationisaverycleverandcomplexdynamicofshrewdPublishersthatuseveryinnovativesitedesignsinordertodriveusertraffictoAdvertisersitesforthepurposeofpurchase.AllsuchAffiliatesstartout smallbut theircontent isarranged inamethod thatattractshighvolumesof traffic.These affiliates are sometimes referred to as “Aggregators” and often break away fromAffiliateNetworks to become largePublishers in their own right, attractingAdvertisers

andAgenciestospendwiththemdirectly.

Just as with Search activity, the Advertiser or Agency will just end up paying the adservingproviderbasedontheclickactivity.ToconcludethischapterwewillturnnowtotheoutlyingchannelsofDigitalAdvertising.Fromanadservingperspective thereareanumberofrestrictionsplacedonthesechannelswhichseverelylimittheexistingtrackingandservingcapabilitiesofadservingtechnology.SuchrestrictionsexistnotbecausetheAdvertisersdonotwant to track theiractivitybut thatPublisher-side technologiesplacerestrictions on the ability for a true independent audit to occur. These channels willcontinuetosufferfromfundingproblemsunlessthePublishersopenthedoors,andoffergreatertransparencyandahandofintegritytothosewiththemarketingbudgettospendacrossawidereachofchannels.

Chapter3-Section9Email,in-gameandSocial• PermissionMarketing

• Figure3.9.1

• SendingEmailsonMass

• EmailMarketingadvertisingmetrics

• ThirdpartyadservertrackingforEmail

• In-gameanddigitaloutdoortracking

• SocialMediatracking

• Figure3.9.2

• Theversatilityoftrackingpixels

This section rounds up the email, in-game and social channels for digital advertising.Thesechannelsarenotsmallchannelsbyanystretchoftheimagination,theygetalotofuserattentionandvastvolumesofinternettrafficthroughthemeverywakingmomentoftheday.Howeverwhentryingtousethirdpartyadservingaseyesacrossallmediabuys,thesechannelsareoutliersandforonereasonoranotherlimittheamountoftrackingandservingthatcanbedone.Thankfully themeagretechnologyof trackingpixelsandclicktrackers allows a very independent eye on delivery and click activity, but beyond thisAdvertisersareessentiallyblindandmustsubmittoPublisherreportingattheverybest.

Email advertising is anoldchannel in theworldofdigital advertising.FewAdvertisersextend their budget to cover it beyond their own existing user base but if it is a trafficdriver to an Advertiser site, users need to at least be tracked for the email channel toappearintheadserverreporting.In-gameisclassedasstandalonesoftwareeitheronaPCor secondarydevice includingconsole.AlongwithEmail, In-gameas a channel suffersthe same setbacks as trying to track activitieswith applications forMobile advertising.Lastlywewilllookatsocial:adominatorintheworldofcontentproductionbutveryslowtomonetizetheirproperties.Socialnetworkscontinuetokeeptheirnetworkslockeddownto third party ad servingwhich seems daft, given that it shuts off potential revenue infavourof less transparency for theAdvertiseronly adding to the speculation that socialmaynotbeaseffectiveforAdvertisersasinitiallythought.

PermissionMarketing

Email advertising gets a bad rap because Advertisers believe that email advertising issynonymoustoSpam.Strictlyspeakingthedifferencebetweenemailmarketingandspamis that the user has provided consent to be sent emails from a third party versus beingdelivered unsolicitedmail. The difficulty with presenting email marketing as a fruitfulchannel for Advertisers is that the methods to obtain consent for email marketing arelumped intosmall illegibleboxesasacondition toaccessaservice.Withsuchmethodstheconsentisprovidedbuttheuserneverreallywantedtoreceiveadvertisingfromathirdparty.

Emailadvertisingfallsintotwobroadcategories:emailsthataresinglepageadverts,andmonthly subscribed emails from a Publisher directlywith ad space around the sides orembeddedamongthecontent.Toreceiveeitherofthesetwoformsofemailadvertising,ausermust signup to the service andopt in to receivemessages from thirdparties.Thestrict opt in requirements demanded by regulators such as theDMA (DirectMarketingAssociation) in theUKmean that theemail advertisingchannel is sometimesknownas“PermissionMarketing”aphrasealsocoinedbytheauthorSethGodin.

Figure 3.9.1 The cursed double-opt-in boxes. On a Publisher site a user might beinterestedintothecontentastooptintoanewsletter.Onthenewsletteroptinpagetheremaybedisplayed twoboxes: (1) is for firstpartyadvertising:can thePublisherdeliverPublisher-specific ads to the user’s inbox? The second (2) asks about third partyadvertising,uponwhichemailadvertisingisfounded.Publishersregularlytrytousethemanipulationoflanguagetoconfusetheuserintooptingin.Examplesinclude“tickheretonot opt out of thirdparty advertising”;whichwill only serve to de-value email as amarketingchannelevenfurther.

Byconsentingtoreceivesuchemails,theuser’semailaddressbecomesacommoditythatthePublisheroremaillistownercanselltoAdvertisersor,withauser’spermissiononasecond tick box, the “double opt-in” can be sold on to other list owners to receiveadditionalmailwherefurtheradvertisingspaceisthenmadeavailabletoAdvertisers.

SendingEmailonMass

ConsentedemailsfromrespectedPublisherscanbecommoditisedbecausetheemailsareopened and read and the ads attract clicks,which can result in conversions (sometimeswithbetterresultsthanstandarddisplay-SeeChapter4).InorderforaPublishertosendthousandsofemailstheyeitherusetheirownlistorbuyoneinfromadataprovidersuchasAxiom.As such the list of emails becomes the property of thePublisher for a givenperiod of time and it is their responsibility to act in accordance to the wishes of theAdvertiser,theuserandtheoriginallistowner.

Emaillistqualityisahugefactordeterminingthesuccessofacampaignanditessentialthat addresses are cleaned regularlyby thedatabaseowner to ensure that anyaddressesrequestinganunsubscribeorwherea“bounceback”isoccurring,areremoved.Tosendthousandsofemails,aPublisherneedstouseamassemaildeliverysystemsuchasEmail

Vision.The list is simplyuploaded to thedeliveryplatform, theHTMLemailcompiled,andthensentasanemailoveragivenperiodofdelivery.Tobecertaintheemailsarenotblocked, thePublisher needs to be sure they are sending from awhitelisted IP address.TheycanworkwiththeemaildeliveryplatformandmajorPublisherslikeMSNorYahoo!to check that the IP is not on a blacklistwhere other spam email sender addresses areblocked.Here,theemailproviderisactingasthegate-keeperforwhatcanandcannotbeloadedintoemails.IfanIPaddressisdeemeddangerous,linksorimagescanbeblockedandinsomeinstancesemailscanendupinjunkmailfolders.

EmailMarketingadvertisingmetrics

Anemaildeliveryplatformcanmeasurefourcoremetrics.

1. Open rate, detecting howmany of the emails sentwere viewed in themainreadingpaneoftheemailprogramorweb-basedapplication.

2. Click through rate (CTR) for the subsequent clickable areas or URLsembeddedintheemail.

3. Bounce rate (i.e. the number of emails thatwere never delivered to inboxesbecausetheyreceivedabounce-back).Theseemailaddressesaresentbacktothe listownerafter thecampaignends to investigate to see if theyarenowwhollyinactive.

4. Unsubscriberate;needlesstosayahighunsubscriberatefollowingthelaunchofanemailcampaign, isastormcloud thatneitherAdvertisernoPublisherwishesfor.Theuserthatclickstheunsubscribebutton,whichshouldappearat the bottom of every Publisher-owned email, will be removed from theoriginallist(whichwilleitherbePublisherownedorhired).

ThirdPartyadservertrackingforEmail

Connectingthethirdpartyadservertotheemailmarketingchannelisimportanttoensurethat thischannel is independentlytrackedandthat theemailchannel isconsideredasaninfluenceintheuser’spathwaytoconversion(seeChapter4).

The trafficker in the third party ad serving interfacewould lookon themedia plan andmightseeaPublishersnamefollowedbyamentionofthenameofanewsletter.Thissamereferencewouldneedtobeincludedinthethirdpartyadservingsitelist.Thetraffickersselectsthisasasiteandgeneratesa1x1pixelsandclicktrackersandsendstheseontothePublisheradoperationsteamtoensuretheyembedthemintotheemailcreative.Truethirdpartyadserving(evenofimagesinsideemails)isneverundertaken.Moreoftenthannot,theimpressiontrackerisalsowhollyabandonedandonlytheclicktrackerisimplemented.Here,theemailproviderisconsideringthethirdpartyadserversIPaddresstobelongtoablacklistbecause there is suchawideusageof the IPacrossmillionsofotheremailsatsomepointinthepast.Thismeansthatpixelneverloadsandthereforeneverrecordsanimpression.

In addition not all email provider interfaces are setup to receive emails that can renderimagesproperly.Thosethatarenot‘HTMLfriendly’willrenderatext-onlyversionoftheemail. In such adelivery the imagepixelwouldnever loadbut theusermaychoose toclick throughvia thesupplied link,whichwould route through the thirdpartyadserver

beforelandingtheuseronthelandingpage.

Duetothesimplenatureofemailandemaildeliverythereisnopossibilityofrecordingdwell rates or interaction rates, but conversion tracking is possible so long as the clicktrackersareinplaceontheemailcreative(moreinChapter4).

Email lists have evolved significantly in recent years. Formerly they were largedatabasesofPublisherowneddataaboutuserswhereuserswererecognisedbyanemailaddresscollectedwiththeirpermission.NowlistownersandPublishershavegone a step further and logged the cookieIDof the users using first party cookieinformationandhavealsolabelleduptheusersintoaudiencetypesbyloggingtheirinterests basedonwhat they arebeing exposed to andwhat they are clickingon.Rather thansell thesedatabases toAdvertiser’sasemail lists, theyarenowbeingsold as sources of data to be inserted into DSP’s and DMP’s to improveprogrammaticbuyingandtrading(seeChapter8).

In-gameanddigitaloutdoortracking

Despite the promise ofmass in-game advertising, high reach games remain a premiumofferingtoplayersandthereforeadvertisingrarelyhasmuchimpactinthisspace.Ingameadswoulddelivercreativetoanonlinegameviaafeed(suchasanXMLfeed).Thesameistrueofdigitaloutdooradvertisingaswell(contentcanbekeptregularlyuptodateviaanXML feed) but the nature of the games console and standalone gaming applicationsoutsideof thebrowser,mean that the thirdpartyad servingcookiescannotbedropped.ThishinderstheabilityforAdvertiserstoestablishhowmuchvaluethegameenvironmentreally provides to the overall cross-channel ad campaign. Until such time as in-gamecookiesaredroppedandcaneasilybematchedwith thoseof theadserver; there is lessstrengthinrunningsuchcampaignsastheimpactofthecreativeitselfinthecontextofamulti-channel campaign is lost. There may be some hope in second screen audiowatermarking technologies but at time of writing, nothing scalable has come onto themarket.

SocialMediatracking

Digital advertising through social channels is still in its infancy and as yet the linkbetweenthirdpartyadservingandtheseenvironmentsisrathertentative.SocialmediainthecontextofDigitalAdvertisingcanbedividedupintothreeparts:

1. ThemassadvertisingthatappearsontherightsideofthepageforFacebook.Thisismanaged by an advertising exchange program called FBX (see Chapter 8). Thisscaled technology ismuchmorecloselyaligned to theworldofPaidSearch thanthatofdisplayadvertising.

2. Adsthatappearintheuserfeed.Theseadspotsaretechnicallyunlimitedanddependontheuserengagementwiththeprivatefeed.Adsarepurposedfromcontentonfanpagesorelsewhereonthewebbuthavelimitedengagementcapabilities.

3. Thehomeof theAdvertiseron thesocialnetwork. In thecaseofFacebook, this isthe fan page, where is has been notoriously challenging to embed tracking andtagging technologiessuchasconversion tags (seeChapter4).Meanwhile ‘tweets’and ‘likes’; the measures of user engagement endorsed by the social networks

TwitterandFacebook,donotpassthisinformationbacktothethirdpartyadserverusinganyreservedmethod(soofficiallythemetricsarenotsupportedbythirdpartyadservers).

Figure3.9.2Still in its infancy: thead ecosystemof thepopular social networking siteFacebookconsistsof3corepropertiesforAds.FBX(whichcatersforthepaidscaledandtargeted advertising on the right navigation: drawing from user data), the Advertiser’sFanPage (forsomeAdvertisers this ismoreofahub than theirownwebsites)andadswhich appear in the user’s news feed (commonly content is repurposed from the FanPage).

In all of these situations one fact rings true over all others as a cry of utter restriction.FacebookandTwitterdonotallowJavaScripttagsfromanexternalsourceontheirsitesas amatter of policy.Thismeans that should these sites appear on themedia plan, thetraffickermerelysetsupimpressionandclicktrackersforeachplacement.Thetraffickerthen sends the resulting tags along to theMediaAgency to either embed into theFBXaccesspointorworkwithtechnologyprovidersthatarerepurposingAdvertisercontentasads (suchasUpcast orExperian’sAlchemySocial product suite). This data leaves verylimited performance information (clicks and conversions) and delivery information(impressions)insideadserverreportingfortheAdvertiserandAgencytosee.

Theversatilityoftrackingpixels

ItisclearthattrackingpixelsandclicktrackersareveryversatiletechnologiesandallowfortrackinginavarietyofchannelsincludingamajorityofthenewestchannelsavailabletoAdvertisers.Theconcernwithpawningoutavastqualityoftrackersandnotconductingfulladservingisthatthereisonlytheconfirmationthatapixelwaseverdeliveredonasiteandthataclickwasinstigated.Thereisnoproofassuchthatthetrueadwasdeliveredorwho saw it. Publishers that today only accept tracking pixelswill need to go a stepfurtherandprovidegreatertransparencytothirdpartyadserversandtechnologyprovidersiftheyaretoattractfullinvestmentfromAdvertiserswithbigbudgets.

This concludes theChannel byChannel campaign setup chapter of this book.We havegonemeticulouslythroughtheworkofthetraffickerinthesetupofadcampaignsviathethird party ad serving interface, from Display, RichMedia, DCO, Instream Video andMobile,toPaidSearch,NaturalSearch,Affiliates,EmailandtheSocialchannel.Butthecampaign setup for all channels by the trafficker is only complete when the trafficker

generatesnotjusttheadtagsbutalsotheconversiontags.ThenextChapterlooksindepthat the importance of Conversion Tags and how the trafficker uses the concept ofconversionattributiontodiscerntheoptimumtargetingandaudiencesettingsavailableinthecampaignsetup.

Chapter4AttributionandAudiences

Chapter4-Section1ConversiontagsandAttribution

• Whatareconversions?

o Figure4.1

• Wheretoplacetheconversiontag?

o Figure4.1.1

• Wheretohaveconversiontags(table)

• CPA

• Conversiontagtypes-SalesVs.CounterTags

o Figure4.1.2

• Extendeddata/ExtendedParameters

• Figure4.1.3

• Figure4.1.4

• Howdoyoutestconversions?

• ConversionTagFunctionality-Piggybacking

o ContainerTags

o Certifiedcontainertagproviders

o Whichconversiontagfiresfirst?

o SelectivePinging

o Figure4.1.5

o PiggybackingwithExtendeddata

o Figure4.1.6

• ConversionTagFunctionality-ConversionWindows

o WindowSettings

o Post-impressionandPost-clickwindows

o Longandshortconversionwindows

o ShadowTagging

o ReportingonConversions

o RepeatConversionWindows

• ConversionTagFunctionality-AttributionModelling

• Figure4.1.7

• AttributionSettings

• EngagementMapping

• Figure4.1.8

• CustomAttributionModelling

• Challengesinimplementingconversiontags

• TagManagementSolutions

• Figure4.1.9

• UniversalTaggingSecurity

InChapter3weexploredthecampaignsetupchannelbutall theeffortsof thetraffickercome down to one question for the marketer: What is the goal of the advertisingcampaign?Whatdoessuccesslooklike?Forbigbrandclientsthegoalisusuallytoagetmessageout toa targetaudience, tobuildawarenessandbrandaffinitybut theultimategoal (as is the goalwithmore direct response styleAdvertisers) is to drive sales. Thissectionwillgointotechnicaldetailabout“conversiontags”;atechnologythatcanbeusednotjustformeasuringsalesbutalsoeventsthatmightbedeemedagoalforthemarketingteam. Beyond this it is important to establish exactly what activity drove the use to aconversion;untanglingalloftheirmarketingengagementswithanAdvertisertoestablishhowtoattributetheconversiontoapaid-foractivity.

Whatareconversions?

It’s all very well measuring the delivery of ads and collecting information about howmanyuniqueuserssawandclickedadsbutthevastmajorityofAdvertiserswanttoknowhowmanyusers then landedon theirwebsiteandhowmanywenton tobuyagoodorservice.ThiscanbeascertainedbymonitoringthetrafficdrivendirectlytothepurchasepagesoftheAdvertiserwebsite.

Conversions by definition vary by vertical. Here are some examples of uses for theconversiontag:

• Automotive-TrackingBrochurerequestsortestdrivesign-ups

• CPG/FMCG-Trackingsamplerequestsorcoupondownloads

• Retail-Trackingsalesthroughthesiteorclickstothesalesphoneline

• Entertainment-Salesthroughtoticketpurchaseorlocalcinemalistings

• Finance-LeadsgeneratedthroughInsurancequotes

• Telco-FormcompletionforafreeSIMcard

AconversiontagisapieceofJavaScriptcode;generatedinsidethethirdpartyadserverinterface by the traffickerwhich is sent to the sitemanagement team of theAdvertiserwebsite to upload into the source code of the page. The trafficker can output theconversion tag invarious formats (suchas iframeVs. JavaScript)dependingonwhat iseasiestfortheAdvertiser’swebsitemanagementteamtoimplement.

Onceimplementedonthepage,theconversiontagwillallowtheAdvertisertowatchforincoming traffic and recognise it asoriginating froma thirdparty servedor trackedad.

The tag, once present on the webpage can recognise the difference between ad-driventraffictoitselfwhichiscalledisa“Conversion”,andnon-addriventraffic(whichcarriesvarious names depending on which technology is being used; theAtlas third party adservercallsthisan“Action”).

Figure4.1Conversionsdetectforthepresenceofanexistingthirdpartyadservingcookieonauser’sbrowser.Thiswillhavebeendroppedontotheuser’sbrowserbythedeliveryofanadthattheuserwouldhaveencounteredpreviously.TheconversiontagthenpassesthecookieID toad server reportingwhen it loadson theAdvertiser site toallowad serverreporting to place the conversion event, marked with a timestamp to the cookieID incookie level data. In aggregated reporting the Publisher which featured the ad wouldshowasingleconversion.IncomparisonifauservisitstheAdvertisersitedirectly,havingnot previously encountered an ad or having deleted their cookies, will still trigger theconversiontagonthepage.Thisconversiontagwillupdatetheadserverreportingwithanactioninsteadofaconversion.ThiswilladdarecordtotheconversionsandactionsdatabaseintheadserverandnotbemarkedoranycookieleveldataagainstanyexistingcookieID.

Oncethe“conversiontag”(alsocalleda“Spotlighttag”inDoubleClickoran“Actiontag”inAtlas) is generated and placed on a page it begins to work, feeding conversion andaction informationback to the thirdpartyadserver. Itdoes thisby looking to theusersbrowser when they arrive on the page and once the tag code renders, seeks a cookiebelongingtothethirdpartyadserver;acookiewhichwouldhavebeendroppedbyathirdpartyadserverviaanadinthepast.

Iftheconversiontagfindsacookie;theadservermatchesthecookieIDtotheactionoftheconversiontagfiringandintheadserverdatabaseaddsanewrowtoamastertable,torepresentthattheconversionoccurred,whenitoccurredandwhichcookieitoccurredon.Ifnocookieisfound,theadserverstillmakesalogthattheconversiontagwastriggeredandatwhattime,labellingitasanAction.

Wheretoplacetheconversiontag?

TheAdvertiserisadvisedtoplaceaconversiontagonwhateverpagesrelevanttomeasurethesuccessofthecampaign.PutaconversiontagonalandingpageandtheadservercanmeasurethenumberofusersthatgotthroughtotheAdvertisersite.Placeonedeeperonthe pathway to an actual purchase, such as the basket page, and more information isgleaned such as “drop-off rate” which indicates how many users are not making thejourneyallthewaythroughtoapurchase.

Themost useful position for a conversion tag is on the “confirmation” or “thank you”

pageofane-commercesite.Whenausertriggerstheconversiontagontheconfirmationpage; the Advertiser knows that the user has purchased; a purchase which can beattributedbacktotheadwhichdrovetheusertothesiteinthefirstplace.Itisimperativethereforethateachconversiontagisuniquetoitspage‘type’;placingaconversiontagona landing page and then duplicating the code to use on a confirmation pagewill skewresults.This is because the arrival on the two different page types could be counted asmarketingsuccessgoalsbuttheyneedtobesegregatedsinceanarrivalisalessvaluable“conversionevent”thanapurchase.

Figure 4.1.1The paid-formedia channel/Publisher site, drives the user through to theproduct landing page via an ad. There could be a conversion tag on this page (C1) toassesshowmanyusersgetthisfar.Usersthatmakeitasfarasthebasketpagecouldhavethisconversionevent trackedbyanadditionalconversion tag(C2)and thoseusers thatpurchase will trigger the last conversion tag shown (C3). In ad server reporting it ispossible to establish the drop-off of users throughout the process and correlate thedifferencebetweenusersthatforexample,cameinoffofapredominantly-socialpathwayversusthosethatcameinfrommoreofadisplaypathway.Theresultsmightindicatethatusers that seemoredisplayadsrather thansocialadsaremore likely tocauseC3-typeconversions.Sitesideanalyticscanalsomeasuresuchactivitybuttheadserverfactorsintheuser’s engagementwithall ads in the campaign to establishanattributionpathway(coveredofflaterinthissection).

ThepagesontheAdvertisersitesoonbecomemarkersinthemarketingfunnelfortheadcampaignasusersarrivetothesite.Thosethatdropoffatproductpagesareconductingresearch.Those that drop off at the basket page are closer to the buying decision.Thisprogressshouldbereflectedinthenamingconventionadoptedbythetraffickerwhentheconversion tags are created so that it ismuch clearer in reporting howvaluable certainconversioneventsare.

Conversiontagsarealsousefulinotherlocations:

CPA

ThesuccessofacampaigncanbemeasuredbythenumberofcompletedconversionsbutinadditiontheactualsalescanactasatriggertopayingthePublishers.ACPAisacostmodelthatagreesthattheAdvertiserpayoutsapercentageofthevalueofthefinalsaletothePublisherthatdrovetheusertotheconfirmationpage.InreturnthePublisherdelivershowevermanyimpressions,andthusprovideshowevermuchinventorywillberequiredtomeettheincomegoalreturnedthroughtheconversions.

As an example: an Advertiser might need to sell ten thousand suits and have anAdvertisingbudgetof£100,000todothis,butapproachingaPublishersite,findsthattheCPMis£1.Thiswouldmeanthateverytenthuserthatseesthesuitadwouldneedtobuythe suit through the website. The Advertiser might know that the likelihood of thishappening isvery small; a conversion ismore likely towork foroneuser in ahundredthanoneinten.TheadvertisercouldnegotiatewiththePublishertoagreeinsteadtopayonaCPAof£1.ThiswaythebudgetisusedtosellalltenthousandsuitsandthecostislowerfortheAdvertiser.

ClearlyCPAdoesnotwork foreveryPublisher.butwillwork forPublishersdisplayingads toniche audiences for high cost purchases. In such an example theCPAcan returnmorerevenuethanaCPM.

IfanAdvertiser sitecontainsmore thanoneproduct for sale, theconversion tagcanbeusedtoworkoutwhichproductwasactuallypurchasedandhowmuchitcost:

Conversiontagtypes-SalesVs.CounterTags

Conversiontagscomeinafewvarieties;mostadservershavethecapabilityofgeneratingconversiontagswiththesolepurposeofcountingthetrafficorthenumberofconversions(calledacountertag).Thiscanbecomparedtothegenerationofwhatisreferredtoasa“Sales Tag”; which can pick up more information about the conversion in addition tocounting,suchasvalueoftheuserpurchaseornumberofitemsinthebasket.

Technically speaking these tag types are very similar. The differentiation helps thetraffickertorememberthepurposeofthetag.Thisgivesanindicationofthetaglocationwhencampaignsaresetupwithoutconsistentnamingconventions.

Figure 4.1.2The trafficker is best off creating counter type conversion tags for pageswheretheAdvertiserjustwantstomonitortheuser’sjourneytothatpage.Salestagsarebettercreatedforpageswherethemostvaluableconversioneventisoccurringsuchasthepointofsale.

Extendeddata/ExtendedParameters

The sales type of the conversion tag can also be setup to capture specific informationabout a sale. This information ismade up of three groups of parameters referred to as“ExtendedData”or“ExtendedParameters”.Inthecollectionofeachoftheseparameters,theAdvertiser sitemanagement teampass theparameter into thecorresponding field intheconversiontagusingamechanismlikeasimpleJavaScriptcode,beforeitdeliverstheconversioninformationbacktotheadserver.

Figure 4.1.3As the usermakes the purchase on the confirmation page; extended dataparameterscanbesuppliedtotheconversiontagviaascript,andsuppliedintoadserverreporting.

Thethreemaingroupingsofparameterare:

1. “OrderID”:TheOrderIDisauniquefieldgeneratedbytheAdvertiseruponpurchasetodifferentiatebetweenthesales.

2. “Metric Parameters”; number value associatedwith the sale. Themost commonlycollected are two parameter fields called “Quantity” and “Revenue”. Quantityusually reflects the number of items in the basket andRevenue reflects the totalbasketvalue.

3. “StringParameters”; areavailable toaddcontext to thepurchase.Somead serverscallthisinformation“Categoryfields”.Categoryfieldsfollowstrictguidelines;theAdvertiser is permitted to pass information about the sale such as the names ofitemsinthebasketordescriptionoftheitemspurchased,butitisstrictlyprohibitedfortheAdvertisertocollectPersonallyIdentifiableInformation.

Someadservers limit thenumberofMetricandStringParameters thatcanbecollectedbut there israrelyaneedformorethanonehundredfieldsofdata.Mostcampaignsutilisingsalesconversiontagswillcollectbetweenthreeandsixdatapoints.

Figure 4.1.4 An example of a Mediamind sales conversion tag. Such a tag is bestimplementedasclosetothetopofthesourcecodeontheAdvertiserpageaspossible(toensure the conversion event is captured if the user leaves the page before it is fullyloaded).Thisscriptwouldbegeneratedby the traffickerwhowouldsend it through theAdvertiserwebsitemanagementteam.ThereareaseriesofextendedparametersthatthetraffickerhasaddedtotheexampletagshownwhichincludebothMetricParametersandStringParameters.ProductIDisanexampleoftheformerandProductInfoisanexampleof the latter. The tag shown has two portions: the part labelled <noscript> can beimplemented on a page that does not accept JavaScript but only the <script> or the<noscript>portionoftheHTMLshownneedstobeimplementedatall.

Itbecomespartof the traffickingprocess for the trafficker toaddtheMetricandStringParametersasrequiredtothetagbeforesendingitovertheAdvertisersitemanagementteamwithimplementationinstructions.

With regards to the collected data, the ad server simply records the data against aCookieID (which cannot be traced back to an individual’s personal details). ThatCookieIDisassociatedwithanOrderIDwhentheconversiontagfiresandthedetailarerecordedinasinglerowintheadserverreportingdatabase(seechapter5).Hereinsidetheconfinesoftheadserver,noinformationatallaboutthepurchasercanbegleanedexceptthebehaviourofthecookieinquestionbylookinginthedatabasetootherreferencestothe Unique CookieIDs engagements with other ads or conversion tags. A consistentcollectionofinformationforacampaignfromaconversiontag,willallowextendeddataparameters to become columns in ad server aggregated reporting (see chapter 5) and

purchasescanbegroupedsothatusersthatpurchaseaparticularproduct,canbeanalysedfor their buying patterns, ad engagement habits, expenditure and complimentarypurchases.

Once the traffickerhasexported theconversion tag fromtheadserver interface,addingmetric and stringparameters; the sitemanagement teamwill implement the tag.At thisstage,testingouttheimplementationiscrucialtoavoidmishapsindatacollectiondowntheroad.

Howtotestconversiontags

Conversiontagsworkrightoutofthebox.Onceimplemented,theywillcountconversionsin ad server reporting if a user’s cookie has encountered an ad for the sameAdvertiserfromthesameadserverbefore.This isnotuniversal;someadserversallowconversiontagstoonlycountconversionsifusershaveencounteredadsfromacampaignassociatedwiththatconversiontag.

Asanad serverbegins tracking theconversiondata, the reported figures should remainrelativelyconsistent,howeverifadserverreportingshowsafall inconversionsorif theconversionsdonotstartappearingrightafteracampaignisputlive,thentheconversiontagitselfneedstobechecked.

Animplementedconversiontagcanbevisitedjustbylandingontherightpage;iftheuserhasnotencounteredanadbefore,thetagwillfire;triggeringan“action”whichwillshowupinadserverreporting.TheuserarrivingtothepagewheretheconversiontagsitscanuseanHTTPtracer(which,asareminder,measurestheservercallstoandfromthepage).

IftheURLoftheadserverappearsinthelistofservercallswhenthepageisloaded,thenthe conversion tag is present. To check its functionality deeper to determine if theconversiontagfiredoffan“action”versusa“conversion”requirestousertoknowwhatasuccessful conversion count from the conversion tag to the server looks like. Thisinformationcaneitherbeobtainedbycontactingthethirdpartyadservingproviderorinsomeinstancesthecallisthesameandonlyadserverreportingwillshowthattherehasbeen a conversion count. In this call to the server will be carried the extended dataparametersaswell.

The problem here is that ad server reporting processes conversions slower than theprocessing of impressions and clicks, so the conversion activitymay not display in adserverreportingforuptotwentyfourhours.Thatisalongtimetowaittofindoutthattheconversiontagisnotfiringcorrectlyandthatnoconversioneventwasrecordedintheadserverreporting.Forthisreasonit isrecommendedthatthecreationandimplementationoftheconversiontaghappenatleastafewdaysbeforetheadcampaignisduetogolive.If there is a problem with the implementation, it can be tested, remedied and retestedbeforethecampaigngoeslive.

Althoughitisrecommendedthatthetraffickercheckstheconversiontagisimplementedcorrectly before the campaign goes live, checking for the “conversion” in ad serverreportingfortestingpurposeshastwoproblemsforthetester:

1. Themost relevant conversion tags (usually sales tags)willpredominantlybehousedonapageattheendoftheAdvertisersite’suserjourney(alsocalleda

“booking pathway”). If this is indeed a thank-you page, the tester willnormally be required to undertake a transaction to get there. It isrecommended that Advertiser’s therefore have ‘testing’ transactionalinformation that can be inputted into a form (such as a fake credit cardnumber)sothatthetestercanreachthesalestagwithoutincurringacost.AsurprisingnumberofAdvertisershavenot foreseen this requirement andassuchconversiontagtesters(insupportteamsontheAgencysideorworkingforthethirdpartyadservingprovider)undergopersonalexpensetoconducttesting with reimbursement later on. This can result in an embarrassingsituationforanyAdvertiserwhosesitemanagementteamhaveimplementedtheconversiontagincorrectlyonthethank-youpage.Expensivetransactionssuchasholidayscanbeparticularlytoughonthetester.

2. If the conversion was actually measured then the reporting will contain therecordofaconversion-howeverthisconversionwouldnotreallyhavebeenatruesalebutatestsale.Forthisreasonsomeadservershaveconversiontagtest functionality built in. This allows the conversion test data to either bequarantinedfromtheactualconversiondataforwhenthecampaigngoesliveorallowsthetestertoomittheirIPaddressfromthecollectionofcampaign-relevantconversiondata,sothatitonlyappearsintestreporting.

Inreporting,adropinconversionsmostcommonlyoccursbecausethewebdevelopmentteamoftheAdvertisersiteareunawareofthesignificanceofthecode.Aweekendreleasecycleof theAdvertiser site could easily accidently remove the conversion tag andon aMonday morning, when the development team realise the tag has been removed, theywould be alerted in ad server conversion reporting and need to replace the tag. Thetraffickerwouldneed to justexport thesameconversion tagcodeagain to fix this. It iswise for the trafficker tocheckconversion reporting in thead server reporting interfacefrequentlytoensurethattheconversiontaghasnotbeenremoved.Unfortunatelydespiteprograms designed to educate the site management teams about the importance ofconversiontags,suchtechnicalmishapswillstilloccur.

Aconversiontagthatissetupandtestedasworkingcanbeusedtosupportthetrackingofconversionevents forawholecampaignbut thereareadditionalconversion tagsettingsandfunctionality thatareutilised inalmosteverycampaignutilisingconversions.Thesesettingsfallintothreegroups:

a. Piggybacking

b. ConversionWindows

c. AttributionModelling

Theremainderofthissectionwilllookatthesesettingsandexplainhowtheywork,howthetraffickercansetandadjustthemandforwhatpurpose.

a. ConversionTagFunctionality-Piggybacking

ContainerTags

Conversion tags (spotlight and action tags) can be setup to trigger the loading ofotherconversiontags.Thistriggerusually‘fires’whenthepagewiththeconversion

tag on it is loaded. Conversion tags with this functionality are called “Containertags”or“Floodlighttags”andthevastmajorityofconversiontagshavetheabilitytoperform this function.Thecontainer tagcanholda largenumberofother tagsfrom a variety of sources, it can also be used to carry third party code. If theseadditionaltagsoriginatefromanothertechnologyprovider:thesetagsarereferredtoas4thpartytags.

What other conversion tags or third party code would an Advertiser want to“piggyback”ontoaconversiontag?Giventheusefulnessofconversiontagsitisnotjust third party ad servers that want to measure conversions as they happen.PublishersmeasuringCPAmightwanttoputasmallpieceofcodeinthecontainertagsotheyknowstraightawaywhensalesarebeingmade.Inadditiontothis,anytechnologythatdropsacookieonauser’smachineatsometime,willwanttocreateaconnectionbetweenthateventandthearrivalofthatusertothiskeypointinthepurchaseprocess:theconfirmationpageconversiontag.

ItisalsoworthbearinginmindthatsometimesAdvertisersareusingmorethanonetrackingtechnology(doubleadserving)andtheythereforewantthenon-dominantad server’s conversion tag to be piggybacked into the container. The aim of thecontaineristoreduceclutteronthepagebehindthescenes.

WhenanAdvertiserorAgencycompletesatraffickingsheetforthetraffickeratthestart of a campaign, it is customary to supply the trafficker with instructionsregarding piggybacking. The additional conversion tags to piggyback into thecontainer tag are sent across as code and the trafficker inserts this code into thepiggybackingscreeninthethirdpartyadservinginterface.Someadserversprovidethe facility to label each piggybacked tag with a name to identify its origin orpurpose.Thiswaywhenthetraffickermakeschangestothecontainertag,itisclearwhich tag doeswhat. This is particularly usefulwhen there aremultiple tags fordifferent purposes by originating from the same technology provider (thereforehavingvirtuallythesamecodesyntaxandbeingdifficulttouniquelyidentify).

Certifiedconversiontagproviders

SomeadserverssuchasAtlas,alsoemployedaprocesswherebyallpiggybackedtagsneeded tobelong to recognised and certifiedproviders.Topass certification,thetechnologyprovidersoftheadditionalconversiontagswouldneedtosignlegaldocumentation to verify that the piggybacked tagswould not be undertaking badindustry practices (such as the collection of PII or data collection which couldbreachPublishertermssuchasPublisherdataleakage).Thisprocessofcertificationgetsamixedreactionacrosstheindustryassometraffickersfindthatitsignificantlyslowsdown theprocessofpiggybacking. In some instances technologyprovidersrefuse to sign legal documentation which can mean changing plans to use thatproviderortheadserveraltogetherforthatcampaign.

Piggybacking essentially means being able to adjust technology sitting on theAdvertiser’s sitewithoutdisrupting the sitemanagement team (whichwill almostalways slow down the process of tag implementation). This level of accessibilityusually means that swapping tags in and out of container tags becomes a daily

activityforthetrafficker,sittingcompletelyoutsideofcampaignlaunchdates.

Whichconversiontagfiresfirst?

Whena container tag contains several conversion tags, the containerwill call thethird party ad server first, then the piggybacked tags fire off using a randomalgorithm.Thisway thepiggybacked tagsareall treated fairlyandpriority isnotgiven to a particular one. Sometimes the trafficker may want to over-ride thisdefaultwhichiswhereselectivepingingcomesin.

SelectivePinging

On a high trafficweb page, such as the confirmation page of amajorAdvertiserwithanecommercesite, thecontainer tagmighthousetwentyormorethirdpartyconversiontagsortrackingpixels(thephrase‘pixels’isusedbecauseitistypicallythe <noscript> code in the generated conversion tag code,which is implementedinto the container). Calling all of these pixels every time the confirmation pageloads,riskstheuserleavingthepagebeforeallpixelshavehadachancetoload.Ifthe user abandons the page before the container fully loaded then some of thosethirdpartieswon’ttrackanythingatall.

The solution to this is to have particular rules in place on the container tag(controlledfromwithintheuser interfaceof the thirdpartyadserver)whichhavepiggybackedconversiontagsfireoffatdifferenttimes.Themostcommonmethodis to limit the firing of those tags which had nothing to do with the sale. Forinstance,theadservercancheckinamatterofmicrosecondsifthepurchasingusercookieIDwasreferredtotheAdvertisersitefromanadonaspecificPublishersite;itthenfiresthecorrespondingpixelandignorestheotherPublisher-suppliedpixels.Thismethodissometimescalled“SelectivePinging”.

Figure4.1.5 In selective pinging the trafficker has received several conversion tags orpixelcodefromtheAdvertiserorAgencywhohaveexplainedthatthecodecomesfromacouple of the suppliers on the media plan. These conversion tags (referred to in thisdiagram as 4th party tags) are loaded by the trafficker manually into the ad serverpiggybackinginterfacefortheconfirmationpagesales/containertag.ThetraffickerthanadjuststheselectivepinginglogicsothatiftheuserarrivestotheconfirmationpagethatwasoriginallydirectedtotheAdvertisersitebyanadonPublisherA,the4thpartyAtagwouldfireandthe4thpartyBtagwouldnot.

PiggybackingwithExtendeddata

Piggybackedpixelscancapture information that thecontainer tagalreadycollectssuch asOrderIDor anymetric and string extended data.To pass the informationthrough,theadserverutilisesan“adservermacro”whichisa tinypieceofcode,alreadydefinedbytheadserverwhichcanbeplacedintopiggybackedtagsjustasatokenwouldbeinsertedintoanimpressionorclicktracker.Anexampleofsuchamacrowouldbe%orderID%whichwouldbeappendedtotheendofthe4thpartypixelcodetoautomaticallyrecognisetheOrderIDinthecontainertagandpopulatethetokenspacecontainingthismacrowiththesameidentifier.

Beyondtheextendeddata,the4thpartytagcanalsocapturepre-definedadservermacrossuchastheConversiontagID.It isworththetraffickercheckingwiththethirdpartyadservingproviderforafull listoftheavailablepre-definedadservermacros.

Figure4.1.6Theextendeddatapresentontheconfirmationpageispassedintothesalesconversion tag by the customised code on the page (written by the Advertiser sitemanagementteam).Whenthesalestagfires,itcarriestheconversioninformationwiththeextendeddatathroughtoadserverreporting.Thesalesconversiontagisoperatingasacontainertagandfiresthecorrect4thpartytagusingtheSelectivePinginglogic.Uponfiring the 4th party tag calls the 4th party reporting server, carrying with it the sameextendeddatavariablespresentinthesalestag(namelytheOrderIDandtheRevenue).

The conversion tags described here as being piggybacked in the container and thecontainer tag itselfwill report data back to the ad serverwhen it is fired but ad serverreportingwill treat thisdecidewhetherornot theactivitycountedasaconversioneventusingtheconversionwindowsettings.

b. ConversionTagFunctionality-ConversionWindows

WindowSettings

Aconversiontagsendsinformationtoadserverreportingeachtimeitisfired,buttheadserverhastodecideifthisinformationconstitutesa“Conversion”ornot.WehaveseenhowtheadservermakesadecisionbetweenaConversionandanActionbutconsiderthisscenario:

AuserseesanadforproductbelongingtoanAdvertiserandthenswitchesofftheirPC, goes on a long round-the-world trip and then returns to the machine twomonthslater.Ontheirreturn,theuserswitchesontheirPC,opensthebrowserandtypesintheURLoftheAdvertiserwebsitebeforepurchasingthesameproductandfiringtheconversiontagontheconfirmationpageafterthepurchase.

Isthisaconversion?Somereaderswouldsayno.Somewouldsaythatonlyiftheuserclicksonthead,didtheyintendtopurchase.Butinthatinstancewouldatwomonthgapbetweenthatclickandthefinalpurchasestillmeanwecanattributetheconversiontothatclick?Theanswerliesintheabilityofthetraffickertocustomisetheconversiontag’sconversionwindowsettings.

Post-impressionandPost-clickwindows

There is no “right method” for measuring conversions; but the two schools ofconversionmeasurementarecalled“PostImpression”and“PostClick”conversion.

• ApostimpressionconversioniswhenauserhasseenanadbutnotusedthatadasaroutetovisittheAdvertisersitetotriggertheconversiontag(theusermay have typed the URL into the browser directly to visit the site). Theconversionisthenclassedasapost-impressionconversion.

• ApostclickconversioniswhenauserseestheadandclicksonittoarriveattheAdvertiser site.The user then triggers the conversion tag from their byvisitingthepagewheretheconversiontagisplaced.

Thesetwomodelstendtofitdifferentproducttypesandpurchasecycles.Considertheuser exposure to a high impact video ad for a vacuumcleaner.The ad serverreporting records just one impression to that ad by a unique cookieID. The nexteventrecordedby thesamecookieIDisapurchaseof thatvacuumcleaneron theAdvertisersitejustoneweekaftertheexposuretothead.Ifnootherdataexistsforthe same cookieID theAdvertiser can infer that the user converted as a result ofonlyseeingtheadandthisisrecordedasapostimpressionconversion.

The time-lag of oneweek is a consideration that needs to be taken into accountbasedon the fact that theproduct inquestion isahighvaluepurchase (avacuumcleaner)andthattheusermayhavetakensometimetocometoadecisiontobuytheitemafterseeingthead.

The “Post Impressionwindow”canbe controlled and set by the trafficker on theconversion tag sitting on the confirmation page of theAdvertiser site.Given thatvacuumcleanershavelongerpurchaseconsiderationsthatsmallerretailgoods(sayaweekormore),thepost-impressionwindowshouldbesettobeover7days.Thetrafficker goes into the conversion tag settings inside the third party ad serverinterfaceandaddsthissettingbeforetheconversiontagisimplemented.Here,thepostimpressionwindowshouldbesettoavalueover7days.Inthisexamplelet’ssuggestthatthetraffickersetsthewindowat12days.

Thismeansthatwhentheadserverreportingreceivestheconversioneventfromthefired conversion tag, the ad server would count this user as havingmade a postimpression conversion. If the user had made the purchase on the 13th day afterseeingthevideoad,theadserverwouldnotcountthisasaconversionatall.

Post-click conversion windows work in the same way but consider the time lagbetweentheclickontheadandtheconversionactivitybythesameuser.Mostadserversdefault thevaluesofthesewindowsat30daysforpostimpressionand90days forpost click. It is also the case that as adefault, the clickingof anad is amore valuable metric that the impression count on an ad by the same user. If auser’sactivityforanadcampaignleadinguptoaconversionconsistsofaclickinthe first week and an impression in the second (for a conversion window of 3weeks), thentheclickwillbeattributedas themaindriver to theusermakingthefinalpurchaseinadserverreporting,oncetheconversiontagisfired.

Longconversionwindowsandshortconversionwindows

Long conversion windows give the user plenty of time to make a considerationbeforeconvertingbutthequestionis:howlongistoolong?

Typically if a trafficker is too liberal and leaves a conversion windowopenfortoolong,thenuser’sconversionswillbeattributedtoanadthatmay have had no real-world influence on the user’s decision topurchase/convert. Here, a user may scroll past a DCO ad and hardlynotice it on day 1 and then see amore engagingTV ad on day 20 andconvertonday21.InadserverreportingtheattributionfortheconversionwouldbeawardedtotheDCOad.Perhapsiftheconversionwindowwasshortenedto5days(sothattheTVcampaignwouldnotinterferewiththeonlinecampaign)theDCOadscouldbedroppedfromthecampaignsincetheywouldappeartoprovidenodirectinfluence.

However,leavingthewindowtooshortmeansthatthetruevalueoftheadcampaign may not be correctly attributed. This may be because someuserswerebeeninfluencedbyaspecificrichmediaadbutdelayedtheirpurchase for too long, meaning that the influence of that ad was notcapturedinadserverreporting.Whenthetraffickerlooksinreportingandsees no conversions for the Rich Media ad (because the conversionwindowsettingsweretooshort)andthetraffickerdecidestostopservingtherichmediaad,thetotalnumberofsalesmaydrop.

Ad server reporting can also provide good reports for “time to convert” suchreportingshowswhenmostusersareconvertingafterseeingorclickingonadsinthesamecampaignastheconversiontagisassociatedwith(seeChapter5).Insuchreports, considering the collected extended data can help the Advertiser todeterminewhattheconversionwindowsettingsshouldbe.Usingthismethod,whenthepurchasesof theproduct featured in theadcampaignarebeingmade,usingastring such as ‘Product Name’ the Advertiser can spot the difference betweenconversions for the product featured in the ad from those thatwere not featured.Thoseproductsfeaturedintheadaremorelikelytohavebeenpurchasedasaresultoftheadstheyfeaturein,solongastheconvertinguserssawthoseadsinthefirstplace.

Giventhattherearenohardandfastrulestosetthesewindows;experimentationiskeytofindingahappymedium,otherwisetoomanyortwofewconversionscouldbeawardedtothewrongadsorattributedtothewrongPublisherswhenpayingoutonaCPA.

ShadowTagging

SomeadserversallowtheAdvertiserorAgencytoviewtheirtotalconversionsinreporting by more than one conversion window so that stats can be compared(answering the question; is there a difference in the total number of conversionsbetweenshortandlongwindows?).Thoseadserversthatofferafixedwindowontheconversiontagwithnooptiontomakechangesinreportingneedtoutilisethefunctionalityof thecontainer tag to seewhat theeffectwouldbeofadjusting theconversionwindow.Inthisinstancemultipleconversiontagsarecreatedunderthe

same Advertiser with different window settings. These conversion tags are thenpiggybacked onto the container. Comparing the conversions across the differentconversion tags in ad server reporting exposes the effectiveness of the differentwindows.Thismethodissometimesreferredtoas“Shadowtagging”.

ReportingonConversions

Thepresenceofconversionwindowsmeanstheadserverhastodoalotofworktocalculateconversionsasitmustlookbackthroughthecollecteddatafortheentirelength of all windows in order to allow reports to be pulled with an accuratereflectionof totalconversions.Thedataprocessing foradserverswhich typicallyprocessbillionsof impressionseachweekmeansthataccurateconversiondata(attime of publication) is not currently available in real-time unless the ad servingtechnologyhasconsiderablylessinformationtoprocess.Dataistypicallyavailablewithintwentyfourhoursofcollectionforreportingpurposesowing,in-part,totheprocessingofconversionwindowinformation.

RepeatConversionWindows

Sometimesuserswillpurchasesomethingand then refresh theconfirmationpage.Toprevent theadserverfromcountingmorethanoneconversionevent, thereareadjustablesettingsinplacecalledthe“repeatconversionwindow”whichisusuallysetataboutaminuteormore.Thiswayanyrefreshesoftheconversiontagwhichtake place by the same cookie in the same minute, are ignored in conversionprocessingandarenotcountedasconversions.

c. ConversionTagFunctionality-AttributionModelling

Itwouldbeunfairtoawardtheconversiontothelastadthatauserclickedon,sincethiswould suggest that this last adwas completely responsible for a purchase orconversion that followed it. As consumers we usually experience a number ofengagementswithabrandbeforeapurchase.Theseengagementsor“touch-points”(a referenceusedbyFlashtalking) should eachbegiven somecredit for the finalconversion. In practical terms this means that rather than paying the Publisherresponsibleforthelastadthatwasclicked,(aswouldoccurunderanagreementtopay the Publisher for CPA) all Publishers that contributed should be awarded aportion of the CPA. The division of that amount of capital (even at time of thisbook’spublication)veryrarelylandsinthelapofmorethanonePublisher,owinglargelytothecomplexpaymenttermsthatwouldneedtobemetbyallparties.Thisincludes the lack of technology to support such payment systems, and a lack ofuniversal industry agreement in any attribution model beyond what is called the“lastclick”attributionmodel.

Figure4.1.7IfattributionmodellingdidnotexistthenonlythelastAdtouch-pointontheuser’sjourneythroughtoconversionisresponsibleforthepurchase/conversion.Here,theusersawthreeadsforthesamecampaignacrossthreedifferentPublishersbeforeclickingthrough to theAdvertiser sitewhere a conversionwas counted. In ad server reporting,under the “last click” attributionmodel, only Publisher 3 deserves the full CPA. As adefaultmostadserverscountconversionsunder the lastclickmodelunlessadjustedbythetrafficker.Usingthemodelcanbeadjustedundertheconversiontagsettingsscreeninthethirdpartyadserverinterface.

AttributionSettings

Attributionsettingspresentthemselvesinthirdpartyadserversasawaytoadjustthe reporting of the total number of conversions as they appears in ad serverreporting when looking at conversions driven by each of the Publishers andchannels on themediaplan.These settings are typically simple adjustments fromthedefaultlastclickmodel,toperhapsamodelthatweightsimpressionsandclicksequallyandwill thereforeaward theconversion to the lastPublisher (even if thatPublisherdidnothavetheuserdirectlyclickonthatad,butrathertheuserjustsawit-countingapost-impressionconversion).

Adjustmentstothesecoreattributionsettingsareusefultorevealthetruenumberoftotal conversions per Publisher and per media channel, when a last impressionmethodologydoesnotseemtoawardenoughconversionstoamorecostlychannel.

Forinstanceconsiderthisscenario:

• The trafficker begins by setting up the ad campaign according to themediaplan.Theconversiontagisgeneratedwitha“lastevent”(clickorimpression)attributionmodelselected.

• During an ad campaign for ‘ski boots’ Paid search was themost expensivechannelusedbytheadvertiser.

• AlongsidethistheAdvertiserhasretargetingtagsontheirsiteandisrunningDCOadsdisplayingcreativechosenbasedonretargeting.TheDCOdisplayinventoryiscostingtheAdvertiserlessmoneytorun.

• Theuserstartsbyactivelysearchingfor‘skiboots’,seesaPaidsearchadandnavigatestotheAdvertisersite(theskibootsproductpage).

• Theuser then leaves the site andconsidersbuying thebootsbutdefersbyadayorso.

• TheuserthencheckstheiremailandtheDCOadisservedtothemfeaturingthe ski boots, the product appears to have followed the user to their emailinbox.

• Rather than click on the DCO ad, the user navigates directly to the site topurchasetheboots.

• The user buys the boots, on the thank-you page the conversion tag fires tosignalaconversion.

• Inad server reporting the totalnumberofconversionson theconversion tagattributedtotheDCOadis1andthetotalnumberofconversionsattributedtothesearchadis0.

Ifthetraffickerhadselecteda“lastclick”attributionmodel,thenPaidSearch,beingthe lastclickbefore theconversionevent,wouldbeawardedwith theconversion.Here,theattributionsettingscanbeusedtobetterrepresentthevaluethatsearchisdriving,awardingthewholeconversiontothePublisherandchannelresponsibleforthelastclick.

If in thesamescenario, theuserhadactuallyclickedtheDCOadaswell,amorecomplexattributionmodelwouldhavebeenrequiredtotelladserverreportingthatPaid Search (regardless if it was the last click or not) is always more valuable,perhapsDCObecausePaidSearchbetterrepresentstheusersintenttopurchase.

EngagementMapping

Someadserverspresentmorecomplexattributionsettingsthanjust“lastclick”or“lastevent”.Likethemoresimplesettings,thesecanbeadjustedinsidethesettingson the conversion tag by the trafficker. Rather than make changes to the“conversions” column in ad server reporting, an additional column is presentedalongsidetheoriginalconversioncolumn.InsidethethirdpartyadserverAtlas,thetrafficker can observe the “e-conversions” column alongside the standardconversions column so that there is a yard stick to understand the context ofmeasuring conversions under new models. E-conversions consider how allPublishers,adsandchannelscansharepartofthewholeconversionsothatthesplitconversion, in its parts, are properly attributed to a combination ofPublisher,AdandChannel.InthiswayPublisherscouldappeartoget0.2ofaconversionintheE-conversionscolumninadserverreporting.Thisisdeterminedbythevalueofthefactorsinvolvedintrueconversionattribution.

AsaresultofthisinformationtheplannerorbuyercanbettercalculatethevalueofthemediaspacebacktoCPAsinceaCPAreflectsthevalueoftheattributingmediaspace that drove the conversion event.Where a Publisher agrees to be paid on aCPA,thecontractualobligationtopayoutonconversionsshouldbetiedbacktoanagreement about the attribution model associated with that conversion. Thisencourages an industry based around a “last clickwins”model, to amodelmorealignedwithhowuserschoosetoengagewithAdvertisingandsubsequentlyspend

withtheAdvertiser.

Figure4.1.8On the user’s pathway to conversion, theremight be an engagement withtouchpoints acrossmultiple channels.Once theuserarriveson theAdvertiser site andtriggersaconversiontag,theconversioneventisfedthroughintoadserverreportingforstandardconversioncalculation.Withinreportingoneconversioniscounted.Inadditionthe conversion might be processed by a layer sitting on top of the reporting interfaceknown as Attribution Modelling. Shown in the diagram the trafficker has set theattributionmodelforthisspecificconversiontagasa‘balancedmodel’meaningthattheattribution is evenly splitacrossallpublishersandchannels. Insidead server reportingthethreetouchpointsreceiveequalweightingande-conversionsareapercentageoftheoverallconversion.Hereshownat0.3conversionsforeachtouchpoint.Underthismodelreportingwould show that theDisplay channel ismore effective at driving conversionsthanSearchsincethetotalforDisplaywillbe0.6whereSearchwillstandat0.3.WithoutattributionmodellingthefinaltouchpointwasDisplaywhichunderthe‘lastclick’modelmeansfullattributiongoestoDisplay.Withthisasareportingmethodology, thecorrectactionwouldbetotakeSearchoffofthemediaplanforfutureactivity.Suchactionmayeasilybedamagingtoacampaignsincesearchmaybeplayinganimportantpartofthispathwaytoconversion,anattributionwhichishiddenwithoutAttributionmodellingbeingenabled.

CustomAttributionModelling

Trying to determinewhich ads are driving actual conversion events is a complextask. On the average media plan, there is a diverse range of Publishers acrossmultiplechannelsshowingdifferentpiecesofcreative.Ifauserencounterseachofthesetypesontheir“pathwaytoconversion”,wheredoestheattributionlie?

Inscienceandmathematics,breakingdownaproblemintoitssimplestpartsmakesit easier to understand. In third party ad serving, particularly inside theAtlas adserver for attributionmodelling these parts are called “Factors”. Possible factorsinclude:

• Thedimensionsofthead

• Theformatofthead

• Howrecentlytheadwasencountered(termed“recency”)

• ThequalityofthePublisherdelivery(seeChapter5-Verification)

• Thequalityofthecreative

• Theviewabilityofthead(seeChapter5-Verification)

• ThePublisherorsitename

An advertiser can rank each of these factors in terms of prioritisation in how todeterminethevaluethateachPublisher,AdandChannelisbringingtoacampaign.This prioritisation produces an algorithm to calculate what Atlas terms “e-conversions”.

These algorithms are heavily scrutinized as being far too subjective, because theAdvertiserhasthefreedomtotweakthealgorithminfavourofcertainPublishersorformats.This canunfairlyweight a campaignbasedonAdvertiser andmarketingbiasratherthanacloserexaminationofthehistoricalcookieleveldata.

Seasonal fluctuations in thedata should alsobe taken into account, and for someattributionserviceproviders,machinelearningisintroducedtomakelightworkofbuildingacomplexanddynamicattributionmodel.Given thecomplexityof suchtasks,variouspoint solutionshavecroppedup toprovidesuchserviceswhicharecommonlyintegratedwiththirdpartyadservingtechnologybyreceivingafeedoftheir data or connecting via an API. Examples include Encore Metrics andClearSaleing.

Many Agencies and Advertisers prefer to make their own minds up with theirPublishersaboutwhatattributionmodellingmeansforthemandtheywilltypicallyuse the data from the Exposure to Conversion (E2C) or Pathway to Conversion(P2C) reports to do so (SeeChapter 5 -Cookie level data). In this reporting, alltouch-points canbe exposed and theAdvertiser can look at thewholedata-set todecidewherethevalueisbeingprovidedonacasebycasebasis.LargeAgenciessometimeshavein-housesystemstocalculationattributionmorefairly.

Challengesinimplementingconversiontags

Advertisers find conversion tags exceedingly useful and become enthusiastic about adserver reportingwhen conversion counting is added to themixof reportingmetrics buttherearetwohurdlestoovercometomakethemostofconversiontags:

1. Conversiontagstakethesitemanagementteamalongtimetoimplementandalongtimetoremove.SometimesitcantakeamarketingteamsixmonthsormoretogetnewtagsplacedonanAdvertisersite.

2. There are so many pages on the Advertiser site: because each conversion tag isdifferent,codingupalargesitecanbeanightmareifconversiontagsandcontainertagsarecreatedonebyone

These two needs alone birthed a new frontier in conversion tag management called“Universaltagging”.

TagManagementSystems

Togetacompleteviewoftheuserjourneyandgetaproperhandleonconversion

attribution, conversion tags should be placed on asmanypages of theAdvertisersiteaspossible.ThismeansalotofworkfortheAdvertiserwebsitemanagerswhokeep dropping these pieces of code, received from the trafficker onto pages.Keeping a log of conversion tags across the site can be a full time job in itself.Althoughcontainertagshavereducedtheworkbymakingindividualpagesalittletidier(keepingallconversiontagsinsideonecontainer),alargeAdvertisersitemayhavethousandsofpages.

TagManagement Systems are single pieces of code that (like site side analyticstools)canbeincludedonallpagesoftheAdvertisersitemakingtheimplementationveryeasytoscale.Thesitemanagementteamcansimplydeploythesinglepieceofcode using the CMS (which will usually grant the ability to update all pages atonce).This single act allows theAdvertiser tomap any existing conversion tags,newconversion tags, retargeting tagsand thirdpartycode to existingURLs frominside the ad server interface. The role falls to the trafficker to generate andimplementtheconversiontagsallfromwithinthesamesystem.

Figure4.1.9UniversalTaggingisenabledthroughaTagManagementsystem(whichcaneithersitwithintheadserverorbesuppliedbyapointsolution).ThetraffickeraccessesthetagmanagementsysteminthefirstinstancetodeploytheuniversaltagcodewhichtheAdvertiser site management team deploy across all pages of the site. Now via the TagManagementsystem,thetraffickercanmapexistingconversiontagsandcontainertagstopagesorURLsontheAdvertisersitewithoutneedingtobotherthesitemanagementteamtomakechanges.

UniversalTaggingSecurity

Universal taggingallowsmarketersa‘backdoor’ intowebpagesontheirownsitewithout needing to wait for the website release cycle to come round.With suchpowercomes the responsibility toensure that any implemented tagsandcodearetestedinfullbeforedeployment.Somesuchsolutionsalsooffertheabilitytoissueaproposal to the sitemanagement team for them tograntpermission to implementtags via the ad server on a tag by tag basis. This is a security mechanism thatpreventsthetraffickerfrominundatingthepagesofthesitewithheavytagsorcodethatisoutsidethepoliciesadoptedbythesitemanagementteam.

TagManagementSystemsaresometimespackagedtoincludefeaturesthatmeasure

thespeedatwhichthepagesload,whichcanbeanimportantfactorindeterminingauser’sinterestintheAdvertiserspagesbeyondsimplyfiringaconversiontag.

TypicallyanAdvertiserorAgencywilldeploysuchasystemlongbeforecampaignsareduetogoliveandthenconversiontagswillmappedtotherelevantURLsclosertotheadcampaignlaunchdateorduringthecampaignlifetime.

Onceagain“pointsolutions”havecroppedupintheindustrytoprovidethisserviceat an additional cost to theAdvertiser, althoughUniversal tagging solutions builtinto the ad server tend to be cheaper but with fewer features. Examples of suchpoint solutions include Tagman, Qubit and the server-side tagging solutionBrightTag,allofwhichcanbesomewhatintegratedintotheavailablethirdpartyadservers using piggybacking. Beyond this Universal tagging is very useful to theworldoffirstpartydatacollectionandtheDMP(seeChapter8).

Conversion tags providemuch needed accountability in digital advertising.Advertisingdollarscanbeattributedtoeventsthatprovideadirectreturnontheinvestmentintheadcampaign. Conversion tags allow marketers to learn more about their customers,transforming the behaviour of users into the definitions of audiences. Audienceidentification allows marketers to correctly target relevant ad campaigns to users andavoidwastingvaluablemediaspendonshowingadstousersthatwillnotbeinterestedinanAdvertiser’sgoodsor services.Conversion tagsdo thisbyallowing themarketers tocorrelateconversiondataacrossallusersthathavebeenengagedinanadcampaignandcheckthattargetaudiencesarethesameusersmakingthepurchases.

Thenextsectionlooksindepthataudiencesandtargetingtotransformthedeliveryofadsfrommassreachtorelevanceinordertodrivevaluableconversions.

Chapter4-Section2Audiencesandtargeting

• AudiencesandTargetingatamedia/inventorylevel

• AudiencesandTargetingatacreativelevel

• thirdpartyadservertargetingrules

• LayeredTargeting

o Figure4.2.1

o Geo-targeting

o Behaviouraltargeting

o Figure4.2.2

o Contextualtargeting(page-scraping)

▪ Figure4.2.3

o Searchkeywordtargeting

o TargetingbasedonVerificationData

o AdditionalDatatargeting

• A/BTesting

• Figure4.2.4

• Trulift

• Figure4.2.5

• AudienceAssignment

o Channelswhichbendtherules

Setting up a campaign for any channelwithin the third party ad server can be a trickyundertaking for a trafficker new to the process, but practice makes perfect. Beyondassigning the ads to placements and sending tags to Publishers, the trafficker has theopportunity to take advantage ofmore advanced settings in the ad serving interface toensurebetterperformance(improvedinteraction,clickandconversionrates)fromadsinacampaign.

Audiences and targeting settings on a placement level, within the third party ad serverinterface is a different, more subtle approach to improving relevance than make suchadaptionsintheplanningandbuyingphaseofthecreationofthemediaplan.Thesetwodistinctionscansometimesgetconfusing:

AudiencesandTargetingatamedia/inventorylevel

PartoftheroleofthemediaplanneristolocateanAdvertiser’stargetaudienceforanadcampaignamong theavailablePublishers andchannels in theonline space.Choosingtherightaudienceandprocuringthatinventoryisthefirststepinensuring

thatadsareseenbytherelevantusers.

WhenPublisherssell‘audiences’toplanners,theyaresellingadspacewhichtheybelieve attracts a certaindemographicof individual.Themethodologybehind thedemographics should be questioned in depth by theAdvertiser before purchasingtheadspace.Tocheck that theaudiencebeingsold ismarketer’s targetaudience,planners can check independent audience verification tools like Neilsen orComscore before the campaign goes live, or verification reporting during thecampaign(seeChapter5).

SometimesPublishersdefinetheaudiencetrafficthemselvesbyallowingasamplesize of logged in users (with completed demographic information on the site)visitingthosepageswheretheadsareshowntobescaledupbasedonthevolumeofimpressions to those pages. Publishers package these audiences for sale in thepremium space and label it up for sale directly to planners and buyers or bypackagingitforsaleonAdNetworksandAdExchanges.InprogrammaticbuyingtheaudiencecanbedistinguishedusingadditionalaudiencedataorwiththehelpofSSP’s(seeChapter8).

Despite the importance, Audience labelling is not industry standardised soparticularlyintheExchangespace,itpaystoaskquestionstobeabletoverifytheaudiencebeingpurchased.Oncethepurchaseoftheinventoryhasgoneaheadandthe labelling of the purchased audience by the Publisher has been accepted, thetraffickercanusetheadserverasasafetynettoensurethatmorecostlyadformatsandcreativemessagingareshowntotherightaudience.

AudiencesandTargetingatacreativelevel

An ‘audience’ in thirdparty ad serving termsmeans agroupofusers labelledbytrafficker as having a common set of properties (such as users in a particulargeography). These groups can be ascertained by either capturing web-page levelinformationinrealtimeorlookingbackatauser’scookiehistoryandIPdata.Thislevelofaudiencerecognitionandtargetingdoesnotusuallyomitthedeliveryofadstocertainusersbut ratherensures relevantandcost-effectivecreative is shown tospecificaudiences.

This sectionwill look in-depthat theabilityof the trafficker toapply rules toadservertargeting,todecidewhichadtodisplaytotheuserwhentheadserverreceivesarequestfromaPublishertodisplayanad.

Thirdpartyadservertargetingrules

Asadefaultsetting,theadserverdoesnotencounteranytargetingrulesonaplacementlevelwhendisplayinganadotherthantoshowthemainadwhere-everpossibleandfallbacktoabackupimageiftheuser’sbrowserdoesnotsupportflash(andfallingbacktoHTML5forthemobilechannel).InChapter3wealsolookedcloselyattheabilityforthetraffickertosetadcontrolstorotateadsorshowntheminasequence.

Targetingrulesaddanewlayertoadecisiontreethatthethirdpartyadserverwillexecuteupon,todecidewhichcreativetodisplaytotheuser.Essentiallyaseriesofquestions;therulescanbelayereduptoformcomplexqueries,butthereisalwaysan‘other’bucketat

thebottomofthedecisiontreewhichcontainsanadthatwouldbeshowniftheuserdidnotfitintoanyofthetargetingruledescriptions.Suchanadwouldbedescribedasanon-targetedad.Asanexample:

• Themarketer purchases ad inventory from a Publisher to target only userswith aSpanishIPaddress.

• The trafficker sets up aplacementwith the targeting criteria to target userswith aspecificadfor thoseusersbasedinSpainandsetsupamoregenericadincaseauserseestheplacementoutsideofSpain.

• ThetraffickertrafficstheplacementtothePublisher.

• The Publisher does not honour the targeting criteria for some of the purchasedinventory because there are not enough Spanish IP addresses viewing the site tofulfil the agreed number of impressions on the IO. Perhaps even if the site is aSpanish site, Spanish speaking users are still expected to arrive to the site fromotherlocationsglobally.

• AuserfromaGermanIPencounterstheplacement.

• TheadserverreceivesarequestfromauserwithaGermanIPfortheplacement.

• Sincetheuserdoesnotmeetthecriteriaforthe‘SpainIPtargetedad’theadserversdelivertheadinthe‘other’bucketsetupbythetrafficker.

• When the trafficker checks ad server reporting, the trafficker will know that thePublisherdidnotfulfiltherequirementontheIOandmayaskforthePublishertoremuneratethecostofthemislaidimpression.

Inthisexampleitisclearthatthethirdpartyadserverprovidesasafetynet.Suchafeaturebecomes more relevant if, for example, there are strict advertising guidelines foradvertisingmedicalproductsinGermanycomparedtothoseinSpain.Thethirdpartyadservercanbeused toensure that theAdvertiserdoesnotbreachsuchguidelineseven ifthePublisherfailstofulfilthetargetingcriteriaontheIO.

LayeredTargeting

Giventherangeofavailabletargetingcriteria,itiscleartodistinguishhowmanytargetingrulescouldbesetupintheadserverbutthereneedstobecorrespondingcreativeforeachdefined audience. Complexity can easily become inefficient, if it takes more work tomanage the decision tree than the revenue that the targeting generates. If there is noobvious uplift in ROI as a result of the targeting rules in place, it is important try tosimplify the decision tree. Branches which attract larger numbers of impressions,conversions, clicks or interactions are worth keeping but those with just one or twoconversions or limited engaging behaviours should be scrapped. Complex targetingrequiresalotofcreativewhichcanbecostlyandevenDCO(seeChapter3).

Figure4.2.1AuserencountersaplacementonaPublisherpage.Theplacementmakesacall to the ad server to decidewhichad to display basedon targeting rules or criteriasetupbythetrafficker.EachrulecorrespondstothedefinitionofanAudienceinsidetheadserver interface. In thisexample, if theuser fits thecriteria forAudience1-Ad1 isshown. If not, the ad server checks to see if the user fits the criteria forAudience 2 inwhichcaseAd2isshown.IftheuserfailstomeetthecriteriaforeitherofthesedefinedAudiencesthentheuserisshownthe‘other’non-targetedad-Ad3.Ifauser’sbrowsercannot supportFlash, then the userwill not enter the decision tree and insteadwill beshownthefall-backbackupimage.

Thethirdpartyadserverinterfaceallowsthetraffickertosetupvariousformsoftargetingdesigned tomatch themarketer and planner’s intended audience criteria. The traffickercannavigatetotheplacementlevelwithinthecampaignsetuptoapplytargetingrulestotheadswhichareassignedtotheplacements.Thesetargetingformsare:

• Geo-targeting

• Behaviouraltargeting(alsocalledretargeting)

• ContextualTargeting(alsocalledSemantictargeting)

• TargetingbasedonVerificationData

• AdditionalDatatargeting

Inadditionsomeadserversofferthechancetoimplementanadditionalfrequencycapattheaudiencelevel.Thiscanactasanoverridetotheruleinplace,bringingtheuserdowntothenextsectionofthedecisiontreepre-maturely.Thisisusefulsothat theAdvertiser can ensure that the targeted audience does not suffer creativefatigue.Thisdiffersfromplacementlevelfrequencycapping(whichshowstheuseradifferentadwhenthecapismet)here,whenthecapismetatanAudiencelevel,theuserdropsdownthedecisiontreeandmightslotintoanotherruletobeshownadifferentplacement.

Geo-targeting

Beforetheadserverdeliversthead,itreceivesthecookieIDandtheIPaddressoftheuser.TheIPaddresscanbeusedtoworkwhatgeographytheusersitswithin,withadecentdegreeofaccuracy.UponreceivingtheIPdata,theadserverrunsthisagainstamappingdatabasebasedonadditionallypurchaseddatafromIPmappingproviders such as Digital Element. This will return a geographical locationprimarily at a country level but then with a greater granularity in somemarketswhen the internet is more heavily used. The US is one of such countries wherebreakdowntoDMAandstatelevelretainsadecentaccuracy.

ThemappingdataacrosstheboardisnotonehundredpercentaccuratebecausenotallIPaddressesarecorrectlymappedtothecountrytheuserissittingwithin.ItissensibletosuggestthatIPmappingiscloserto80%accuratebutthisisstillenoughto perform geo-targeting, especially if this data is pairedwith other data sets fortargeting.BeyondstateorregionlevelsIPdatadoesnotfaresowell(suchasfallingtopostalcodeorzipcodelevel)intermsofaccuracysoitisrecommendedtoavoidthislevelofgeographicalgranularityusinggeo-targetingalone.

To utilise geo-targeting, the trafficker needs to setup placements so that the adsassignedtothemaregeo-targetingbyselectingthespecificcountryorregionfromthe ad server ad-assignment interface. Once the assignment is complete and thecampaign goes live, users in the specified geographies will be shown a targetedpiece of creative, if not then they are shown a non-targeted ad. Geo-targeting isuseful toadd relevance to thecreative, soauser sitting inChicagoseeingaRichMedia ad from a flights provider such as American Airlines may see a datacollectionforminsidetheadwhichcancalculatethecostofatripfromChicagotoother destinations. Geo-targeting would ensure that the creative is used whereChicago isapre-populated field in the form.Thiscreativeoptimizationmaysavetheusertimewhenengagingwiththeadandprovetoincreasethetotalnumberofconversioneventsasusersnavigatetothesitetopurchaseflights.

Although it is possible for the Publisher ad server to perform IP level targeting(thereby geo targeting ads) the control would then sit with the Publisher. ThetraffickerinthethirdpartyadserverinterfacewouldneedtosetuphundredsoftagstoprovidePublisher-independentgranularreportingandwouldnotbeabletostopaPublisherservingahighcostrichmediaunittothewronggeography.

Behaviouraltargeting

Behavioural Targeting or Retargeting is now so extensively utilised in digitaladvertising that it has been given its own Chapter in this book (Chapter 7). Tosummariseverybriefly:usersthathaveperformedaparticularactiononawebsite,such as making a purchase of a particular product, are added to a retargetingaudienceorcookiepool,so thatspecificcreativecanbetargetedto thoseusers inthefuture.ItiswisetoaddretargetingpixelstotheAdvertisersitepro-activelysothatacorrespondingaudienceandrelevantadcanbesetupinsidetheadserverontheplacementlevel.

Figure4.2.2Audienceshaveadsassignedtothem.WhentheadserverreceivesarequestfromthePublisherpagevia theplacement, theaudiencedecision logic in theadserverselectsthecorrectaudience(suchasthecorrectgeography)andservesthecorrespondingad.

Contextualtargeting(page-scraping)

Thepagethattheadisservedontoprovidesgreatcontextualinformation,sincetheuser has navigated to the page for a reason and this is reflected in the contentsurroundingthead.Aplacementcanbesetuptherefore,togatherinformationaboutthepageitselfbeforechoosingwhichadtoshow.Thekeywordscodesorvariablespresentonthepagecanbeusedtoinformthethirdpartyadservertargetingsystem;thetraffickeronlyneedstosetupanaudienceandthereforeanadtomatchwiththecorrespondingkeywordorvariable.

Togoaboutgatheringtheinformationonthepage,theplacementcaneitherscrapethepageautomaticallyorthesitemanagementteamcanwriteascripttoaccompanythe thirdparty ad tag,passingcontextual information into it.The formerof theseoptionsisthekindofcontextualtargetingthatFlashtalkingoffers,whichisapagescrapingmechanismwhichcanliftkeywordsfromthepagetoinformthetargetedad. The latter example is what is referred to in theDG/Mediamind ad server asKeyword targeting; here the Publisher passes a variable across such as‘female54years’ which may not appear in the visible context of the page but isavailableasavariablebecausetheuser issignedinto thePublishersite.KeywordtargetingcanbeusefulfordemographictargetingwheretheaudiencedataisbeinginformedbythePublisher’sowndatabase.

Gatheringsemanticinformationisinvaluablewhenconsideringthedeliveryofadstopages inBelgium.BecauseBelgiumisa tri-lingualsociety(French,GermanorDutch),thesemanticofthepageinformstheplacementwhichlanguagetodisplaythead in. In thisexample the traffickerneeds tohavesetup threeaudienceswithcorrespondingadsineachofthethreelanguagesalongwithanon-targetedad.Thepage can pass a variable or keyword across when the placement requests the adfromtheadservertodisplaytheadwiththecorrespondinglanguage.

Contextual targetingcanalsobeutilised incombinationwithDCOso that theadpassesavalue intoadynamiccreativewhichappears tomirror thecontextof thepageinrealtime.

Figure4.2.3PlacementAisanadsupportsbykeywordtargeting.AscriptisdevelopedbythePublishersitemanagement teamtopassaspecifickeyword into theadrequest.Thekeyword is processed by the ad server decision logic and requests an ad from thecorresponding audience assignment. In placement B, the placement code contains ascrapingmechanismwhichprocessesthesemanticofthepage,passingavaluethroughtotheAudiencedecisionlogic.HerethesameactionistakenaswithPlacementAandtheadrequestservestheadforthecorrespondingaudience.

Searchkeywordtargeting

SearchEngineownersGoogleandMicrosofthavetheadvantageoflookingbeyondtheadservingdatatolookatauser’sfullsearchhistoryeveniftheydidnotclickonthepaidsearchadstrackedbytheadserver.Thissearchhistorycanbetiedtotheuser’slocalcookiedataorsignedinWindowsLiveorGoogleaccountsonthelocalmachine.Theadserversofthesetwocompanies(DoubleClickandAtlas)canthenmatch the ad serving cookie with the activity of the first party network/searchengine cookieoffering awealthofnew targetabledata toAdvertisersusing thesetechnologies. Advertiser can setup their ad campaigns to target users based onspecificsearchtermsaskeywordsthattheyhaveactivelysearchedoninthepast.

Asfarasotheradserversgo;thekeyworddataiscapturedinpaidsearchcampaignswhich are tracked under the search channel by those ad servers and there is anassociation between those keywords and the cookie. In the future it is likely thattheseuserscouldbetargetedbasedonpaidsearchclickactivityalthoughthiswouldprobably constitute a form of retargeting rather than be considered true searchkeywordtargeting(sincetheuserhastoperformtheactionofaclickforthedatatobecollected).

TargetingbasedonVerificationData

Verification data (See Chapter 5) is already being catered for in the world oftargetingonmostlevels:

• GeographicalverificationiscateredforinGeo-targeting

• BrandSafetyandContentCategorisationiscateredforinContextualtargeting

• Device verification is partly catered for using the ad server fall backmechanismforHTML5

Theoutliers forVerificationdata includeverificationmetricsbeyondbrandsafetywhichdeterminethequalityofthecontentonthepage.Suchmetricswouldinclude

thepresenceofa largenumberofads(called“adclutter”)or thepresenceofusergeneratedcontent(UGC)orcommenting.Todayadserversarenotdetectingtheseoccurrences as part of the standard placement delivery (for premium directinventorybuys)but it is likely theywill do so in the future.Within theDSPandExchangespacesthisisadifferentstorycompletely(seeChapter8).

With regards to targeting for viewability data, the placement upon loadingwouldonlybeabletogatherviewabilitydatainrealtime,andassuchthe%surfaceareaseen and the duration of time the user sees the ad forwill be subject to constantchange. Ad server targeting is not yet sophisticated enough to contend withtargetingadsbasedonthisinformation.Insteaditmaybepossiblefortraffickerstousepreviousdatacollectionregardingaverageviewabilitymetricsforadsdeliveredtothesameplacementasabenchmarktodeterminethemostsuitableadtoserve.Once again this level of sophistication has not yet been introduced into targetingbecause (at time of publication) viewability is still in its infancy. Once againtechnology is available in the exchange space to buy inventory and target usersbased on a rudimentary viewability detection method; detecting for ads servedaboveorbelowthefoldoftheuser’sbrowser.

AdditionalDatatargeting

Allofthetargetingmethodslistedsofarinthissectionareactuallyusingadditionaldatatomaketargetingdecisions:

• Searchtargeting-SearchEngineData

• Verificationtargeting-VerificationData

• Contextualtargeting-PageLevelData

• KeywordTargeting-Publisher-suppliedData

• Behaviouraltargeting-RetargetingDataintheadserverretargetingpool

• Geographictargeting-AdditionDatasetfromIPmappingprovider

ThereareasignificantnumberofotherdatapointsthatcouldmeanthatAdvertisersaretargetinguserswithmorerelevantcreativemessagingandmoreimpactfulads.Unfortunatelyasyetmanyofthesecapabilitieshavenotbeenbuiltdirectlyintothead server and require a complex customisation of bolt-on or partner technologiessuch as DMPs or Data Providers (See Chapter 8). These data providers can sithalfwaybetweentheplacementand theadserver tonourish theplacementcall totheadserverfortheadwithmoreinformationabouttheaudience.InthisinstancebothRetargetingandKeywordtargetingcanbeadjusted tobuildoutaudiences inthe ad server from external sources of data. This more complex capability isexploredinmoredetailinChapter8.

Beyondexternalsourcesofdata,theadserveriscollecthighvolumesofdatawhichcanbetransformedintoadditiontargetingoptions.Thesearestandardisedinsomeadserverswhileothersmayneedtosetupcustomscriptstoperformthem:

• Targetingbytheageoftheuser’scookie

• Targetingbythedatethecookiewasdropped

• Targetingusersbasedontheirhistoricalcookieleveldata

• Predictive Behavioural targeting (comparing the actions and engagement ofconvertinguserstonon-convertinguserstomaketargetingdecisions)

• Targetingusersbasedonover-zealousfrequency

Targetingmethods in a decision tree-typeworkflow allows a trafficker to use trial anderror tonarrowdown theaudience left seeing thenon-targetedad.Thismethodology ishighlyusefulfortargetingusersthatperhapsareaccessingtheinternetfromanon-USIPaddressbycustomisingthedecisiontreeandmakingfulluseofthenon-targetedad.Inthisway,traffickerscancalculatewhatimpactcertaincreativemessagingishavingaswellascreatingvarious“experiments”tosplitouttheusersseeingadstodetermineifoneadisbetterthananother.Twoexamplesoftheseare“A/Btesting”and“TruLift”studies.

A/BTesting

A/Btestingmeanssplittingtheusers thatseeaplacementoutso thatdifferentusersseedifferentadsandthentheresultsbetweenthetwogroupsarecompared.Thisenablesthesegmentingofcookiesintorandomtestorcontrolgroupssothatthecontrolgroupalwayssees thesameadwhile the testgroupseesanad thatchanges.Theresulting increase ineffectiveness of the new ads can be attributed to the creative displayed rather thanunaccountabletrends.

A/Btestingcanbeconductedintwoways:

1. The first time the user encounters the placement, they are forced into a group(eitherthecontrolortestgroup).Iftheuserseestheplacementagain,theywillseetheadforthegrouptheywereoriginallyforcedinto.

By creating a control group the trafficker can test various creative formats,creativemessagesandevenpost-impressionwindowsagainstusersthathavebeenservedtheadinthecontrolgroup.Withthismethod,ifanofferisnevershowntoauserinthecontrolgroup,yettheyactivelysearchforitandconvert,thatmediain other channels (such as TV or word of mouth) may be attributed to theconversion.

2. Everytimetheuserencounterstheplacementtheyencounterthelogicfromthetopofthetargetingdecisiontree.

Thiswaythecontrolgroupchangessothatthedivisionbetweenusersisrelevantonly to that session in time. This way if users in the control group had apropensitynot toconvert anyway, thecontrolgroupwouldnotalways stay thesame so thosenon-converters couldbebalancedbetween the control group andthetestgroup.

ThisisnotatrueA/Btestasindividualuserscommonlyvisitbothexperiments.Howeveritcontrolscreativeexposureequallyandmorefairlyataspecificpointintime,toestablishthevalueofcreativemessaging.

Figure4.2.4AnA/Btestallowsatraffickertocontrolthedeliveryofspecificmessagetoasampleofthetotalnumberofusers.WhentheuserfirstencountersthelogictheyentertheA/Btest,90%of theuserswillsee theusualAdB,while10%willsee theexperimentalmessageAdA.Whentheuserencountersanotherplacementfromthesamecampaigntheyareshownthenextusualmessageinthesequence.DidthepresenceofAdAinsteadofAdB improve the campaign performance? A trafficker can check reporting to see. If theexperimentfailedtopointtoaperformanceincreaseorworse;ifAdAhadadetrimentaleffect,atleastonlyasmallsampleofthetotaluserswereeverexposed.

A/B testing helps expose the strength of the creativemessage and its relationshipwithAdvertiser KPIs such as conversion events. A “tru lift” study is the name given to anexperiment usingA/B testing that allows theAdvertiser to demonstrate that any digitaladvertising is having an impact on conversionsversususers that arenot exposed to thecampaignatall.

Trulift

Atruliftstudytries toprovethatanexposuretoanadvertisingcampaigncauses(at themostrudimentarylevel);anincreaseinthenumberofconversionsforanAdvertiseroveragivenperiodoftime.

To conduct a tru lift study the trafficker needs to split the audience of the advertisingcampaignintothosethathaveneverseenanadfortheAdvertiserandthosethathaveeverbeenexposedtoanad.Giventhelimitationoftheimpressionbasedcookiewindowitisactually very difficult to create such a clean break in the audience; some audiencemembers may have seen Advertiser ads from previous campaigns and therefore beinfluencedtopurchasewiththeAdvertiseronthatbasisalone.

In an ideal world if the user has never been exposed to an ad from this campaign orAdvertiser before, the trafficker should set up targeting to create a retargeting pool forusersthathaveseenadspreviouslyandallowthecontrolgrouptoseethenon-targetedad.

Thisnon-targetedadwouldnotbefortheAdvertiserbutratherapublicservicea(suchasananti-smokingad).Iftheadserverobservesthesameruleeverytimetheuserencountersone of the placements for the campaign they will never be shown the targeted ad andtherefore the group remains controlled.The third party ad server allows audience to besplitandweightedsothatthefirsttime,50%oftheusersenterthecontrolgroupand50%are shown the Advertiser’s ad. For a tru lift study to work, the rule which determineswhichadtheusersees,mustpermanentlydropthatuserintoacookiepool.

Oneproblemwith thisapproach is that tobeawater-tightcontrolgroup, theAdvertisermustbegindeliveringretargetingpixels(seeChapter7)withtheiradsfromday1oftheirfirst campaign otherwise some users will be influenced by previous campaigns. Theretargeting cookiewindow also needs to be left open long enough to allow the tru liftstudy tobeundertaken.Theotherproblem is cookie-deletion,which cannotbe avoidedwhereauserwhowasseeingadvertiseradscouldendupnolongerseeingthemastheyareplacedintoacontrolgroup.

Across theboard, tru liftstudiesareapoor indicatorof thepowerofonlineadvertising.Thestudiesundertakenarecommonlyfullofholesthatcannotbepluggedandtheresultstheadvertiserwasexpectingtoseerarelycometofruition.Thismeansthatalotoftimeandmoneycanbewastedundertakingsuchstudieswithlittleornoreturn.

Figure4.2.5Atru-lift study.Theuserarrivesataplacement,neverhavingencounteredthecampaignbefore.Theaudiencecriteriaasksif theuseris intheretargetingpoolforAdAorAdB.IftheuserisinneitheraudiencetypethentheyentertheA/Btest.Heretheadserverrandomlyseparatesouttheusersequallysothat50%ofusersareinthecontrolgroup while 50% of users are in the test group. The corresponding ad includes aretargetingpixelsothattheuserislabelledasbelongingtooneoftheaudiencegroups.Thenexttimetheyencounteraplacementfromthesamecampaign,thelogickicksinandrecognisesthatthecookieisinaretargetingpool.Itshowsthemthecorrespondingadfortheiraudiencetype.NowthecontrolgroupcanalwaysbeshownthesameadorevenanadforapublicservicewhilethetestgroupareshownAdvertiseradsforthepurposesofcreative testing. In reports the performance across the two ads for both clicks andconversionscanshowifthecreativemessageinthetestgroupisshowinganimprovementinperformanceoverusersthatmayhaveonlyseenapublicservicead.

AudienceAssignment

Targeting, experimentation and audience assignment differs between third party adservers.Atitshearttheaudienceisdefinedbythetraffickerandthenthatdefinedaudienceispairedwithaparticularadoradversion.

Aplacement follows the listof audiencesdown the treeof targetingdecisioncriteria todeterminewhichadtoshow.Onewouldthinkthatitmakessensetoalwaystrytotargetanadtoausertoincreasethechancesoftheadgettingnoticedandactionbeingtakenbutalong list of targeting rules for each placement increases thework required to setup andoptimizethecampaign.Highcomplexityinsetup(suchassettingupfiveormoredifferenttargetedaudiencecombinationsforeachplacement)causesthefollowingissues:

1. Muchheavierprocessingrequiredat theadserverwhendelivering thead,whichcouldimpacttheuserexperienceoftheadloadtime

2. Keeping tabs on the effectiveness of various audiences becomes very complexwhenonlyverysmallnumbersofusersconvertoneachtargetedad

3. Committingtosomanyaudiencesandgranulatingthesetupsomuchmeanshavingto createmanyads to support it (althoughDCOwill do thiswell the setup andassignmentcanbelengthyandtheoptimizationwillnotworkiftheaudiencesaresosmallbecauseofthediminishedeffectivesamplesize).

Thesecanbeavoidedbyusing the sameaudiences formultiplecampaigns. Inaddition,the recommendation is to reduce targeting and audience complexity with consistentoptimizationoffofthebackofwellanalysedreportingwiththeultimateaimtouseonlyahandful of verywell targeted and effective audiences. In essence targeting is a scienceexperimentthatifpursuedintelligentlycanyieldgreatresults.

Channelswhichbendtherules

Audience targeting within the ad server will not work for pure Paid SearchCampaigns (since the ad server is not delivering a creative), this is also the casewith Affiliate and Email campaigns. Within the Instream channel, targeting isavailable but since the video is part of a running stream (where a pre-roll isfollowedbyvideocontent),thereisnoneedforanon-targetedad.Inthisinstancetheadserverreceivesarequestfromtheplacementtoserveanaudience-specificad.Ifthecriteriaisnotfulfilled,theadserverreturnsnoinformation,meaningthatanadneverloadsandthereisnonon-targetedadandnofallback.Inthissensetheadserver(atleastforInstream)hasthepowertoaffecttheactualmediadelivery.

Inadditiontothis,Adblocking(afeatureofverification)couldalsobeconsideredafeature of targeting, where, if the page does not meet the targeting criteria forverification, a non-targeted ad is displayed. However, in this instance the non-targetedad isapublic service imageand itsdelivery is specially separatedout inreportingforthepurposesofremuneration(seeChapter5-Verification).

Thisconcludestheworkofthetraffickerinthepre-campaignsetupinterfaceinthethirdpartyadserver.

Thetraffickerwillreturntothisinterfacetosetupsubsequentcampaignsforretargeting(see Chapter 7) as well as adjustments to the existing campaign for the purposes ofoptimization (seeChapter6). The next step for the trafficker is to access the ad serverreporting interface.Thiswill allow the trafficker to checkaddelivery, performance andverification.

SectionCAdServerAnalytics

SectionCAdServerAnalyticsTheprocessofResearchingtherightPublishersforthemediaplan,followedbyplanningandbuyingthemediaforthecampaignsisfollowedbytheprocessoftrafficking,allforthe proof of delivery, verification and performance. Reporting and analytics are theultimate a feedbackmechanism for anAdvertiser, that yields the results and insights toensurebountifuleffectivenessandfuturecampaignoptimization.

Thenextsectionlooksindepthatreporting,analyticsandmetrics.Justknowingwhatdatacanbereturnedfromadservingsystems,canbeenoughtofindthegreatinanycampaignorimproveafailingone.Surprisingly,statisticsarerarelydifficulttodigestinspreadsheetform, where the uniform layout of the data in a standardised to read the data quickly,becomesataskeasilyachieved.

Thestoriesofthedataspringoutofreportspulledintherightwayandattherighttime.The ad serving providers continually improving the look and feel of the interfaces anddataoutputs thatmake it easier todigest that data and find actionable insights.Evenattime of publication, ad server analytics suites are undergoing transformations of thisnature.Thecurrentstateofplaycomesdowntothreecoreareas:adVerification,adDeliveryandadPerformance,thisfollowingchapterwilldivedeeplyintoeachoftheseareasindepthbutfirst it isimportanttounderstandWhatreportingismadeupof,Where the reporting isgenerated,When thedata isprocessedandhow thedataisoutputted.

Chapter5ReportingTechnology

Chapter5ReportingTechnology• Whatisadserverreporting?

• AggregatedData

• Figure5.0.1

• Figure5.0.2

• ReportingAPI

o Figure5.0.3

• WhenistheDataProcessed?

• Realtimereporting

• ProcessingClock

• Whereisthedataprocessed?

• Howisthedatafiltered?

• Whereisthedatadisplayed?

Thissectionisanintroductiontoreportingtechnologywithinthethirdpartyadserver.Itexplainswhatadseverreportingismadeupof.Whenandwherethedataisprocessedandwhattheoutcomeoftheprocessingis.

Whatisadserverreporting?

Thirdpartyadserverreportingisaninterfacethatsitsontopofthedatawhichhasbeencollected during a live campaign. In the core databases of the ad server, the data isvirtuallyincomprehensible,itneedstobecollatedandpresentedinafashionthatiseasytounderstand, theoutcomeare theavailableadserver reports.This sectionwillexplorethe relationshipbetween thecollecteddataand the reporting interface.This relationshipcarriescertainconditionswhichput the reports intoacontextwhichmustbe taken intoaccountwhenanalysingthereportsthemselves.Theseconditionsincludehowthedataisprocessed,whereitisprocessedandwhenitisprocessed.

Conditionsarecommonlypresentedalongside fields insideadserver reportingandgivecluesastowhycertaintrendsmayexistinthedata.

Third party ad server reporting is divided into two distinct types:Aggregated data andCookieLeveldata.

AggregatedData

Aggregateddatadescribesdatathatispresentedtosummarisethecountofcertainmetrics(such as impressions and clicks). Summaries of this data include time periods and adserver hierarchy entities (such as placements and Publishers). These summaries arecommonlyselectableasfilters in theadserverreporting interfacesuch that theycontroltheamountofdatathatwillappearinthegeneratedreport.Thefilteredlistofentitieswillusuallybepresentedinspread-sheetformdownthefarleftcolumntodescribethecontent

ofeachrow.Thetitlesofthecolumnswillbethemetricnamesthemselvesandthecellswillcontainacountoftheoccurrenceofeachmetricagainsteachentity.Thedesignofthespread-sheet is such that a human can analyse the data right across one or morecampaigns.

Figure5.0.1Anaggregatedreportfromathirdpartyadserveraggregatesmetricsacrossanentirecampaign.Thedatawillbeavailableforeachleveloftheadserverhierarchy.To‘pull’areportofthistyperequiredthetraffickertonavigatetotheanalyticsinterfaceof the third party ad server and input several filters such as the Campaign name, thespecifichierarchyentitiesandthedaterange.

Theadservertypicallytakesafewsecondstobringthedatatogetherbeforechurningouta spread sheet or visual representation of the request (referred to as a dashboard).Commonly, the ad server has already processed the data it is returning and the delayaccounts for the front end of the ad server communicating with the backend storagesystemstoretrievethedataintherequestedview.

Cookieleveldatabycomparisonisnotavailableintheadserverinterfacebutisavailableviaafeedmechanism,sendverygranulardataonagiventimeperiodtoalocationoutsideoftheadserver.Thisinformationisusuallytoogranularforahumantoprocessandwouldnormallyfeedintoanexternalreportingsystemordatawarehouse.CookieleveldatawillbecoveredoffinmoredetaillaterinthisChapter.

Figure 5.0.2 Third party ad server reports are available in three main formats.Aggregatedreportsdisplayhighleveldatawhichcanbepivotedacrossvarioushierarchyentitiesandmetrics.ADashboardwillbeamorevisualrepresentationoftheaggregatedreports(useful toliftout trendsandtopresent thedata.Thecookieleveldataisamorerawoutputwhichtendstobeabroadspreadsheet,tableordatabaseinthemostgranularformavailable.Hereeachengagementwithanadorconversiontagisdisplayedrowbyrow.

ReportingAPI

InChapter3-Displayadserving,wemadereferencetothecapabilityofthethirdpartyadserver to supply access to reporting data via an API. Different ad servers are able toprovide different levels of reporting granularity via their reportingAPIs.Almost no ad

servers todaywill allow a developer (with the right level of security access to theAPIfromoutsidetheadserver)toreturncookieleveldata.OnlyaggregateddataissupportedbytheAPI.Inadditiontothisnotallmetricsandadserverentitiescanberetrieved,sotothisextenttheAPIislessdetailedthanthedatathatcanbecollectedfrompullingreportsdirectly fromthe interface.ReportingAPIscanbeofusedbydevelopers inconjunctionwithotherAPIsprovidedbytheadserversuchasaCampaignSetupAPIorPlanningAPI.

AsanexampleanAgencycouldhireadevelopertobuildaplanningtoolforAgencyuseonlywhichsetsupcampaignsandplacementsintheadserverfromadistanceandcolourco-ordinates the buys to correspond with levels of impressions at which ads are beingdelivered.

ThedifferencebetweenthirdpartyadserveranalyticsandSiteSideAnalytics

Chapter2explainedhowanAdvertiserwouldinstallasitesideanalyticstool(suchasGoogleAnalyticsorAdobe’sOmnitureSuite)onto theirwebsite toestablish thesourcesofthetrafficcomingtothesiteandtodeterminehowusersnavigatearoundthe site. Although this data remains very useful, it is still just describing userbehaviouronsiteandcannotexplainhowuserengagewithorseeadsontherestoftheweb.ItistruethatSiteSideAnalyticscantrackreferraltrafficintothesitebutattribution beyond the last engagement is challenging unless the third party adserverisintegratedwiththesitesideanalyticstool.

Figure5.0.3Site-sideanalyticsisusefulforpage-levelmarketingoptimization.Thedataretrieved fromSiteSideanalyticsshowshowusersarrive toanddepart from individualpagesandthejourneysthataretakenwhentheuserisonsite.Bycomparison,thirdpartyadserverswillonlydetectthetraffictoahandfulofkeypageswhereconversiontagsaresitting but reporting will extend out to the plethora of served and tracked ads acrosschannelsandPublisherstorepresenttheeffectivenessofpaid-foradvertising.

Having said this, the technologies cannot be perfectly synchronised because themethodologies are fundamentally different. One example is that the ad serverdefinitionofauniqueuserisdifferentfromsitesideanalytics(whichcanadjustitsmethodologydependingonthereporttimeframe.Unique’spermonthandUnique’speryearisnotcomparablewithwanttheadservertrackswhichisoneuniqueuserper lifetime of the cookieID.) Site side analytics can also not detect for postimpressionconversioneventsanddoesnotofferCookielevelreporting.Therearenumerousotherdifferencesaswellaroundtheclickmethodologiesandthepositionof thecodeon theAdvertiserpage(sitesideanalyticscodeversus thepositionof

theconversiontag).

There is still value to be attained from Site Side analytics and integrations arepossiblethroughtheuseoftokenvaluespassedthroughintheuserclickpathway.However the two systems cannot replace one another, they are doing two verydifferentjobs.

WhenistheDataProcessed?

Thewayanadservercollectsdataandprocessesit,differsaccordingitscorearchitecture.Adserversarestructuredtotrytobringdatacollectionasclosetoreportingaspossibleintime.Whileimpressionsandclickscanbepassedintoreportssoonafterusereventshaveoccurred, it is not as easy for conversion data, where the ad server must compareconversion events across Publishers to establish the attribution. This means that theconversion informationrequiresa longerprocessing time.At timeofwritingconversionmetricsarestillavailableafterawindowofaround6hoursformostadservers(or4timesaday).Thisfigureisdecreasingasadserverprocessingpowerimprovesyearbyyeartotryandreach“realtimereporting”.

Realtimereporting

RealtimedataallowstheAdvertiserorAgencytomakedecisionsimmediatelyafterusersengagewithads.Whileretargetingmakesreal-timetargetingpossiblealready,ifthedataneeded to be processing by human eyes before decisions were made then the processwould slow down. For human eyes to seewhat is going on, reporting needs to presentdata,which in itself is subject to aminor delay.His human touchmight also be calledmanualoptimization(SeeChapter6)sotospeedeverythingup,decisionenginesarebeingcreatedandmanyalreadyexist,toautomaticallyoptimizecampaignsastheyruntoavoidthedelaycausedbyhumaninteraction.Suchsystemseradicatetheneedtopullreportsforthepurposesofoptimization.

Having said this, real time reporting remains in demand for direct responseAdvertiserswhere a trafficker needs to react to adjust the purchase of media to reflect the datadisplayed in ad server delivery. For instance: over the course of a day, if a big eventhappens unexpectedly, the trafficker cannot react to a spike in search traffic for certainterms (so as to paymoreor less for certain keywords) since the spikewould appear inreportingwhichmaynotbeavailableuntilafewhourslater.

If there is ever a fault with the ad serving technology, ad server reporting canusuallyexpecttobedelayedwhilethecollecteddataiscorrectlypresented.InsuchaninstanceAdvertiserteamscanoftenbewaitingtomakechangestoamediaplan;literally prevented from doing their jobs until the conversion data has becomeavailable.

ProcessingClock

Conversionprocessing can take several hours, sooften the ad server has a hard cut offtimefordatacollectiononaparticulardayorprocessingperiod:suchas11.59pm.Thentheconversiondatawillbemadeavailablewithinafewhours,leadingupthethiscut-offpoint. The freshness of the data is therefore impacted not only by location of theprocessingserversbutalsofortheamountoftimeittakestoprocessandreturnthedata

sets.Whenpullingreportsandcomparingdatasetsfordiscrepanciesitisanimportantstepto ask questions about when the cut-off point was for data processing, what time thereporteddatawasavailablefrom.SometimesanadserverreportmaybepulledintheCETtime-zonebutthedataitdisplayswascut-offonanESTtimezone,owingtothelocationthat the data was processed from. This calls into questions where the data is beingprocessedexactly?

Whereisthedataprocessed?

Duetotheglobalnatureof thirdpartyadserving, thebackendsystemshaveaveryrealandcomplexconsideration tomakewith regards towhatarecalled“serverclocks” (i.e.thelocaltimeoftheserverwherethedataisbeingprocessed).Runningalltheconversionprocessing from a single location and then distributed worldwide means the data isavailableatdifferenttimesindifferentcountriesandreallyallcountieswouldlikethedataavailable as early as possible in theworking day, so thatAdvertisers andAgencies areworkingtooptimizeagainstthefreshestdatasets.

Matchingupdatabetweentimezonescanbeabigheadache.Accountsettingswillusuallyallowtheaccounttobesetupfordataprocessingacertaintimezonebutthisisn’talwaysenoughareliableindicatortoexplainwhattimeperiodsthereportistrulypresenting.Itisworkenquiringwith thesupport teamat the thirdpartyadservingprovider toseewhatclocksettingsareinplaceandhowthisimpactsthereportingdata.Insomeinstancesthesequestionscanhelptoremedymanydiscrepanciesbetweenreportswhichappeartobeforthesametimeperiodbutwereprocessedfordifferentcutoffpoints.

Howisthedatafiltered?

Data processing does not just mean reporting, it means cleaning up the data. This canmeanmappingdifferenttablesofdatatogetherintoasinglereport,calculatingmetricsandof course filtering the data. Third party ad servers have their own blacklists for IPaddressesthathavebeencompromisedandhavewaysofsweepingthedatasetsforbotsand data deemed to be created by unauthorised machines creating fraudulent clicks orwastingimpressionsnotseenbyhumaneyeballs.Someadserversallowthisinformationto be presented in reporting. It is also worth asking the ad serving provider if testimpressions,clicksandconversionscanbefilteredfromreporting.Someadservershavethecapability todo thisby filteringout activity fromcertain IPaddresses as abackendservice. It is also not fair to expect an ad serving provider to divulge themethodologybehind the filtering of the data for fraudulent activity since the availability of themethodologymightcompromise the filters.With this inmindadserverswillusuallybeindustryaccreditedformethodologieswhichfilterfraudulenttraffic.

Whereisthedatadisplayed?

Processed data can be accessed via the third party ad server reporting interface.Aggregated reports will exist in templates so can quickly be run (such as a reportcontaining all the standard delivery metrics). The interface will also allow thecustomisationof theaggregatedreports toaddspecificmetricsor filters to thereportsothat it can be generated by a particular time period or for specific ad server hierarchyentities. Run reports will usually output an excel spread-sheet or in a visual interfacewherethetraffickeroranalystcananalysethedata.Aggregatedreportscanalsobeshared

oncecreatedorscheduled(deliveryanupdatedcopyofthereportdirectlyaninboxonagiventimeperiod).

Withanoverviewofaggregatedreportingnowcompleted,theremainderofthisChapterwill detail out the different kinds of reporting availablewithin the aggregated categorybeforecoveringoffCookieLevelreporting.Thesectionswithinthischapterarestructuredin the order that the trafficker or analyst might retrieve the data from the reportinginterface in the ad server once a campaign has been set live. The first stop over isVerificationreporting,thefirstsetofaggregatedreportsthatthetraffickerwillencounterwhenanalysingalivecampaign.

Chapter5-Section1BasicAggregatedReporting• QAReports

• Globalreporting

o Figure5.1.1

• Deliveryreports

• VerificationReports

• PerformanceReports(ConversionandAttributionReports)

• CostandROIreports

• 404errorreporting

• Expiringcostpackages

• ChannelSpecificReports

• TheStandardMetrics

o Impressions

▪ Anewimpression?

▪ Figure5.1.2

o ClicksandClick-throughRate(CTR)

o Conversions&Conversionrate

Thereareaflurryofstandardisedaggregatedreportsthatthetraffickercanpullfromthethird party ad server reporting interface. The following section will detail these reporttypesandthenexaminethemostcommonmetricstobefoundinsaidreports.

Thebasicaggregatedreportingtypesare:

• Globalreports

• Deliveryreports

• Verificationreports

• QAreports

• CostandROIreports

• 404errorreports

• ExpiringCostpackages

• Performancereports(ConversionandAttributionreports)

QAreports

QAreportsarereportsthatcanbepulledbeforeacampaignisputliveorfromtheveryfirstfewminutesofacampaignlifetimetomonitorkeymetricstoensurethat

the implementationof tagshasbeenconductedcorrectly.Somead serversbolsterthisofferingbyconnectingittoamonitoringplatformwithafasterdatarefreshratethan the other data sets. Monitoring platforms such as this can be connected tospecific email alerts that can be setup to warn a trafficker when a campaign isdeliveringorperformingunexpectedly.Thisensuresatraffickercanquicklyreacttoanysuddenturnsindatatrendstoperformtherelevantoptimization(seeChapter6)or test the live tags for errors. QA reports are particularly useful to check thatconversion tags are live and pulling in the correct extended data parameters.WithoutQAreports,thetraffickerwouldhavetowaituntildatahasbeenrefreshedinmorecomplexreportingsuchascookieleveldata,evenincaseswhereextendeddatacollectionisjustbeingtested.

Globalreporting

ThirdpartyadservingtechnologycansupporttheneedsofinternationalAdvertisersandadvertisingcampaignsbyofferingthesingleco-ordinatingteamanaccesslevelinadserver reporting thatgivesaviewacrossallcountries.Within thisstructure,eachlocalcountryteamcanonlyviewtheirin-countryreportingdataandlayerscanbeaddedtosupportregionaldataaccessaswellwhileensuringthatonlythosewithglobalaccess,seeeverything.

It is important to remember that when running global campaigns that it ischallenging tocomparecostdatasinceeachcountrymightpay formedia in localcurrencyandassuchcostsneedtobestandardisedtoasinglecurrencysuchasUSdollarsbeforebeingcomparedwitheachother.

Figure 5.1.1 Within global reports, there are various access levels to the ad serverreportinginterface,grantedtotheanalystortraffickersdependingonwhatlevelofdatatheyrequireaccessto.Centrallymanagedcampaigns(campaignsmanagedforthewholeworld from a single location) benefit from this style of reporting,where the supportingteams in local markets see less information in their reports. Access can be granted atGlobal, Regional and Local levels. Consistent naming conventions are essential toproducing“cleandata”inglobalreporting.AconsistentglobalAdvertiser,campaignandmediaplannamingconvention(1)allowstheglobalAdvertisertocompareperformanceanddeliveryacrosstheregionsatabrandandproductlevel.AconsistentPublishername(2) allows for global benchmarking so that the delivery and performance of the sameformatscanbecomparedshowingifPublishersandPlacementsarejustasstrongregiontoregion.Aconsistentplacement,adandcreativenamingconvention(3) identifieshowthecreativeandexecutioncomparesbetweenlocales.

Deliveryreports

Deliveryreportsaggregatedataaboutthecoreadservingresponsibility:i.e.Servingtheads.Such reports canbeveryuseful toAdvertisers thathavea strictdeliveryplan, tocheck that theyareoncourse todeliver thepaid-for impressionsover theintended time period. Such reports are commonly shared with participatingPublisherswhocanuse thedata toadjust thecontrols inaPublisheradserverforover or under delivery. It is best practice for Advertisers andAgencies to checkdeliveryreportsatleastonceperdayduringacampaign.

Verificationreports

Thetraffickerwouldrunandlookataverificationreportwhenareasonableamountofimpressionshavebeenservedtoproducealargeenoughdatasetforthepurposesofadandcampaignverification.Itisrecommendedthatatleastonedaypassesonlonger termcampaignsbefore checkingverification reporting.Verification reportsareveryextensiveandagrowingareaforthirdpartyadserving,assuchthereisafullsectioncoveringthisareafollowingthissectionwithinChapter5.

Performancereports(ConversionandAttributionreports)

Oncetheadshavebegundeliveringthetraffickercancheckinonhowtheadsareperforming to drive certain marketer KPIs. Performance reporting displays thenumbersofclicksandconversionsbrokendownbyPublisher,Site,PlacementandAd.Foreachoftheseadserverentities,thetraffickercanviewthetotalnumberofclicksonthead,butthenabreakdownordrop-offtoeachoftheconversiontagsontheAdvertisersite.Thisshowswhichcreativeisdrivingthemosteffectiveness.Inaddition, the e-conversions attribution, explained in Chapter 4, can be displayedalongsidethefigurefortotalconversionstoeachentity.Whereitispossibletodoso, the metrics from performance reports should be combined with verificationreporting so that post impression conversions that occurred on impressions thatwerenotinview,areomittedfromtheconversiontally.

CostandROIreports

Byinputtingthemediacostdataintothecampaignmanagementinterface,reportingcanshowthegranulardetailofactualexpenditureovertime,showinghowmuchithas cost the Advertiser to advertise to different Publishers, Sites and Channels.Furthermore this style of aggregated reporting links performance (conversions,clicksandinteractions)backtocosttoshowhowmuchengagement,exposureandconversion iswon from howmuchmedia spend. Some ad serverswill go a stepfurther and display the ad serving costs alongside the media costs so that theAdvertisercanworkoutthetotalcosttoserveanddeliveranad.Whenthemetricsfrom cost and ROI reports are combined with the metrics from verificationreporting, theAdvertiser is informed howmuchmedia spend is beingwasted onineffectiveimpressions.

404errorreports

Thirdpartyadserversthatdonothaveaclose-to-real-timemonitoringandalertingreport functionality,offerweeklyreports thatshowwhichadsarereportingwhichadsareclickingthroughtonon-existentpages.Thisisaveryusefulreporttocheckat thestartofacampaign toensure thatadsareclicking throughcorrectly.Thoseredirectsthatlandusersonnon-existentpagesthroughupanHTTP404errorwhichcanbereportedbytheadserver.

Expiringcostpackages

Someadserversofferareportwhichprovidesanotheralert-stylesystem,showingwhencertaincostpackagesandplacementsarenearing theirenddate.Sometimesad campaigns can last a year or more and although traffickers will insert thecampaignenddateatapointinthefuture,theseenddatescanquicklycreeponanditthecampaignhitstheenddate,startservingfallbackcreativesuchasbackupgifs.

ChannelSpecificreports

Some third party ad servers group reports together (which are really channelspecificreports)intocategorieslikeEngagementreports(forRichMediainteractionreporting). As there are an extensive number of channel specific reports, withmetricsthatneeddetailedexplanation,thereisanadditionsectiondedicatedtothis

topicwithinthischapter.

TheStandardMetrics

Outside the channel-specific metrics almost all aggregated reporting comes back to ahandfulof coremetrics.Thesemetricshavealreadybeen referencednumerous times inthisbook,thefollowingisadeepfocusonexactlywhateachmeans:

Impressions

Most ad server reports, flooded with numbers, represent the increment ofimpressionsforaparticularmetric.Tobefoundinsidethecoreaggregatedreportswithin the ad server reporting interface is a column purely called “Impressions”.Strictlyspeaking,impressionsareacountofthenumberofadrequestscomingfroma user’s browser for a single ad. The request is countedwhen it is received (seeChapter1)inaccordancewiththeIABindustryapprovedstandarddefinitionofan“Impression”.Thisiswhatisreferredtoasthestandardindustry“tradingcurrency”,since the majority of cost models are performed on a CPM (cost per thousandimpressions).AdservinginparticularischargedtotheAdvertiseronthissamecostmodel.

Anewimpression?

Impressions may have had their day. There are now a few potentialcandidatesintheworksthatmightpushimpressionsoffofitsheldspotasthe“tradingcurrency”fordigitaladvertisingbutthereiscontentionoverwhatmighthappen.Thefirstcontender,whichhascroppedupanumberof timesover theshortyears thatdigitaladvertisinghasbeenaround, is“Delivered Impressions”, also referred to as “ad loads” and “ad starts”.Rather than count on the ad request, the count is made when the firstframeoftheadisloaded.Thedelaybetweenthestandardimpressionandthe delivered impression can be significant enough for Advertiser toquestionthestandardmetricimposedbytheIAB.

Thesecondcontenderistheviewableimpression(seethenextsectioninthischapter)whichonlyregistersanimpressionwhentheadisinviewontheusersscreen.Again therecanbea largedifference in thenumberofcounted impressions on the ad server versus the number of viewableimpressions.

Figure5.1.2AstheuserarrivesonthePublisherpage(1)arequestfortheadissenttothe ad server. The counting server registers an impression as it deploys the impressiontrackerfollowedbythead(2).Adsareusuallyloadedinthisway,sothosebelowthefoldoftheuser’sbrowserwillalsoloadupandanimpressionwillbecounted(3).Iftheuserscrolls down, any ads which then are shown above the fold will trigger a “viewableimpression”metric(5).

Some argue that the industry factors in the difference between the cost of animpressionandthecostofadeliveredorviewableimpressionbykeepingalowcostwhereadeliveredimpressionorviewableimpressionmaybeprovidedatahighercharge,becauseofthedeliveryverificationthatsurroundsit.

ClicksandClick-throughRate(CTR)

Clicksarearecordedincrementintheadserver,countingauserclickingonanad.As a component of the click redirect, the ad server is pinged and the masterdatabaseoftheadserverappendsarowstatingthataparticularcookieIDclickedonanadataparticulartimefromaparticularPublishersite.

Formanyyears, the clickwas theonlyway a user could interactwith an ad andshowsomeintent topurchase.On theearlyweb,AdvertisersandPublisherswereobsessedwiththeCPCmodel.Withothermethodsofcountingnowavailablesuchasrichmediainteractionorviewbasedconversions,theclickhaslostalotofvalueinthedisplayspace.Havingsaidthis,aclickissomeindicatorofinterestandmanyAdvertiserstrytoseparateouttheiraudiencesbythosethatshowsomeinterestfromthosethatdon’t.

Clicks remain integral to digital advertising owing to the support of themby theperformance channels (Affiliate and Paid Search).Clickthrough-rate composes ofthenumberofimpressionsdividedbythenumberofclickstodeterminehowmanyviewerssuccessfullybecameclickers.Usuallythiscanbeexpectedtobeaverylownumber, certainlymuch lower than1%of the total impressions.Onabigenoughscale this is still a considerable number of interested users, but it is difficult todeterminewhatcaused theclick itself.Sometimes thecreativemayhavebeen themostenticingpartoftheclickaction(creativeoptimization)andsometimesitmay

just been down to the correctly targeting audience (media optimization). AsexplainedinChapter4,experimentationcanhelpdeterminewhatisdrivingtheusertoclick.

Conversions&Conversionrate

Justaswithclicks,conversionsarepresentedinaggregatedreportingasameasureof campaign success, particularly where the user is driven to make a purchaseonline.Conversions are really; impressions forwhich a conversionwas reported.Conversion reporting will typically displayed so that the total number ofconversionsisattributedtoeachadserverentity(channel,Publisher,site,package,placement,ad,creativeandversion).Conversionstoaparticularconversiontagwillbe labelled in reporting along with a conversion rate (conversions divided byimpressions)tocalculatehowmanyoftheusersthatwereexposedtothead,wentthroughtoconvert.

AsareminderfromChapter4,conversionscanbebrokenoutintopost-impressionand post-click conversion events, allowing the Advertiser to dismiss conversionsbased on impression only exposure to the user. Furthermore it is importantwhenpulling basic aggregated reports (such as performance reports) to remember thatconversionsarecalculatedbasedonthesettingsshownontheconversiontaginthecampaign management interface of the ad server. Both attribution settings andconversion windows will have an impact on the total number of conversionsdisplayedintheconversionscolumn.

With a good grasp of the basic reports, a trafficker can go into the ad server reportinginterface and pull the report directly or setup a schedule so that it delivers to thetrafficker’sinboxfrequently.Thirdpartyadserversdesignthesebasicaggregatedreportsintheinterestoftimesavingsothatthetraffickerdoesnothavetoassembleareportfromscratchbycollectingupandfilteringbyawiderangeofmetricsandhierarchyentitiesthatmay not fit together. An example would be trying to report a channel specific metricacrossacampaignrunninginachannelthatdoesnotsupportthesamemetric.

Further in this Chapter we explore Advanced aggregated reports, which the traffickerwouldattendtoaweekorsointothecampaign,sothatenoughdatahasbeencollectedtobeabletoactiontheresults.

Leavingbehind thecampaignsetupandhaving thePublisher set thead tags livemeansthattheworkofthetraffickernowmovestomonitoringandreportingontheprogressofthecampaign.Thefirststageinthisprocessof involvescheckingthat thecampaignhasactually kicked off (i.e. are the ads delivering, clicking through and are the usersconverting). It is recommended therefore for the trafficker to check basic aggregatedreportingwithintwentyfourhoursofthecampaigngoinglive.Theinitialdataavailableindeliveryreportswouldjustbeanindicatorthatthingskickoffwell.Thetraffickerwouldthen move onto looking at verification reports before taking a deeper dive into theavailabledataviamoreadvancedaggregatedreports,fullycustomisedreportsorchannelspecificreports.

Thenextsectionlooksatthissecondstep:verificationreporting.

Chapter5-Section2AdVerification&Viewability• EffectiveandIneffectiveImpressions

• Reactive,PreventativeandPre-EmptiveVerification

o Figure5.2.1

o PublisherandPlacementqualityindicators

o ContextualVerificationQualityIndicators

▪ ContentClassification

▪ BrandSafety

▪ Figure5.2.2

▪ DemographicorAudienceVerification

▪ Thepanel-basedAudienceVerificationMethod

▪ Figure5.2.3

▪ Thesocialprofile-basedAudienceVerificationMethod

▪ Figure5.2.4

o DiscoverabilityVerificationQualityIndicators

▪ GeoReporting

▪ DeviceVerification

▪ Viewableimpressions

▪ Figure5.2.5

▪ Double Implementations and Fraudulent Servingtechniques

▪ Figure5.2.6

o PageLevelVerificationQualityIndicators

▪ DomainURLReporting

▪ OtherusesforPageLevelinformation

▪ WhiteListsandBlackLists

▪ Figure5.2.7

o Seethrough,ViewthroughandRecordabilityRates

▪ Figure5.2.8

• Preventative-Adblocking

• Pre-emptive-VerificationinProgrammatic

• TheVerificationlandscape

Aftercheckingdeliveryreportsbrieflytoseethatthecampaignstartedwithoutahitch,thenextmostimportantareaforthetraffickertolookintoisVerificationreporting.

EffectiveandIneffectiveImpressions

Verificationisameansofauditingthedeliveryoftheads.ByconsultingVerificationanddelivery reports early and often in campaign management, the wastage of ads can beminimised. Impressions which are not wasted are sometimes referred to as “effectiveimpressions”whilethewastagewouldbe“ineffectiveimpressions”.

Verificationtoolshaveonlyjuststartedtoappearintegratedintothethirdpartyadserver.Upuntil2012theroleoftheVerificationtoolsethasbeenplayedbysmallerindependenttechnology companies but separate tools mean a bigger headache for the setup of thecampaign and a larger room for reporting discrepancies. Ad servers with integratedVerification mean that enabling Verification can be as little work for the trafficker asflickingasingleswitch.

Verification asks a simple question “Is my campaign being delivered to the rightaudience?” Incredibly this is still a conundrum. It’s up to the Publishers to ensure theadvertgetstotherighteyeballsbutbeingathirdparty,theadserverallowstheAdvertiserto verify the delivery independently. Verification tools started life seeing if ads weredeliveredintoa“Brandsafe”environmenti.e.noadultcontent,violenceordisasternewsstoriesandnowoffersomethingwhollymoreattractivetoAdvertisers.

Reactive,PreventativeandPre-EmptiveVerification

Atthetimeofthisbook’spublication,VerificationisoneofthehottesttopicsonthelipsofAdvertisers,AgenciesandRegulators.VerificationcanspelltroubleforPublisherswhodefytargetingandaudiencecriteriaandareouttomakeafastbuckbybendingthetruthabout what they are selling to Advertisers. Verification is available to Advertisers andAgenciesthroughthreemethods:

• Reactive:asindetectingthattheadwasservedintoanunsuitableenvironmentandthenreporting the fact in ad server reporting. Such reporting would be used toleverage conversations with Publishers for better targeting in future orremuneration.

• Preventative:as in preserving the delivery of ads and creativemessages from unsuitablePublisherpagesbyservingapublicserviceadinstead.Thisiscurrentlybeingcalled“AdBlocking”.

• Pre-Emptive:asinverifyingthatanadwillbeservedonasuitablepageandinasuitablespotbefore the inventory is evenpurchased (currentlyonlyavailable in theprogrammaticandRTBspaces-seeChapter8).

Figure5.2.1WhentheuserarrivesonthePublishersite,itispossibletoverifythatboththe page and the user are suited to the ad being displayed. An Advertiser has threeoptions.Either theAdvertiserbuys thead spotprogrammatically (seeChapter8) or bybuyingdirectlyfromthesite.Inthefirstinstanceapre-emptivetoolcanbetriggeredwhentheplacementloads,checkingthepagecontenttoverifyifitisagoodplacetoserveanad.Ifthepagedoesnotappeartobebrandsafe,thepre-emptivetoolwillensurethattheadspotisneverpurchasedinthefirstplace(sendingtheavailableimpressionbackintothe Ad exchange). If the Advertiser does not use a pre-emptive tool then PreventativetechnologieslikeAdBlockingcanbeenabledontheadserver(sothatifthepagedoesnotlookbrand safe, thead serverwill serveaPublicServiceAd (PSA) instead). If neitherpre-emptivenorpreventativemethodsarebeingused,theAdvertisercanbereactiveanddetect the delivery of the ad on an unsafe page. This detection (occurring for bothpreventativeandReactivemethods)feedsthroughintoVerificationreporting.

The remainder of this section will look in-depth at these three classes of verificationbefore looking at how the definition of verification is adapting and changing the wayAdvertiserslookattheirreporting.

PublisherandPlacementQualityIndicators

Reactive verificationworks across a series of quality indicators.When these indicatorsmountupagainstanindividualplacementorPublishertheymightpaintthatplacementinanegativelight(inwhichcasethe traffickershouldconsiderhavingthePublisher targetaway from thepage,domain,geographyordevice inquestion)or inapositive light, inwhichcasemoreimpressionsshouldbepurchasedagainstsuchaplacementinfuture.Inthisway,verificationbegins to eradicatewastage asAdvertisers andAgenciesuse theircollected verification data to map out and benchmark the safest and most valuableplacementsinthePublisherecosystemtomatchtheneedsoftheirbrandsandcampaigns.

Differentbrandsandcampaignswillhavedifferentcreativemessagesanddifferenttargetaudiences.Thismeansthatwhatmightbeconsideredanegativequalityindicatorforonebrandmay not be true for others. For instance, an ad delivered to a news site featuredagainst a disaster news story featuring a plane crash may be dangerous for an airlineAdvertiser but might be seen as good opportunity for a particularly risqué financialAdvertisersellinglifeinsurance.ItisimportantfortheAdvertiserorAgencytomakethisconsiderationwhenanalysingverificationreporting.

Thesequalityindicatorsare:

1. Contextual:whichincludesContentClassification,AudienceverificationandBrandSafety

2. Discoverability: which includes Geographic and Device reporting,ViewabilityandFraudulenttagImplementation

3. Page Level: which includes Domain, subdomain and URL reporting andWhitelistsandBlacklists

ContextualVerificationQualityIndicatorsContentClassification

Verificationtoolswhichspecificallyanalysethecontentofthepagearealsoreferredtoas “CV tools”.CV toolsoperatemost effectively in real time.Thismeans thatwhenthethirdpartyadisservedtoapage;thedomainname,subdomainandURLare passed into the CV tool, where-upon the tool either looks the page up in itsdatabase of ready-scraped pages or it scrapes the page in real time.The scrapingprocess captures information on the page including words, phrases, wordcombinations, multimedia; including video, and digests the relationship betweentheseelements.

Theseelementsarethencategorisedsothatforinstance,apage-levelreferencetoaparticular celebrity mean that the URL of the page is categorised in a celebritycategory. Superior CV tools can analyse the semantics of the page in multiplelanguages and even calculate the cultural significance of certain phrases. Theoutcome is a statistical analysis which places an affinity of the page towards aparticular categorisation. Thismeans that if anAdvertiser paid for ads to appearnext tosportrelatedcontentandtheCVtoolreporteditwasshownagainstmusiccontent, theAdvertiser could take the third party report back to the Publisher inorder to have future ads targeted away from the incorrect pages and onto morerelevant,promisedcontent.Publishershaveaccesstothesamelevelofreportingtoassist theminmanaging theirowninventory toavoidselling inventory that isnotrelevanttoanAdvertiserfromtheoutsetofthecampaign.

CV tools are interesting pieces of kit because they do not utilise user cookieinformationinordertoprovideinformationaboutpageclassifications.Thismeansthat they do not fall subject to any limitations surrounding privacy constraintsaroundcookiebasedtargeting(seeChapter7).

ThetraffickerwouldcheckverificationreportingaroundcontentclassificationandcompareittothegoalsoftheplannerinagreementwiththePublishertoappearonpagesgearedtowardscertaintypesofcontent.Ifthereisamismatch,thetraffickerwouldalerttheplannertospeakwiththePublisheraboutthetargetinganddeliveryconcerns.

Content tends to be classified into a finite number of categorisations which aredefinedbytheCVtool.Until theseareuniversallyacceptedcategorisationsbyallpartiesaroundthetable(Advertisers,AgenciesandPublishers)andwritteninstone

by regulators like the IAB, therewill be disputes about themethodology becausecontentclassification.Tomakenegotiationsmorechallenging,customcontentcategoriesareavailableforAdvertiserstocreateanddefinesothatreportinglaysoutwhereanAdvertiser’sadshave appeared next to content geared towards a certain sentiment or alongsidepraiseaboutacompetitor.InsuchsituationsPublishersmaynothavetheavailabletechnology in their Publisher ad server to avoid delivering ads towards or awayfromsuchpagesinthefuture.

BrandSafety

CV tools are predominantlymarketed as protection against unsafe content ratherthan the interference of a content categorisation. If the ad serves to a pagewithcontentdeemedtobecategorisedasadultinnature,displayviolence,havecontentaboutdeathordisasternews stories, theCV tool canclarify this in reportingandseparateitoutfromstandardcontentclassification.

Figure5.2.2The spider from theCV tool crawls sites across the internet and harvestspage level keyword informationand relates the informationback to thepageURL.Thedata is stored in the CV tool database. When a placement enabled for verificationreportingloadsitcapturesthepageURLandpassesthisbacktotheadserverintheadrequest. The ad server then matches this URL to the CV tool keyword database andpresentsthecontextualinformationinAdserververificationreporting.

SuchoccurrencesareseriousproblemsforAdvertiserssince theycauseunwantedmediaattentionorbadpresswhentheydohappen.AnAdvertisermaynotwantthecarefullypositionedbrandtoappearalongsideunsuitablecontent.Adheringtosuchbrand guidelines, the CV tool allows the planner to gather the data aboutoccurrencestoinformaconversationwiththePublisher.TheplannerwouldinitiallyaskaPublisher to target ads away fromsuchpagewhenprocuring the inventory.TheCVtoolwillshowifthePublisherhashonouredthisagreement.

Whereoccurrencesoffailedbrandsafetydooccur,CVtoolsreporttheoccurrencealongsidethelocationoftheprobleminadserverreporting.

DemographicorAudienceVerification

Publishershavelongbeenplaguedbytheabilitytolabelandselltheiraudiencestoattract Advertising spend. The methods used by the Publishers today are notstandardised by the industry but the more Publishers and Advertisers that use asystemofaudiencemeasurement, themoreaccurateandverifiedtheaudiencecanbe.Therearetwomainstreammethodsforconductingaudienceverificationtoday:

Thepanel-basedAudienceVerificationMethod

ComScore andNeilsenwith their large audience panel data can give anidea(usingsampling)aboutwhataudiencesareconnectedtowhichpages.AsareminderfromChapter2-AgencyPlanning;ComScoreandNeilsenareproviding top listsofhigh traffic sitesbutcanalsobreak thesesitesdownintomatchingdemographicsbasedontheirpanel,aproductNeilsenreferstoasNeilsenOCR.

APublishermaysellemptyadspacesonapagetoanAdvertiserwiththeclaim that it matches the Advertiser’s audience targeting requirements(sayMalesaged18-25).ToverifythattheaudienceisindeedMalesaged18-25,usersontheComScorepanelverifiedasbelongingtothisaudiencetype,wouldneedtobeshownadsfromthetrackedadcampaign.

AComScorepixel,embeddedin thecreativeoraspartof theplacementcodeloadswhentheadloadsandcancrossreferencewiththeComScoredatabase to see if the user belongs to the panel, and if so, whatdemographic they fall into. Here the trafficker needs to ensure theypiggybackaComScorepixelontotheservedplacementcodewithintheadserver.

For the trafficker to obtain such a pixel and to get access toComScorereporting(whichconfirmedthattheadwasdeliveredtotherightaudienceonthepanel)wouldinvolvetheAdvertiserhavingaseparaterelationshipwithComScoreandpayingextrafortheirservices.

Figure 5.2.3 When a user arrives on Publisher page, it is possible to verify thedemographicoftheuserbypiggybackingaComScorepixelontoaservedad.Whentheadloads, the ad detects for an ad server cookie in the usual way but in addition theComScore pixel looks for a ComScore cookie. Relaying this information back to thecomScore servers, the Advertiser can pull a report from ComScore to determine whatdemographic the ad was served to, providing the user made up part of the ComScorepanel.Despitethesamplingofthepanel,acrossthescaleofthewholeadcampaignitispossible to determine an average demographic across the campaign and across eachPublisher.

Thesocialprofile-basedAudienceVerificationMethod

Another methodology utilised by the verification provider Adagooprovides a demographic verification where page content can infer thedemographicoftheaudiencetothatpage.Adagoo‘sspiderscrawlpublicfacing social profiles andmine age, gender and other audience data. Inadditiontothesedatapointsthespiderslookatlinkspostedonthewallsoftheseprofiles,crawltothosepagesandminethekeywordsandcontentavailable there. The content of those pages is then mapped to thesedemographicgroups.

SuchasystemcanthenprocessthecontentofanypagebybeingpassedtheURLofthatpageandreportonthestatisticallikelihoodofthatpagebeing viewed by a certain demographic group based on the alreadygathereddata.Thecatchis that this isacookie-freetechnologyandit isnotactuallyverifying theactualusers thatarriving to thosepagesbut israther reportingon thedemographicsofusers that shouldbe arrivingatthosepagesaccordingtothemethodology.

Figure5.2.4TheAdagoomethod to deduce demographic is a littlemore complex. TheAdagoospider(1)crawlstoapublicsocialnetworkprofilepageandextractstheAgeandGenderinformationthere(2).ThespiderthencrawlstoprofiletoapostedURLandgoestothatURL.Thespiderthangatherskeywordinformationfromthepage(4).TheAgeandGender(2)are inferredto thekeywordsandstoredinademographickeyworddatabase(5).Thishappensacrossmillionsofprofilesandpagestobuildupthedatabase.Whenauser (6) visits any Publisher page, the ad served placement on the page captures thesemantic information on the page before serving an ad (7). The ad server passes thesemantic keywords it finds, related to the ad, placement and site to the demographickeyworddatabase(8).Thedatabasereturnsintoadserverreporting,theinferredAgeandGendertomatchthekeyword.

Inbothsituations,additionalcode(eitherfromAdagooorComScore)needstobeadded to the placements by the trafficker when they are setup. But there is nosolution available today towholly verify audience information at scale.The onlymethods currently available entail a complex process of trying to map deliveryinformationwithextraaudiencedata(availablefromdataproviders)andmatchitatthecookielevelinsideaDMP(seeChapter8).

Advertisers should liaise with Publishers before the campaign begins to setexpectations about the usage ofAudience verification tools. The classification ofcontent and pages must be agreed and accepted by both parties before anynegotiations are made regarding those ads that were effectively wasted. If bothpartiesareinagreementintermsofcontentandaudienceclassification,AdvertisersmayeitherbereimbursedbythePublisherforadwastageorthePublishercanusethesamereportsintheadserverinordertoavoidwastagealtogetherbyidentifyingitearlyonandblacklistingthepages.

DiscoverabilityVerificationQualityIndicatorsGeoReporting

Probably the most well established verification technology is GeographicalReporting(alsocalled“Geo-Performance”).ThirdpartyadserverscandetecttheIPaddress of the cookied user. The IP address can then be run through a mappingdatabasewhichmapsIPaddressestotheirlocation,thesamemethodusedforgeo-targeting.

ThirdpartyadserverswillkeepanuptodatelistofIPmappingsandsoareabletoprovide reports that showwhere the adswere delivered to at the country and insome instances city, DMA and zip/postal code levels. Just aswith geo targeting,regionaccuracydeclinessharplythecloserthatthedatatriestogettopinpointuserexactlocation.

Geo performance reports show the exact numbers of impressions and clicksdelivered to these locations. Those campaigns intended for display purely to anaudience based in the UK for instance and are shown in Geo reporting to bedistributeworldwidecreatemeans for theAdvertiser toapproach thePublisherornetwork to demand the campaign be adjusted to match the right geographicaudience.

Geoperformancecanalsobeusedinconjunctionwithconversiondata,toidentify which regions or countries are responsible for the majority ofconversioneventsacross thecampaign.Thoseregionsfound tobemorelucrativecouldthenbegeotargetedwithaspecificcreativemessage.

Since the Publisher can control the IP delivery inside the Publisher ad server tocorrect the problem of incorrect geographical delivery, third party ad serverreportingisdirectlyactionable.ForthetraffickeritisusuallyjustacaseofenablingthecollectionofIPdataattheAdvertiserlevelwithintheaccountinthethirdpartyad server interface before the campaign goes live or during delivery. To viewreporting, the traffickerneeds toonly lookinto thereporting interface in the thirdpartyadserver.

DeviceVerification

Browsers make publicly available their version number (referred to as a “UserAgent”)whichcanbecapturedbyathirdpartyadserverviaaJavaScriptcall.Theversion number reveals some indication about the device type. For instance thesafaribrowserisonlysupportedonIOS(Apple)devicesandproducts,thereforethepresenceofSafariforanyversionverifiesthattheuserisusinganIOSdevice.

Owing to thevast numberofdevices inuse and the flexibility to install differentbrowser versions to different devices, this verification method is more of anindicatorofdevicetyperatherthanmakeandmodelofthedeviceinquestion.Thethirdpartyadserverneedstohaveageneralmappingdatabasetounderstandwhichversionmapstowhichdeviceinordertodisplaytheresultsinreporting.

Device collection is becomingmore standardised as part of third party ad serververificationreporting,thetraffickercansimplylookatverificationreportinginsidetheadserverinterfacetoseewhatdevicetypesadsweredeliveredto.IfitisfoundthatHTML5ads forexample,arebeingdelivered tounsupporteddevicesornon-targetedusers,thePublishermaynothaveinstalledtheadontoanHTML5specific

entity,domainorapplication.

Viewableimpressions

ThemostcommonlydiscussedverificationmetricamongAgenciesandAdvertisersisViewability.Viewabilitydescribeshowmuchoftheadisinviewfortheenduserand viewability ismeasured on viewable impressions.Viewable impressions at aminimum are impressions served above “the fold” of page as it is viewed in theusers thebrowser.Thedesignofmostwebpagescontainscontentwhich fills theactive browser window and trails off below the bottom fold: hidden from view.Scrolling down the browser using the scroll bar, reveals the hidden content andcontentdisappearsat thetopofthebrowser.Strictlyspeakingeachofthesidesofthebrowserwindoware‘folds’asthecontentcandisappearbehindanyofthemifthepageisdesignedtoscrollinthosedirections.

The viewable impression count is triggered by a JavaScript embedded in the adcode which checks howmuch of the ad is in view and for how long (the sametechnologycanbeusedtorevealiftheadisservedintotheactivewindowortab).SomePublishersoffertheirinventoryclassifiedbywhethertheplacementappearsabove or below the fold, and price the inventory accordingly. Increasingly in thedigitalmarketing space, viewable impression complexity is causing advancementthecapabilitiesoftheverificationmeasurementtechnologies.

Figure5.2.5Viewabilityasaformofverification.WhentheuserloadsthePublishersitesomeadswillbeservedabovethefoldoftheuser’sbrowserwhilesomemayloadbeneath.Theadserverdeliverstheadsandcountsstandardimpressions.Intheadserverinterfacethe traffickerdefines the threshold foraviewable impression (durationofexposureofapercentage surface area of the ad).Under the definition shown,PlacementAwould bereportedasaviewableimpression,aswouldPlacementB,PlacementCwouldnotqualifyas a viewable impression. If the user scrolls further down the page before closing it,PlacementCmayalsocountaviewableimpression.

As the IAB develop a fixed standard for viewable impressions; Advertisers areseeking the ability to have a flexible definitionof “viewable impressions”.Thesedefinitionsconsistofapercentageofthesurfaceoftheadexposedtotheuser,forasetdurationoftime.Userimpressionsthatmatchadefinitionsuchas90%ofanadseenfor5secondsandabove,fallintothiscategorisationofviewableimpressions.Thisiscalleda“viewabilitythreshold”.

Adserverreportingrevealsthetotalnumberofviewableimpressionsagainsteachplacementandwhencombinedwithmediacostdata,canshowtheAdvertiserhow

muchmediaspendtheyarewastingonadsnotinview.SomePublisherswillhavecontractualagreementswithmajorAdvertisersandbigAgencieswhere themediacostforaviewableimpressioninventoryreflectsastandardiseddefinitionbetweenthetwopartiesorthedefinitionsetbytheIAB.ThethirdpartyadserverenablesacustomisedthresholdtobesetbythetraffickerattheAdvertiserorCampaignlevelsandthedataiscollectedandpresentedbasedonthedefinition.

Viewable impressions are under high demand from an industry that believes thatthis newmetricwill challenge theusual definitionof standard impressions in thefuture. Some believe that since ads that are not being seen (according to theviewabledefinition)theyshouldbedis-associatedfrompostimpressionconversionsand subsequent attribution. This presents interesting questions for rich mediacontentwhereaudio is automaticallyenabled;ads that areheardandnot seenarestillagreyareafordigitalmeasurement. Inaddition to theadserversprovidingarecord of viewable impressions, point solutions are also available for measuringviewable impressions such as Alenty; who bring to the table a more “dynamicthreshold”methodology (where the threshold does not need to be set before thecampaign begins, but the trafficker can choose the definition when the report ispulled).

Viewability is harder to measure than content categorisation because it requiresJavaScriptwhichmeansitismorecomplextoexecuteonInstreamandMobileads.ViewabilityisalsomuchharderforthePublishertocontrolandmanyPublishersarewell aware that a fair amount of their inventorymaynever be seenbutwon’t beawareofituntiltheadisserved.SuchPublishersmightarguethattheirpricingfortheinventoryisadjustedtoreflectthepossibilityofwastagefromviewabilitywhichisaconundrumfor theAdvertiserunlessviewabilitycanbeguaranteed in theIO.To assist in eradicating this wastage some technologies are being developed thatonlymakearequesttotheadserveroncetheplacementiscomingintoview,ratherthanloadingtheadtagwhenthewholepageloads.

DoubleImplementationsandFraudulentServingtechniques

Thepublisherisnotalwaysawarethatwastagemaybeoccurringontheirsite,falseimplementations that result in poorly targeted ads (to the wrong geography forinstance)dooccur,especiallywhentheoperationsteamareworkingwiththousandsofadtagseveryday.Pressingthewrongkeycanmeanimplementingthesameadtagtwiceinthesameplaceandevenremovingatinypieceoftheadtagcodewillaffect the delivery of the ad which can result in a double implementation or notrackingatall(allcommonreasonsforPublisherdiscrepancy).

AlthoughthesemistakesareaccidentalsomePublishers(particularlyinthelongtailexchangespace)fraudulentlyadaptadtagimplementationonpurposetheservetheadintheirfavour.ForexampleifaPublisheronlyhadahandfulofpagestheycoulddoubletheiravailableadinventorybydoubleservingads.ThiswaytheinventoryappearstobeservedataratewhichmatchesthePublisherspromise(thattherearetwice as many pages to serve to). Lies and deceit never work in the PublishersfavourandincreasinglytechniquessuchasthesearecapturedbyCVtoolsandthirdpartyadserversresultinginAdvertisersstoppingspendwiththosePublishers.

Othertechniquesincludelayeringadsontopofeachother,hidingadsinspotsonthepagethatarenotseen(butwherecodehasbeenusedtomanipulatethedetectionofaviewableimpression)andloadingadsintolayerswhicharehiddentotheuser.AdvertisersshouldpayparticularattentiontolucrativeadspotssuchasforInstreamads where the opportunity to manipulate the data for the purposes of generatinghigherearningsisamotivationforgreed.

Figure5.2.6FraudulentservingtechniquesseethePublisherimplementingtheadinsucha way so as to perform less in favour of the Advertiser and more in favour of thePublisher. Someadsare served into tiny iframes so that theadcannotbe seenbut stillregistersanimpression(1).Someadsarecleverlylayeredbehindotherads,ensuringtheimpression is countedasa viewable impressioneven though theadcannotbe seen (2).Thesameplacementisloadedupontothesamesiteatthesametimeasacopyofitselfsoastoservemoreimpressionsonfewerpages(3).Instreamvideoplayerssettoauto-play,running below the fold, the video metrics all fire off but the ad is never seen so thePublisherstillgetspaidoncethevideohascompletedplaying(4).

PageLevelVerificationQualityIndicatorsDomainURLReporting

As the ad server is able to capture the URL of the page on ad delivery, thisinformation can be sent back into ad server reporting. By returning a full list oftheseURLsorevenjust thetoprankingsubdomains, theAdvertiserhaspeaceofmindthatthespecificadspotsthatwerepurchasedinagreementwiththePublisheraretheonestheadsareappearingon.Despitethefactthatsomebuysmaybeacross“blindnetworks”(networkswhereinventoryischeaperbecausethebuyerbuystheAudienceanddoesnotknow theendpage theadappearson), ad severs can stillcaptureandreportonthepageURL.

Contractually the Advertiser should ensure (via the signing of the IO when themedia buyer buys the ad space) that there is no room for the Publisher to

intentionally waste the delivery of ad impressions for their own gain. This isparticularly prevalent in the DSP and exchange space where long-tail Publishershave a very disjointed relationship from the Advertiser and “ad brokering” iscommonplace.

OtherusesforPageLevelinformation

PageURLswhichappearinCookieLevelreporting(seeChapter5)allowthe Advertisers more insight into the CookieID itself and therefore theuser.This isbecauseabrowsinghistoryallowsan inference tobemadeabout interest categories for that very specific user (no two browsinghabitsareentirelythesame).Asidefrombeingusefulforverificationthiscan be utilised for targeting and even establishes itself as a uniquepathwaywhichcanbe transformed intopartof aunique identifier for auser.Thismethodology isbeing testedforcookie-free technology in theprivacyspace(seeChapter7).

WhiteListsandBlackLists

AcrosstheinternetWhitelistsandBlacklistsareenforcedtocertainrules.Websitescan end up onGoogle’s blacklist because of SEOviolations and the ruleswhichdetermine Google’s indexing of websites consult the blacklist before displayingresults.Verificationtechnologycanbeviewedinthesameway:thosekeywordsandtermsonapage thatviolatebrandsafetyareconsideredblacklist termsand thosekeywordsthatarewellmatchedandevencontextuallyimprovetheperformanceoftheadonthepagecanbeconsideredwhitelistterms.Advertiserscankeeprecordofthosepages,subdomainsanddomainswhichviolatebrandsafetyandissuetheseto the Publisher in the IO process as a blacklist. A Publisher network may alsocategorisecertainpagesontheirownnetworksasbelongingtoablacklist,suchthatthesepagesareneversoldtoAdvertisersfortheiradspace.

TheURL,domainorsubdomainisessentiallythemarkerbywhichverificationcanbeestablished.Atscale,smartAdvertisersandAgencieswillbegintousethisdatatobuildupapictureofthequalityofthewholeofthepremiuminventoryavailableonthewebandusethisdatatobenchmarksitesandplacementsagainsteachother.Thiswillhelpdeterminebuyingandplanningstrategiesforfuturecampaigns.

Figure5.2.7Placements checking for semantic and content verification can pull in thedomainandURLofthepagetheadisbeingservedon.ThisinformationisfedthroughtoAd server reporting where Advertisers can cross reference the delivery against anyexistingblacklisttheymayhave.

The combination of these quality indicators summarises the most popular methods ofverification,allowingtheAdvertiser tochecktheyaregettingwhat theypaidfor.Butinsome instances the Advertiser wants to go a step further than reporting to actuallyscreenshot the ad failing theverificationquality indicators.Althoughnot a standardisedfeature,somethirdpartyadservershavecrawlerstechnologieswhichcanscreenshotthesepagesandsavethemforthetraffickertoviewinthethirdpartyadserverinterface.Suchtechnologiesdonotalwayscatchtheadintheact(suchaswithgeographyordevice)sointhefuturewemayseeadaptionsthatmovefromascreenshottoashortmovieclipoftheadservedintothePublisherpagefailingVerification.

Seethrough,ViewthroughandRecordabilityRatesDespitetheavailabilityofsucharange of indicators, many of these will fail to report if the ad tag is loaded up in anenvironment which does not support the capturing of this data. Some Publishers(particularlyintheexchangespace-SeeChapter8)willserveadsfromwithinanHTMLiframe.Althoughverificationworksfromwithinaniframeinmostinstances,iframescanbenestedwithinother iframes.Theeffect is that theusersees theadstillbut the tag isactuallysittinginaninvisiblesetofframes.Burstingoutofmultipleframesiscrucialtoreading the viewability of the ad and the context it is placed into. A failing to readviewability is a failure in the “recordability rate” of the ad, while a failure to readcontextualverificationisafailureinthe“seethroughrate”or“viewthroughrate”ofthead.Inadserverreportingthesenumbersarepresentedagainsttheverificationdata.Iftherates are less than one hundred per cent then only a sample of the impressions haveverificationdataagainstthem.Althoughthistransformsreportingfromafull-pictureintoasampling,itisstillpossibletodeterminetheverificationqualityofplacementsandsites.

ThePublisherhasthecontrolinmostinstancestopreventthenestingofiframesandintheinterestofAdvertisertransparencycandotheirbesttoremovesuchnestinginresponsetoanAdvertiserorAgencyquerybasedontheadserverreporting.Publisherswhichretainiframe nesting from the Advertiser perspective look like they have something to hidewithoutsuchtransparencyinplace.

Figure5.2.8Deepnestediframesaffectsboththerecordabilityrateandthesee-throughrateforads.IfanadisservedontothePublisherpagewithintoomanyiframes(producedasaresultofoverlyguardedsecurityorexcesstechnologyonthepage)thecodewithinthe placement cannot look into the rest of the page. This affects verification qualityindicatorssuchascontentverification,domainreportingandviewability.

So far this sectionhasput a lot of focusonReactiveverificationwhich is detecting adverificationaftertheadhasbeenserved.Thedatageneratedinadserverreportingisthenused tohelp thePublisher toeradicate thewastageon futureaddelivery.However, thismethoddoesnotstoptheadfromappearingaltogether.Thisiswhereadblockingplaysarole,preventingtheadfrombeingservedbasedonthequalityindicators.

Preventative-Adblocking

InAdblockingwhen theplacement loadsupon thepage, if the content is found tobeunsafeor ifanyof theotherquality indicatorsarebreached, thenaPublicServicead isdisplayedinstead,usinguptheimpression.Adimpressionswhichareadblockedbythead server enter an ad blocking report which the Advertiser or Agency can view anddownloadinadserverreportingtosharewiththePublisher.ThePublishercantheneitherreimbursetheAdvertiserforthewastedadorcommittoavoiddeliveringtheAdvertisersadstosuchpagesordomainsinthefuture.

Aswith reactiveverification, in preventativeverification, thedomains, subdomains andevenURLscanbeextractedfrommostthirdpartyadserverreportingandgivenorfedtothePublisher.ThePublisheradopsteamcanthenaddthislisttoacampaignorAdvertiserlevelblacklistmanagementfacilityinthePublisheradservertopreventthedeliveryoftheAdvertiser’s third party ad tags to such locationsmoving forwards. This process is notusuallyundertakenbythetraffickerontheAgencyorAdvertisersidebutmaysitwithanadditional service team setup to engage the Publishers in a remuneration discussion:pullingthedataandhavingsuchconversations.UnfortunatelysuchnegotiationscantaketimeandthecampaigninquestionmayendbeforeaPublisher’snewbillingcyclewhichmeansthatrecoveredmediacostscannotbeinjectedbackintothesamecampaign.

ThemosteffectivemethodforeradicatingadwastageindigitaladvertisingistousePre-

emptiveverification.

Pre-emptive-VerificationinProgrammatic

Atthetimethisbookisbeingpublished,Pre-emptiveverificationtechnologiesonlyworkintheDSPspace(seeChapter8)andinvolvesnotbuyingorbiddingoninventorybecausethepagetheimpressiondoesnotmeetthequalityindicatorsforverificationdefinedbytheAdvertiser.Pre-emptiveverificationcollectspagelevelinformationabouttheplacement,siteanddomain,informingthebiddingprocessbeforetheadiseverserved.Thisismuchmoredifficultforviewabilitybutworkswellforalmostallotherqualityindicators.

Before thedevelopmentofbidding technologies, thenetworksandPublisherswereheldon thepromise todeliver ads intoa safe environment andcommit to a legal agreementwiththeAdvertiserthroughcontractsliketheIABsupportediASHinitiative.Asaformalagreement,thosePublishersthatfailedtocomplyandendedupdeliveringtheadsintoanenvironment that fell outside the agreement were told could lead to fines and physicalauditing.IntheUK,iASHhasbeendisbandedinfavourofnewinitiativescurrentlybeingdevelopedbytheIABandsupportingbodies,ofwhichCVtoolsarearecommendationbutasyetarenotaformalisedstandard.

TheVerificationlandscape

There aremore point-solutions available to theAdvertiser for reactive verification thanboth preventative and pre-emptive technologies. Reactive verification point solutionsincludeDouble-Verify and the technology formerly calledAdsafe (nownamed Integral)such tools include an ad blocking feature as well. Just as with survey and researchtechnologies, verification point solutions required an extended campaign setup as anadditional line of code needs to be added to all placements in the third party ad serverpointingatthetechnologyinthepointsolution.IncreasinglyresearchpaneldatasupplierssuchasNeilsen andComScore are investing in verification technologies to improve thevalidityof thedemographicandaudiencedata theysupply.Forpre-emptiveverificationtoolsareavailableseparatefromtheadserverwhichplugdirectlyintothepopularDSPssuchasPeer39.

The line between verification, delivery and performance can be difficult to distinguish,particularlywitharoundthedefinitionofimpressionsandcertaininteractionmetricslikevideo completion rates for Instream (which some ad servers label up as verificationmetrics).

DeliveryandVerificationreportsareusefulinensuringthatthingskickedoffokay.Oncethe data begins to gather in volume, the opportunity arises to conduct amore thoroughanalysis of that data. Beyond the most basic standard reporting templates, there are aseriesofmoreadvancedaggregatedreportsthatallowthemostinsightfulandcost-savingtrends to surface. In thenext sectionweexaminea seriesof theseadvancedaggregatedreports, ensuring that live campaigns are meeting the success criteria defined by themarketers.

Chapter5-Section3AdvancedAggregatedReporting• CrossChannelanalysis

• DayPartAnalysis

• CrossChannelReports

o Figure5.3.1

• AdvancedConversionAnalysis

• TimetoConvert

o Figure5.3.2

• EngagementMappingFactorsreport

• ReachandFrequencyReporting

• UniqueReach

• AverageFrequency

o Figure5.1.3

• EffectiveFrequency

o Figure5.1.4

• ReachandFrequencyReport

o Figure5.1.5

• AdjustedUnique

o Figure5.1.6

• AdvancedSiteandCreativeAnalysis

• UniqueSiteOverlap

o Figure5.1.7

• Adformatreporting

• GlobalBenchmarks

• AdvertiserSpecificBenchmarks

o Figure5.1.8

Adserverreportingisagoldmineofdata.Basedonthefoundationmetricsofimpressionsand conversions, comes awhole series of answers about how to calculate effectivenessand pinpoint which impression and engagement are offering the most value in anAdvertiser’sdigitalcampaign.

ThefollowingsectionlooksatthemoreadvancedreportsavailablefromathirdpartyadserverreportinginterfaceandexplainstheirusestotheAdvertiser,beginningwithreports

related to cross-channel analysis, before moving onto advanced conversion-focussedreporting, followed by reach and frequency reports before concluding with site andcreative-focussedreporting.

CrossChannelAnalysisDayPartAnalysis

Distributionofimpressions,clicksandconversionscanbeviewedonatimelineandused to expose trends in user activity for segmented audiences at different times.Thesereportswillexposespikesinactivitythatmayfallinlinewithinternetusageexpectations (such as increased performance for the commuting city workerdemographic, in both the lunch hour and after the evening rush hour. Day partanalysiscanthereforebeusedbythetraffickertoupdatethetimebasedrotationofcreative messaging for that particular audience (when combined with a form oftargeting such as Publisher-keyword targeting to expose the demographic of theuser). Inaddition,daypartanalysiscanbeused tosee if impressiondeliveryataparticular time is more effective at driving conversions across certain audiencesthan others. Thiswill inform theway the planner and buyer purchases ad space,since some ad spots may be more valuable to accessing their target audience atdifferent times. Since impression volumes will be shown in day-part analysis tofluctuate throughout the day, it might be an idea to use this evidence to justifypayingmoreforatimebaseddelivery,cementedintotheIOfromthePublisherasacommitmenttothatdelivery.

CrossChannelReports

Third party ad serving collects data across each channel for the purposes of de-duplicatingtheusers(uniqueusersseeingadsacrossmanychannelsandPublisherswillproducea completepathway toconversion). Inaddition to thisbenefit is thebenefit of being able to compare the performance of channels as awhole. CrossChannel reporting allows the trafficker to identifywhich channel is addingmorevalue compared to others and if one channel is causing an improvement in theperformanceofotherchannels.

Crosschannel reportingallows theAdvertiser to recognise thevalueofachannelthatmightnotdirectlybecausingconversions,butthepresenceofwhichinauser’spathway to conversion may statistically be causing uplift on the last-clickconversionsofamoredominantchannel.

The more commonly used “search display synergy report” demonstrates thatalthoughpaidsearchappearstodrivethemostconversionsunderthelastadmodel;thedisplaychannelwhenpairedwiththesearchchannel,createsmoreofanupliftinconversionsthanthesearchchanneldoesonitsown.Asmorechannelsgetaddedtothedigitalmarketingmixandgetde-duplicatedbythethirdpartyadserver;moreof this style of reportingwill be in demand, to ensure that there aren’t channelsinadvertently acting as a counter weight to the performance of their siblingchannels.

Figure 5.1.1 An example of how a channel by channel uplift report adds value. If achannel isconsideredon itsown,by just lookingatabasicaggregatedconversions-by-channelreport, itwillappear tobedrivinga limitednumberofconversionson itsown.Here, where the Search channel is the last touch point by users before conversion, theSearch channel is just as effective without a user encountering a Display channel adbeforehand.Therearestill20conversions, sodisplay isnothavinguplifton theSearchchannel. However look at users converting from theDisplay channel as the last touchpointand theanalysisshows that ifusersencountersearchadsbeforeseeingaDisplayad,thereisasignificantuplift.Here, theSearchchanneliscausingathreehundredpercentupliftontheDisplaychannel.

Channelbychannelupliftingreportingallowsthetraffickertoclearlyseehowmuchassistanceonechannelisprovidingtoanotherandtheplannersandbuyerscanusethisdatatoplanwhentorunchannel-specificactivity.Thesamereportcanbeusedtogetmoregranular,calculatingpreciselywhichPublishersandplacementswithinthesesupportingchannelsaredeemedresponsiblefortheup-weight.

AdvancedConversionAnalysisTimetoConvert

Settingthecorrectconversionwindowforaconversiontagisnotaneasytasksinceitispronetofluctuationsintermsofwhattheoptimumconversionwindowsettingsshouldbeinordertothegetmaximumnumberofconversions.ATimetoconvertreport is an advanced aggregated report that plots the conversions on a graphaccordingtotemporalincrementssuchasdaysorhours.Atapoint,foraparticularconversiontagandcorrespondingcampaign,theconversionsdroptoalevelwherethey trail off in a long tail fashion. Chasing this long tail by cranking up theconversionwindowmaymeanleavingthatwindowopentoolongandallowingadsinacampaigntobecomeresponsibleforconversionswheretheconversionwindowlengthexceededareasonableconsiderationperiod.Inparticularthisstyleofreportcanbeusefultoshowwhereadditionalconversionswouldhavebeenreported,hadwindows been increased. Frequently pulling this report keeps the conversionwindowsettingsincheck.

Timetoconvertreporting,shouldalsoconsideradvancedattributionmodelling.Inadditiontoshowingtheaveragetimetakenforausertoconvertviaimpressionorclick based conversion, the report can be adjusted to show how long such aconversion would take under different models. For example, with a customattributionmodelthatawardsagreatershareofconversiontoclicksonrichmediaads,thetimetoconvertreportwouldshowtheamountoftimeittakesforausertoclickonarichmediaadintheconversionpathway,tofinallyconverting.Althoughthere may be other clicks later in the pathway, the report would only take intoaccount, the click-to-rich-media attribution model, to advise the optimumconversionwindowtoexploitthemaximumincrementofconversionsforthistypeofattribution.

Figure5.1.2Atimetoconvertreportshowshowlongittakesuserstoconvertafterbeingexposedtoorclickonanad.Thediagramshowsthatatpoint(1)thetippingpointforthetotalnumberofconversionsoccursatthe4daymark.Thetraffickercouldchoosetosetthe conversionwindowat this point or extend it out further to catchmore conversions.This chart also maps on a consideration for e-conversions, showing how long it takesusers to convert under an alternative attribution model. At point (2) is the conversiontipping point. To satisfy both models on the same conversion tag, the trafficker mightchoosetosettheconversiontagtositbetweenthesetwomarkers.

Thisadvancedaggregatedreport includesdata fromall touchpoints thatoccurredwithin the click or view window for each converting user so as to make aconsideration for the optimum conversion window for e-conversions alongsidestandardconversions.Thereportwillallowthetraffickertoensurethatthewindowsare kept wide or narrow enough to correctly report on conversions under thisalternativeattributionmodel.

EngagementMappingFactorsreport

Basicaggregatedreportingcanbeusedtoshowthetotalnumberofconversionsande-conversions that each placement and site might increment under the selectedattribution model.Atlas specifically offers an advanced aggregated report whichlooksbackatolddataandappliesanewattributionmodel to it, tocalculatewhat

thee-conversionsofplacementsandsiteswouldhavebeen,hadtheybeenrununderthat newmodel. In addition to this, new attributionmodels can be invented andtestedagainstexistingdatatoseewhichmodel,providesthebestresults.Althoughthiskindofreportingmayseemtobemanipulatingtheoutputconversiondata,themodelscreatedmightbeabetterreal-worldrepresentationofhowconversionsareaccuratelyattributed.

Anengagementmappingfactorsreportisusefulindecidingwhatkindofattributionmodeltoapplytofuturecampaigns,providingatrialanderrorreportoutputfromthe thirdpartyadserving interfacewithoutaffecting theoriginallyprocessedcoreconversiondata.

ReachandFrequencyReportingUniqueReach

The concept of ‘uniqueness’ is an important consideration in ad server reporting.Impressions reported in the basic aggregated reporting “impressions” column arereallygross impressions (to steala term from the financialworld)whereas someAdvertisers are interested in actual impressions or to be more specific: UniqueImpressions.Thenumberofuniqueimpressionswillbesignificantlylowerthanthenumberoftotalimpressions,sinceitwillonlycountthenumberofuniqueusersorcookieID’s that are exposed to an ad. In third party ad serving this is called“Reach”;thetotalnumberofuniqueimpressions.Uniquereachcanbecontrolledbythemedia buying process (requesting from the Publisher to only purchase spacebasedonitsuniqueviewing)andthecreativetargetingprocess.CreativetargetingisusedtoensurethateitherthepreviouslyencounteredusersarenotshownanyadbytheAdvertiser, or are shown a different creative from the creative that they havealreadyseen.Thiscaneasilybesetupbythetraffickerusingadserverretargeting(seeChapter7).

Uniquereachreportingcanalsobesetuptoseeifthecampaignisbeingdeliveredtoagrowingnewaudienceoverasetperiodoftime.Thereportcanextendbeyondtheinsight of new unique users being delivered ads to also see if existing users arebecoming new clickers, engagers or converters as a result of adjustments to themediaplantoobtainnewreach.

AverageFrequency

Another of the most common metrics is the “frequency” or the number ofimpressions that sawanadmore thanonce,more than twiceandsoon.Atanadlevelinsidereportingthetraffickercanincludethecolumn“averagefrequency”foranadorcreative.ThiswillshowhowoftenuniquecookieIDsarebeingbombardedwiththesameadorcreative.Thisisanimportantmetricsinceahighfrequencycanhaveanegative impact incampaignperformanceand truly tarnishabrandimage,while a low frequency can cause the campaign to have a lower impact. Thefrequency can be controlled by the frequency cap setting at the ad and audiencelevels (as well as on the Publisher ad server by the Publisher ad ops team) andkeepinganeyeonthismetricavoidscreativefatiguefortheuser.

Figure5.1.3InthisexampleanadisservedbytheadserveracrossmultiplePublishersfor a total of fourteen impressions. By utilise themetrics ofUnique reach and averagefrequencyattheadlevelwithinadserverreporting,thesameadisseenbythreeseparateusers(identifiedintheadserverbyhavingseparatecookieIDs).Eachuserisshownthesameadadifferentnumberof timesbutacrossall uniqueusers, thead is shownatanaveragefrequencyoffourtimes.Asimpleaggregateddeliveryreportwhatnotshowthisinsight,without it, themarketer could infer that theadwas shown to fourteendifferentpeople.

EffectiveFrequency

Effectiveoroptimalfrequencychecksatwhichfrequencylevel thefrequencycapshouldbeset tooptimize thenumberofconversions,engagementsorclicks foraspecificad,placement,siteorcampaign(andsimilartothetimetoconvertreport).Justaswiththetimetoconvertreport,theOptimalFrequencyreportwillpresentanoptimum point followed by a drop off. However it is not advised to set thefrequencycapatthepointofdrop-offbutinsteadtoconsidersettingthecapalittleearlierorlaterdependingonthegoal.Thisconsiderationshouldbemadeatbothacreative and site level to determinewhich sites or creativemight require a lowerfrequency to maintain the target goal. The following goals can be mapped ontooptimal frequency reports so that it is clear what the optimum frequency levelshouldbe:

• Maximumnumberoftotalconversions

• HighestROIoutputorCPA-determinedbytheextendeddatarevenueoutputorthecost input data on the package level (bearing in mind that an Advertiser wouldachieve the lowestCPAon the first impression/frequencybecauseanysubsequentexposureperuserincreasesthecostperacquisition).

• MaximumCTR

• Maximumengagementrateordwell

• Maximumnumberofspecificinteractions

• Optimumupliftinconversionsinaseparatechannel(bycombiningwithchannelby

channelupliftreporting)

• Optimumuniquereach

Figure5.1.4What is themosteffective frequency tohitbothmaximumconversionsandthehighestclick-throughrate?Atafrequencyoffive(1)ahighnumberofconversionsareattained,afterthisthereisadropoff.Butratherthansettheoptimumfrequencyatfive,theclick-throughratedropsoffbefore thatwith theaverageoptimumclick throughrateforthetotalnumberofimpressionsataroundafrequencyofthree(2).Toreachabalancebetweenthetwogoals,thetraffickermayusethisreporttosuggestthePublisherssetthefrequencycapatfour.

This last goal consideration is the most common pairing for optimal frequency.When the two metrics of reach and frequency are combined, an advancedaggregatedreportemerges:theReachandFrequencyreport.

ReachandFrequencyReport

Thetwometricstogetherareinverselyrelatedcontrolsforadspendbecausewithalimitedadvertisingbudgetonlyalimitednumberofimpressionscanbeshown.Thismeanseithershowingthesameadtototalusersjustonceorshowingthesameadmanytimestoasmallergroup.Thisreportisaggregatedtoshowthenumberofnewusersreachedatagivenleveloffrequency.

ReachandFrequencyreportsneedalargerdatasamplethanasingledayinordertoperform the right calculations. Typically reach and frequency reports cannot bepulleddailyfromtheadserverbutrathercanbepulledweekly.Withafullweek’sworthofdataacrossthecampaign,betteroptimizationdecisionscanbemadebythetraffickersinceitwilltypicallytakeseveraldaystoestablishabaselinefromwhichto diminish frequency or improve reach. If new goals are set to increase uniquereach, the report can be used to see if at the same or lower frequencies, thecampaigngoalsarebeingaffected.

Figure5.1.5SinceReachandFrequencyhaveaninverserelationshipthefirstfrequencyoffers the highest unique reach and additional frequency hits the initial diminishinguniqueaudience.Ifthefrequencycapwassetattwo(1),theimpressionsspentonhittingthesameaudienceatgreater frequency,couldbeusedtoestablishgreateruniquereachonthefirstandsecondfrequency.

AdjustedUnique

InChapter1,theproblemofcookieuniquenessandcookiedeletionwasexplored.Sincecookiedeletioncannotbeavoided,theIABhasapprovedtheuseofametriccalled “Adjusted Unique” which counts the number of unique impressions tocalculateuniquereachwhileaccountingforapercentageofusers thatdelete theircookies. In particular, theDG/Mediamind ad server allows analysts to pull thismetricwhere-everuniquereachisused,suchasincombinationwithfrequencyordetermineamorerealisticviewontotaluniqueconversions.

Figure5.1.6Adjustedunique,isametricthattakesintoaccountcookiedeletion.Here,adservingreportinganalysesallcookietofindthosecookieIDsthathavealongershelflife,spanningseveralcampaigns.Thisactivityisthentakenasasampleandtheresultsscaledup,basedonthelimitedpathwaydetailsavailableforsuchcookiesasCookie2showninthisdiagram.Thedata fromCookies1and3 isusedasayardstickwithin theAdjustedUniquealgorithmtopredictresultsbeyondcookiedeletionandmultipledeviceusage.Themetricisadjustedforchangesincookiedeletionactivityovertimeandbyusergeography.

AdvancedSiteandCreativeAnalysisUniqueSiteOverlap

ThestrategyoftheAgencyplannerbuyeristooptimizethedisplayofadstousersinsuchafashionsoastogetthebestresponse.ReachandfrequencyreportinggivestheAdvertiserinsightsintotherelationshipbetweentheoverallcampaignreachandthefrequencyatwhichusersareexposedtoadsandcreative.Thesiteoverlapreportinformstheplannerbuyeraboutwhichaudiencesoverlapforparticularsites.Sincethe aim of buying on multiple Publishers and Networks might be to extend theunique reachof the campaign to encounternewusers,having siteswhichheavilyoverlaponeanotherisusefultoboostfrequencybutcandamagereach.SiteoverlapshowshowmanyusersbelongtomorethanonesiteorareuniquetoasiteandiscalculatedbytheadserverbasedonthemovementofuniquecookieIDscrosssiteandchannel.

Figure5.1.7UniqueSiteOverlapreportstaketheinsightsattainedfromacross-Publisherde-duplicatedthirdpartycookieandallowplannersandbuyerstodetermineif theycanreachthesameeyeballswithfewersites.Thethreeuniqueaudiencesareshown(1,2&3)revealingaclearoverlap.ThereportindicatesthatSiteDcanbeomittedfromtheplan,iftheplanistoreachauniqueaudiencebututilisingtheimpressionsdeliveredfromSiteAandSiteB.

Anadditionalconsiderationiswhetherspendtoreachusersonapreviouscampaignare affecting this one. Although the Advertiser may be advertising a differentproduct, theoldmessagingmaystillhavehad some impact,potentiallyadding touserexposureintheformoffrequency.Inthisway,thespendonthoseplacementsmay be contributing to the performance of the new campaign. Pulling a reportwhichlayersuniquereachandfrequency,crosscampaign,channelandsite,allowsthe planner and buyer to consider reducing spend after a high spend campaign,owingtothemarginalimpactitmayhaveontheupcomingcampaign.

AdFormatreporting

AlthoughitispossibletopullbasicaggregatedreportsfordeliveryonacampaignlevelbyPublisherandtoshowthebestperformingadformat,itismorechallengingtojustseeifoneformatoutperformstheothersineverysituation.Someadserversmake this easier for the analyst in the thirdparty ad server reporting interfacebyaggregating click through rates, effective frequency, effective improvements in

uniquereach,totalimpressiondelivery,conversionratesandeffectiveROIleadbyadformat.SuchreportingisusefultojustifyspendforRichMediaforexampleandismostusefulforCreativeAgenciesandplannersindecidingwhatdimensionsofadstobuyanddevelop.

GlobalBenchmarks

Theurgetolookoverafriendsshoulderinanexamisnotalwaystocheatbutratherto check that your answers look right.Thirdparty ad serverbenchmark reportingtakes campaigns results fromallAdvertisers that are opted in, to aggregate thoseresults.Resultsarebrokenbycountrytomakethemmorerelevanttoanalysts.ThismassaggregationallowsAdvertiserstoseewhichsites,placementsandformatsarethemosteffectivebeforerunningsuchactivitythemselves.Theaggregationisfullyanonymoussothatitisnotpossibletodeterminethesourceofthedatabyanalysingthe results. This would normally be possible if an Advertiser from a particularverticalwas the onlyAdvertiser in that vertical to either be utilising a particularPublisherorbesaturatingdigitaladvertisinginaparticularmarket.

Tomakethedatamorerelevant,adservingproviderscompartmentalisetheresultsbyAdvertisermarketingverticalsothatAdvertisersintheFinancialverticalcanseehowwell300x250expandableadsmayperformasaCTRonspecificsectionofaspecificPublisher site basedon the benchmarking results from the previous year.Conversionratedataisalsoavailable,butsinceconversiontagscanbeusedonallmannerofAdvertiserpages, the ad serverwill group together all conversion tagslabelledas“salestags”todetermineaggregateconversionrates.

TypicallyadservingproviderswillofferbenchmarkdataataveryhighleveltoallAdvertisers and Agencies using its interface while offering deeper insights to adifferent set of benchmark data aggregated by specialist opt-in Advertisers andAgencies,whoonlygetdeeperinsightsbyofferingupmoreoftheirowndatatothepool. These deeper levels of collaboration are tricky to operate, they require ablanketlevelofconsensusamongallparticipantswhichinpracticeiscomplicatedtomaintain.Someparticipants to thedata setmay feel that they are giving awaymorethantheyaregettingbackandconvincingcontributorsthatthereisuniversalfair access is challenging to prove, especially if Advertisers differ in size. Toprovide a fair balance, there is usually a minimum contribution requirement forreporting.Thiscanmeanthatdespitethehugeamountofdataavailableinadserverreporting, the benchmark report may display blank results where only oneAdvertiserinacountryhasanyadservingdatafora300x250expandableadonaspecificPublishersite.

AdvertiserSpecificBenchmarks

LargeAdvertisersandAgenciesalsogeneratesuchbenchmarkreportingoutsideofthe ad server, by feeding their ad serving data into a separate reporting systemwheretheycanlayerincostdataandmorespecificperformancedata.Thecostdatawould also be a record of what was paid previously to determine which sites,placements and formats that are the most cost effective for their expectedperformance. Such benchmarking information would likely feed directly into an

Agency or Advertiser planning tool or Agency trading desk to better inform theplanningandbuyingprocess.

Figure5.1.8Benchmarkingdataisuniversallyavailablefromadservers,fullyanonymousshowing a breakdown of performance benchmarks per vertical and ad format. SpecificAdvertiserscouldalsoopt in toapoolwheremoreof theirdatacontributes toabetterindicationabout performancedatadown toa site orPublisher level.Beyond this largeAdvertiserrunenoughactivityacrossPublisherstoallowtheirplannerstoviewspecificcost and performance data by format and placement. This informs the decision to buythoseadspotsagaininthefuturebuttypicallythisisnotinformationthatissharedwiththePublisher.

This data is not just useful for future planning but also to compare with existingperformance to check that the campaign is on track. Numbers reported to be animprovementonthebenchmarkareapositivestoryfor theAgencyanalystwhocanusethebenchmarktosetexpectationswiththeAdvertiserormarketerforthereportedresults.

Advanced aggregated reporting is a method to pivot and adjust the way that data isviewed,buildingontopofbasicaggregatedreports.Advancedreportsareusefulbecausetheyprovideawaytomakemoresenseofthedatawithoutdivingdeeplyintothecookielevel reporting, also available from the ad server.The next section looks atmore nichechannel-specificreporting,whichisusefulinsilo,allowingmarketersheavilyfocussedonSearch or onVideo, to just look at the delivery, engagement and performance of thosechannelsinisolationratherthanlookingatthebiggerpicture.

Chapter5-Section4ChannelSpecificReports• DisplayChannelReporting

• RichMediaChannelReporting

• DCOChannelReporting

o Multivarianttesting

o DangersinVersionOverwriting

o Versionreporting

▪ Figure5.4.1

• InstreamChannelReportingandMetrics

o VideoStartedanduniquevideostarts

o Adduration

o Incrementadplay-through(25,50,75,100%)

o Average%videocompletion

o VPAIDcustominteractions

• MobileChannelReporting

o CookieDuplication

o CookieAbsence

o DeviceFingerPrinting

o Cookiesinapplications

o DeviceandCarrierreporting

• PaidSearchChannelReporting

o SearchOverview

o KeywordPathwayAnalysis

• NaturalSearchReporting

o DoublecountingandSingleCounting

▪ Figure5.4.2

• AffiliateChannelReporting

The third party ad server reporting interface offers traffickers and analysts reports thatcontainmetrics relevantonly to specificchannels.Thesechannel specific reportsmatchwell to the same major channels described in Chapter 3 - Campaign Management byChannel. The following sectionwill outline the proprietarymetrics,methodologies andanalyticalconsiderationswithinchannel-specificreporting.

DisplayChannelReporting

Reportsspecifictothedisplaychannel(omittingrichmedia,DCOandvideofromthedefinitionofthedisplaychannel)arethemostbasickindofreportingavailableand would be the same basic aggregated reports described as delivery andperformance reports.Here reports predominantly relate to impression informationand click information for the Publisher sites, packages, placements and ads. Theadditional considerations,moreunique to thedisplaychannel are thosewhere thenaming convention can be adapted to give more information about the creative.These fields in reporting would be “creative name”, “ad name” and ad andplacementclassifications.Creativenameandadname,asdescribedinChapter2 -should match a description of the imagery used in the creative or relate to theproductadvertisedwithinthead.Withinreportingthesenamescanbeaggregatedsothatiftheyareused.Theadandplacementclassificationfieldsarealsousefulforthispurposebutservetoprovideadditionalmetainformation.Adclassificationonemight be populated by the trafficker to contain the date that the creative wasswappedoutforanotherwiththesamename.Rowswithinreportingcouldthenbeaggregatedtoshowthedifferentinperformancebasedonthisinformation.

Also relevant toDisplay reporting is the ability for the ad server to reporting onbackupimagedelivery.Sometimesabasicaggregateddeliveryreportwillhidethenumber of impressions where users saw a backup image instead of the flashcreative,channelspecificreportingshouldprovidetheanalystwiththeabilitytoseethisinformationaswell.AnindicationtothedeliveryofalargenumberofbackupimagesmaymeanthatunsupporteddevicesarebeingtargetedatthePublisherlevel;a control which can be adjusted by having the trafficker feed this reportinginformationtothePublishertotargetawayfromnon-flashsupportedresultsbeforeahighdeliveryofbackupimagesequatestowastage.

RichMediaChannelReporting

TheprimarydifferencefromareportingperspectivebetweenStandardDisplayandRich Media, is Rich Media’s ability to track interactions and subsequently passthembackintoreporting.Interactionscanbedividedinto twogroups;defaultandcustom interactions. Default interactions apply to the counting of the trigger ofcertain Rich Media specific functions such as Expansions. Total number ofexpansions, for instance is a standardised metric in RichMedia reports. CustominteractiontrackingforRichMediaallowsthecreativedevelopertoplaceatriggerinside the creative anywhere, so that an actionwithin the creativewill count thatcustom interaction occurrence. Creative developers are encouraged to work withmarketersandAgencyteamstoputusefulcustominteractionswithinthecreative,trackinganythingfromauserhoveringoveranobjecttoclicking,pressingabuttonorevensittingidle.

In reporting, interactions can be tied back to unique impressions and conversionevents.Theoverall time spent interactingwithin the ad (such as amousepointerhovering over the creative for an extended period of time) is reported inmetricssuch as “interaction time” or “dwell”. The measure of such metrics is protectedfromaccidental triggerbyputtingsuchsafeguardsinaswaitingforonesecondto

passwiththemousepointeroperatingwithintheconfinesoftheadbeforecountingbegins.ThekeyhereisthatRichMediacreativecanalsosupportatimersuchthatananalystcanpullfromreportinga“totalbrandexposureduration”metricwhichisatotalcountofinteractionacrossallusers.Suchmetricscouldbeusedtocomparewith TV ad campaigns that are measured by minutes of exposure to the totalaudiencebase.

Someadserversofferaheatmappingtool,whichlayershotspotsoverthecreativetoshowanalystswhereuserswerepredominantlyengagingorinteractingwiththead.Althoughsuchreportingisaestheticallyrewardingtheanalyst,theinterpretationof such reporting can be subjective and difficult to optimize from creativelyspeaking.

DCOChannelandAudienceReporting

DCOreportingcanbeviewedatahighlevel;suchastoviewtheeffectivenessofversionsbasedon theversionname,butdeeper insightsareavailablebyensuringsuchreportsdisplaythedetailofthedynamicelementsofthatversionsothatresultscan be tied back to the correct recipe of those elements. Multi-variant testingspecificallyallowsfortheanalysisandidentificationofbestperformingversionsorvariantsbasedonthosebestperformingelements.

Multivarianttesting

Withthedynamicelementsoftheadexposedinthecreativethereportingmay expose trends such as that a particular price point across lots ofdifferenttypesofcreativealwaysseemstopullahighnumberofclicksorconversions. The same could be said of a background image, colour orfont type used although clearly some co-incidental trends may surfacewhichfall to thebest judgementof thecampaignmanager todismissorde-prioritise in favour of optimizing to elements which logically makeattract a better performance (such as price point or offer type). Inreportingformultivarianttesting,theusualcolumnfortheversionnameis replaced with the name of the varying asset and reporting such asimpressions,clicks,conversionsandinteractionscanbelayeredupbysiteand placement so as to find out if a combination of assets equates to abetter performing version when they are combined together. Suchreportingisofparticularinteresttothecreativedeveloper.

DangersinVersionOverwriting

Updating a price point or any other dynamic element for a specificversioninthesetupofaDCOcampaignisabadpracticeintheadserverwithoutchangingthenameoftheversionentirely.Someadserversdonotplace restrictionsonversionupdating simplybecause restrictionswouldadd extra steps for Advertisers and make the setup more challenging.However;theamendmentofadynamicelementinanexistingversionisnota recordedevent so insideDCOreporting,a topperformingversionwhichwaschangedatsomepointinthedaycouldnotattributeitssuccesstotheamendment.

Even at the cookie data level theDCO version amendment is not timestamped so the traffickerwould need tomake a full record of changesmadetoexistingversions.Thesolutionistoalwayscreatenewversionswhen any dynamic element is updated and ensure there is somerelationship between the newversion and the old version to understandthe version evolution (this can be done with a good version namingconvention).

VersionReporting

DCO reportingwithin the ad server reporting interface costs of reportsthatshowsversiondelivery,performanceandengagementat theversionlevel insteadof at the ad level.Detailed reportingof this typewill alsoshowwhichaudienceorplacementtheversionwasassignedto,toallowthe trafficker to optimize the delivery, targeting and rotation of bestperforming versions. Version level reporting can become complex toanalyse if only the version number was displayed for each row in thereportsincethenumberorIDdoesnotgiveawayanaccuratedescriptionofhowversionsdifferfromoneanother.

For this reason, larger version reports that include a list of the content of thedynamic assets is useful although the reporting quickly begins to replicate thespread-sheetbasedmassversioninterfacewithinadservercampaignmanagement.Toreducereportsize,someadserverreports forDCOwillomit rowsorversionswhich recorded no delivery or poor performance, reducing a potential pool ofthousandsofversionsatabyday,weekormonthgranularity,down to justa fewrows.Thismakeslifemucheasierfortheanalystandtrafficker.

Figure5.4.1Versionlevelreporting(shownleftinthisdiagram)revealsthedeliveryandperformanceofthevariousversions.Eachoftheversionsisseentoberunningagainstadifferentdefinedaudiencetype,herebygeographyandclickingthroughtoadifferentpartof the MSN. com website. Beyond this the differences between the versions are notrevealed in the report.Navigating the ad server campaign setup interface revealsmoreinformation about each version by how they vary from one another. The analystwouldneed both these data sets to adequately analyse each version and explain whether theaspectsofeachversionorthetargetingareresponsiblefortheperformance.

InstreamChannelReportingandMetrics

ThemetricsthatcanbetrackedintheInstreamchannelhavegrowingsynergywithTVmeasurement.HistoricallybigAdvertisers,advertising through themediumoftelevision have been able to track a sample of ad delivery and exposure. ThisallowedAdvertiserstopayforTVadsinnewways:suchasonlypayingifthevideooftheadwasplayedinfull.ApartfromtheDisplaychannelandSearchchannel,theInstreamchannelistheonlyotherchannelwhereAdvertisersarepayingPublishersnot on ad delivery, but on channel-specific metrics: video plays and vide

completions.The following is a list of notablemetrics available only in Instreamspecificreports.

VideoStartedanduniquevideostarts

Atthecreativelevel,theanalystcandeterminehowmanytimesthevideowas watched and howmany unique users watched in (some users willplayavideothroughmorethanonce).Videostartedisusefultocomparewith thenumberof impressionsfor thesamead.Since thevideostartedplayingaftertheimpressiondelivery,someAdvertisermayprefertopayPublishersbasedonthevideostartedmetric.

Adduration

Adserverscanprovidemetadataabout thevideocreative itselfsuchasthequalityofthevideostream(inbitsperminute)andthetotallengthofthe video. This information can help analysts perform side by sidecomparisons between different video ads which may be of differinglengthsandquality.

Incrementadplay-through(25,50,75,100%)

The amount of time that the user views the video ad for, is anotherInstreammetric(butonethatisalsorecommendedtotrackinrichmediavideounits).Thesemetricswillshowuserdropoffasthevideonearsitscompleted play. A video which is too long to hold user interest wouldshowhighdropoffofthetotalnumberofuniqueimpressionsastimegoesby.Asanactionthetraffickerwouldrequestashortervideolengthfromthecreativedeveloperandtobuyshortervideoadspacefromtheplannerandbuyer.

Average%videocompletion

Attheadleveltheanalystcanseewhichpercentageofuserswatchedthevideotocompletion.Thiscanactasamarkerfordelivery,performanceorverificationdependingonhowtheanalystinterpretsthedata.

VPAIDcustominteractions

Just as with Rich Media, Instream video supports custom interactions.These are inputted by the creative developer before the campaignlaunches andcanbepositionedanywhere in thevideocreative.Buttonswhichallow theuser to share thevideoona socialnetworkoropenanInstream video gallery from within the VPAID creative can have suchactionstrackedandincrementedwithinInstreamspecificreporting.

MobileChannelReporting

Mobile specific channel reporting, although very similar to standard Displaydeliveryreporting,hassomeconsiderationstomakewhichrelatetheidentificationofauniqueuser.InstandardDisplayreporting,knowingacookieIDisuniqueornotuniqueallowstheuserpathwaytoconversiontobemappedcrossPublisherforthesamedevice.Oncea seconddeviceoramobileapplication is introduced into the

mix, the cookie dependent technology related to establishing unique users andconversions,islost.

CookieDuplication

The human race now accesses the internet more frequently through amobile device than a PC but the environments from a browsingexperience across devices is unbalanced. In thePCenvironment, the adserverdropsacookie,labellingtheuserandtrackingthem.SolongastheuseronlyusesthatsamePCandbrowsertoaccesstheweb,theAdvertiseronlyneedstoworryaboutcookiedeletion,butMobilepresentsaproblemofcookieduplication.

Asingleuserthatswitchesbetweendevicesmeansthattheadserverseesmultipleusersinitsdatacollectionatthecookielevel,thanthetruesingleuser. If there aremultiple cookies in existence at the same time for thesameusertheseneedtobetiedtogethertoestablishthetruereachofthecampaignandforpathwaytoconversiondata(seethefinalsectioninthischapter).Toovercomethisproblemasingletechnologyneedstobeusedonbothdevicessothat inreportingthere isacommonlinkbetweenthetwocookies.Anadditionaldatasetcanbeapplied to themobilecookielevelreportingtoperformwhatisknownas“CookieMatching”tofilterthereporttoestablishtheuniqueusers.

Such datasets are available forAdvertisers to buy; themarketerswouldthen pull this data into the ad server cookie level data inside a DMPenvironmenttocookiematch.Datasetsthatcouldbeusedareoneswheretheuserhasloggedintothesameplaceonseveraldevices(anexampleofwhich would be signing into a browser as is possible with Google’sChromeorloggingintoacloud-basedaccountsuchasanemailprovider).Thisisallwellandgoodwhiletherearecookiestomatch;butthestingeristhatthirdpartycookiesarenotubiquitousacrossmobiledevices.

CookieAbsence

Applemade adecision tohave theSafari browser not allow third partycookiesasadefaultsetting.ThishasdamningconsequencesfortrackingactivityonAppledevices.Ifacookiecannotbedroppedthereisnowaytoknowingifanindividualbrowsinguserhasseenanadandiftheyhaveseen it more than once. Inside the third party ad server database, thecookieissimplygivennoID.Asapplegrowsinadoption,sothenumberoftraceableusersdecreasesusingthecookielevelmethod.

Totrytogetclever,technologyprovidersrealisedthattheydidnotneedacookiepersay,butratherjustneededacommonidentifierforthedevice.The identifier would need to be unique so that the user on the devicecould be detected as a single individual and bemeasured as a separate“cookie”fromtheothersbeingtrackedbythetechnology.Thisidentifierisreferredtoas“DeviceFingerPrinting”.

DeviceFingerPrinting

At the core of the Apple device sat the “UDID”. A Unique DeviceIdentification label which technically speaking could be accessed bybrowser based technology and then amatching “cookie” could be builtremotelyintheserverenvironment.Theeffectisthattheadserverwould“see”asingleuser.ButtheUDIDwasnottobethesaviourforthecookieproblemthatmanyhadhoped.Applescrubbedthelabelfromalldevices,well aware of the workaround, and at time of Publication had begunpresentingaccesstoanewidentifiertoaddevelopers,specificallyforthepurposesofAdvertising.

Thepurpose-builtidentifierhasthepossibilityofusercontrolsothattheycanoptoutofcookie-basedtargetingbyswitchingitoff.ThisexposedthecontrolApplehasoverthirdpartyaddeliveryandtracking,byblockingorcontrolling access to a unique identifier it is not possible to make anindependentauditof the truenumberofuniquecookiesonsuchdeviceswhich is a problem ifApple decides to enter the world of AdvertisingTechnologyandsupplyoftheirownmediaspace.

Workaroundsdoexist to tryandcircumvent theproblemsof identifyingandcountinguniquemobileuserssuchascreatingaunique“sessionID”on the fly,orusing the IPaddress toestablisha singlecookiebut thesetechniques remain flawed and the problem of a lack of common cross-devicecookieremainsinplace.

Cookiesinapplications

Cookiesareenabledthroughbrowserusageandareuniquetobrowserssoin-appadvertisingpresentsachallengefordatacollectionsincethereisnostandardised in-app cookie to speak of. The apple device identifier foradvertising could be used in this instance (captured by the app using aspecificlineofcodeandthenpassinguptotheadserverwhentheadisdelivered)butthisstillmeansthereisaneedforcookiematching.

SomeSDKsforcethemobiledevicetomomentarilyloadupthebrowseronfirstuseoftheappsothat theadservercanlookintothebrowsertodetectthebrowsercookieIDbeforesendingtheinformationontotheadserver. From that point forward the device can be associated with atraditionalthirdpartycookie.Thisisnotatechniquerecommendedtothedevelopment community by the leading app stores but again anotherworkaround that helps demonstrate the jerry-rigging required to datacollectinayet-to-bestandardisedchannel.

With theknowledge that cookies remainelusive in theMobile channel, ad serverreportingforMobilerestrictstheuniversalmetricsavailablefortheanalysttolookat.Impressions,andclicksareavailableataPublisher,Site,Package,PlacementandAd level but reach, frequency, cross-channel and conversion reporting are as yetunavailable or unreliable across the board. Rich Media ads within mobile willsupport interaction reporting for HTML5 ads and flashwhere it is supported. IPdelivery can also be complex due to the portable nature of the device, so

geographical reporting is ill advised. In addition verification has limitations fordevicesthatdonotsupportJavaScriptmeaningthatattimeofPublicationitisverychallengingtomeasureviewabilityinmobile.

DeviceandCarrierreporting

SomeadservershavethetechnologyinplacetoallowthePublishertoreturnvaluesonmobileplacementswhichmayprovidemore informationaboutdelivery.TheseincludefieldssuchasDeviceandCarrierinformationwhichaPublishercouldpassbacktoadserverreportingwhentheaddelivers.Havingsaidthis,fewPublisherswillnotgothatstepfurther toprovidingthisextra informationsincethere isverylittlethatcanbedoneontheirparttocontrolthisformofdeliveryforthepurposesofmediaoptimizationlateron(SeeChapter6).

PaidSearchChannelReporting

Although paid search reporting is available from inside the Search Enginemarketing (SEM) interface, the ad server reporting interface presents the analystwiththeopportunitytoseethissamedataalongsidethedatafromotherchannelsallwithinthesameinterfaceaswellasconductingkeywordpathanalysis.

SearchOverview

StandardSearch channel reporting allows the trafficker and analyst to look at thedeliveryandperformanceofkeywords,ad-groupsandsearchadsalongsidetheCPCcostdataretrievedbytheadserverfromthesearchenginesviathesearchengine’sown reporting API. The difference is the ad server can establish conversionreporting(toadserverconversiontags)aswellaslookattheupliftinconversionsandclicks from thesearchchannel tootherchannels likedisplay (asexplained inChannelbyChannelupliftreporting).

KeywordPathwayAnalysis

A keyword path analysis report aggregates themost commonly searched for andclickedonsearch terms in their leadup to theconversion.Thesecombinationsofsearchtermsandthetimestampsbetweenthemhelptheanalystunderstandwherevariouskeywordssitintheuser’spathwaytoconversionandallowthetraffickertosubsequently up-weight or optimize the bidding on certain keywords back in theengine.Thiscanbedonebyanalysing the reportingorbyconnectingpathway toconversion feed data (see next section in this chapter) back up to the bidmanagement tool. In addition, somekeywordsmay not be driving conversions atall, as such the traffickercaneffectivelyeliminate themfrom thebiddingprocessfortherestofthecampaignorsubsequentcampaigns.

NaturalSearchReporting

Naturalsearchtrackingallowsthetraffickertotracknaturalsearchasthoughitwasanother channel in the paid-for marketing mix. However, the key to making themostofthecollecteddata, issettingupreportingcorrectly.Theriskis that ifuserpathwaystoconversionwereanalysed,mostpathwayswouldendwiththeclickofanatural search keyword in an engine. Under a basic attribution model like “lastclick”, natural search would appear to be driving all conversions. In standard

conversion reporting it would then appear as though all other paid-for channelscouldberemovedfromplanwithoutanyaffect.Infactconversionswoulddropoffor disappear completely, so for the Natural search channel, the trafficker mustensurethatfullattributionisnotawardedtoNaturalsearch.Thiscanbedoneusinganadvancedmethod inadserver reporting toseparateout theactivitycompletelyfromallotherchannelsandthenrebuildtheuserpathwaytoconversionlateron.

DoublecountingandSingleCounting

To still provide visibility on the natural search impact,while not diminishing thevalue of other channels under the last ad model (see Chapter 4), natural searchshouldsitcompletelyseparately fromothercampaignactivitybyhaving it live inanother Advertiser account entirely separate in the ad server to avoid de-duplication.There is anoption in theAtlas ad server, calledSingleCountingVs.DoubleCounting,whichallowsforthissameaffect.

If theuser sees a display ad, and then searchon a search engine for the featuredproduct and clicks a natural search result on the search engine results page thefollowinghappens:

• Thedisplayad,traffickedundertheAdvertiser’smainAdvertiseraccountrecordsanimpressionfollowedbyanimpressionbasedconversionfromthedisplayadtotheconversion tag on the landing page. The impression and conversion are recordedagainstcookieID001incookieleveldata.

• Onthelandingpage,thenaturalsearchscriptfiresaclicktracker,originallysetupasa tracking placement under the Advertisers natural search Advertiser account. Aclick is recorded in this account against this separateAdvertiser but also againstcookieID001.

• Thesamescriptthenfiresaconversiontag,setupunderthenaturalsearchadvertiseraccount, recording a click-based conversion against cookieID001 in cookie leveldata.

If the analyst looked in the pathway to conversion data (see next section in thischapter) for themainAdvertiser account the attribution for the conversion, underthe last admodel would go to the display event, with no natural search activityrecorded in sight. The analyst could then log out and log into the natural searchAdvertiser account. The pathway to conversion data report here would showattributiontothenaturalsearchclickactivity.Thesetworeportswouldneedtobecombined at the cookie level with timestamps matched, to see the user’s truepathway. What this work-around does, it that it preserves the attribution of theconversion to the display ad in the main Advertiser account without needing tocreateacustomattributionmodelorunderminingtheimpactofthepaidforactivity.

SomeadserversofferSEOspecificattributionconsiderationsautomaticallyalongwithaJavaScriptembeddedintofull-sitedistributedconversiontagssuchaswithauniversal taggingsolution(seeChapter4)so that there is littleornoefforton thepartoftheAdvertiserbutpredominantlyadserversrequirethestyleofworkarounddescribedtosupportnaturalsearchtrackinginadserverreporting.

Figure 5.4.2To calculate where Natural Search sits in the pathway to conversion, thesetuprequiresaworkaroundwhich feedreportingdata into twoseparateAdvertisers.Areportcanthenbebuiltfromthecookieleveldatatoseethepathwaytoconversionwhilestill awarding the final conversion to paidmedia channels in aggregated reports. Thisdiagramshowshowasingleseesanadat1pm.Laterinthesamedaytheusersearchesakeywordandnavigatesthroughtothesamelandingpageaswasdesignedforthepaid-forad (Ad 1). The script on the page registers conversions in separate Advertiser as thetracking is setup in thecampaignmanagementsectionof theadserver interface.At thecookielevel,thecookiehistorycanbeusedtorebuildthepathwaytoconversionacrossthesingleuser.

Thissetupstyle fornaturalsearchmeans thatwithinaggregatedreporting,naturalsearchbehavesthesameasanyreportingrelatedtoclickcounting.ThismeansthattotalclicksandclickbasedconversionscanberecordedagainstthischannelwithinthespecificnaturalsearchAdvertiser.

AffiliateChannelReporting

Theideaofaffiliatebasedreportingwithintheadserveristodeterminewhichaffiliateinthe affiliate network, given the use of a single ad, is responsible for the click andconversionactivity.

To calculate which affiliate site was responsible for each sale, the affiliate networkgeneratesauniquereferencenumberforeachaffiliatewhentheygeneratetheAdvertisercampaigncodebeforeplacing iton theirwebsite.When theuser clicks theactivityandgoesontopurchase,theaffiliateIDiscarriedallthewaytothepurchasepageintheclickpathwaywherethedataiscollectedandafeedof informationisreturnedtotheaffiliatenetwork about affiliate so that they can distribute the payment to the affiliate sites. Toachieve this theclick throughURLconsistsof separatepartswhicheach reportback totheirvarioustrackingtechnology.Thecombinationofthesepartsisreferredtoas“DaisyChaining”.

The output from ad server reporting on affiliate activity, lists the affiliate site name orunique identifieroneach rowalongwith the totalnumberofattributedconversionsandthe aggregated ROI. This is based on the extended data revenue passed through at thepointofconversionontheconversiontagextendeddataparameter.

The vast range of reporting optionsmeans that the number of metrics possible to pulldown from the third party ad serverwill be in excess of one hundred different values.

Reporting templateswhich can be customizedwill usually allow users to bringmetricsinto reports which may have no place being there: such as Rich Media metrics incustomizedPaidSearchreports.Itisforthisreasonthattraffickersandanalystsusingthereportinginterfaceareurgedtoplayaroundinsidetheinterface,beforepullingreportsforamoreurgentpurpose.Without thoroughpractice, the actof reportpulling canquicklybecomefrustratingunlessitisclearwhichreportwillreporttherequiredstats.

Thissectionconcludes thesummaryofaggregatedreportingoptions.The lastsection inthis chapter covers off the reports that are usually found outside of the ad servinginterface,althoughthedatastilloriginatesfromthecoreadservingdatabase.Anexampleofsuchdataiscookieleveldatawhichisusuallysodetailedandgranularthatitrequiresmuchgreaterprocessingpowertooutputandpresent.AssuchthisstyleofreportingwillusuallybedeliveredmanuallytotheAdvertiserorAgenciesFTP(FileTransferProtocol)site.

Chapter5-Section5CookieLeveldata• ReportingattheCookieLevel

▪ Figure5.5.1

• Howisthedatasupplied?

▪ Look-backDataDump

▪ CookieLevelFeed

▪ Report delivery, size estimation and storageconsiderations

▪ MetricsinCookieLevelData

▪ Trackingimpulseconversions

▪ PathwaytoConversionReporting

▪ Figure5.5.2

▪ PointSolutionsforAdvancedCookieLevelAttribution

• ReportingonConversiontagActions

• ExtendeddataattheCookieLevel

▪ Figure5.5.3

▪ Figure5.5.4

• ImpressionandClickbasedtokendataattheCookieLevel

• Whatismissingfromadserverreporting?

Aggregatedreportingbringstogethervariousmetricstogiveahighlevelviewaboutthetotalnumberofthosemetricscountedforagivenadserverhierarchyentityoveragivenperiodoftime.

Thismeans that the traffickeror analyst can log into the thirdpartyad server reportinginterfaceanddownloadareportbyagivendaterangetoshowtotal impressions,clicks,conversions,interactionsbyAdvertiser,Brand,Publisher,Site,Section,Domain,Package,Placement,Ad,CreativeandVersion.Toreallylearnexactlyhowuserscameacrossadsandthestorybehindthepathwaytotheconversion,itisessentialtodivemuchdeeper.Incookieleveldata,insteadofeachrowofthereportbeingasummaryoractivitybyentity,eachrowisasummaryofactivitybycookieIDorat themaximumgranularitycanbeasummaryofasinglerecordedactionoreventforacookieID.

ReportingattheCookieLevel

Cookie level data exposes every recorded event in the third party ad server for anAdvertiserasitisrelatedtoeachuser(oreachuniquecookieID)andplacestimestampsagainstvariousactivities.Injustafewrowsorcolumnsofthislevelofrawdata,itiscleartoseehowandwhereausersaworclickedanadinacampaignoveranextendedperiod

of time if or before they reached a conversion tag and converted. This is enormouslyuseful for analysts and for support teams troubleshooting misbehaving ads as itdemystifiestheaggregateddata.Thesimplicityoftheaggregateddatafailstoexplainhoworwhen the aggregated datawas collected andwithout this context, the reports can bemisunderstood.

Thisisparticularlytruewhenaskingtheanalysttoanswermorecomplexquestionswiththedatasuchas:Whatproductsdidmymosteffectiveads,driveuserstopurchase?Oristhere a relationship between the final cost of the purchased product and the Searchkeyword? Only cookie level data can glean these answers without specialist buildaggregated reports to pivot the data in ways that may bemore specific to a particularAdvertiser.CookieLevel data is typically available to theAdvertiser orMediaAgencyfilteredinthefollowingways:

• CookiestoallImpressionsonly(withtheoptionofaddinginteractionsandIP)

• CookiestoallClicksonly

• CookiestoallImpressionsandClicksonly

• CookiestoallConversionsonly

• CookiestoallImpressions,ClicksandConversions(withtheoptionofaddinginteractionsandIP)

• PathwaytoConversion

Eachfiltermethodprovidesadifferentoutput totheAdvertiseralthoughallwillbeinaview that looks like a spread-sheet or CSV format; make it easier to conduct analysisacrossthedata.Thebenefitofchoosingonefilteroveranotheristheamountofdatathatthefilewillcontain.

Figure5.5.1anexampleofasinglerowofcookieleveldatafilteredbyimpressions.HereauniquecookieIDwasseenforaspecificadtogenerateanimpression,withinaspecificplacementonaspecificsiteatsinglepointintime.Acookieleveldatareportwilltypicallycontainmanythousandsofsuchrows.

Cookieleveldatacanbevast.Foreachpassingsecond,millionsofusersareencounteringadsandthentakingactionsorconvertingwhichcreateshundredsofmillionsofrowsinanenormous database behind the scenes at the ad server. Cookie level data is commonlymisconstrued as the raw third party ad server counting database (which is built in themomentsthatadsareservedandmetricstracked).Thisisrarelythecase;cookieleveldatais constructed from the raw data later on. The raw counting database is produced andstored separately in a much more garbled format, that cannot be read by the analystwithoutassistanceordecryption(thisdatasetisoftenreferredtoastheserver“logfiles”).

CookieleveldataistypicallyprovidedtotheAdvertiserorAgencybythethirdpartyadserving provider at additional cost, since it takes a lot of processing power to churnthroughandbuildthedataset,beforeusingupalotofbandwidthtodeliverit.

Howisthedatasupplied?

Cookieleveldatacanbesuppliedintwotypes.

1. Alook-backofcookieactivityoveraperiodoftime(dependingonhowlongthethirdpartyadserverstoresorhasbeencollectingthedatafor)

2. Afeedwhichsuppliesdatafromthepreviousdayorweek.

Look-backDataDump

Third party ad serving providers do not usually collect cookie level data as adefault.Thecollectionofthedata(inaformatthatcanimmediatelybepresentedina cookie level report) needs to be switched on by the trafficker or the ad serversupport team.From the point atwhich it is enabled is the earliest date onwhichcookieleveldatawillbeavailableasareport.Alook-backdatadumpcanprovidetotheAdvertiserorAgencyatapointinthefuture,amassivefilecontainingsomeorallofthecookieleveldata.Thesefilesareusuallyverylargeandaredifficulttoopen and read. It is recommended for analysts workingwith data of this size toinsteadopt for receivinga trickleof thedataeverydayasa feed.ThiswillallowthemtobuildupalargerdatabaseofthedataontheAdvertiserorAgencysitelateron.

CookieLevelFeed

RequestingafeedofcookieleveldatafromanadservingproviderallowsAgenciesandAdvertiser to store themost granular data availablewith the freedom to runcomplex formulae across it at a later date (or in parallel to running aggregatedreportingfromthethirdpartyadserveranalyticsinterface).LargerAgenciesinvestintechnologiesthatpickupthecookieleveldatadeliveredbytheadserver,processit, and present it in customized ways. Some of these in-house technologies thenautomatically take actions as a result of the data such as buying mediaprogrammaticallyforsearchviaabidmanagementsolutionorforDisplayinventory(seeChapter8).

Reportdelivery,sizeestimationandstorageconsiderations

CookieLeveldatasetscanbemonstrouslylargeinthelookbackformandevenasafeed;theamountofdataquicklybeginstoaddup.Thereforethebiggestcostisnottheretrievalofthedatabutthelongtermstorageofit.Despitetheglobalcostofdatastoragestillshrinking,thesedatafilesaretoolargeforeventhebestadserversintheworldtoaffordtostoredataforallAdvertisersforalifetime.Furthermoreasadserversreleasemoresophisticated technology and features, the amount of data that can becollectedquicklyandquietlygrows,creatingfurtherstoragedemands.

Usually an ad serverwill supply cookie level data to a secure FTP sitewhere the Agency or Advertiser will retrieve the data set on a rollingcadence.The ad server is therefore expected to deliver such data at thesametimeeachdayorweeksothatsubsequentanalysisfromthatdatasetcanbeplannedforeffectively.Thetraffickercouldcalculatehowlargethecookieleveldatafileswillbebylookingataggregatedreports.Thetotalnumber of impressions will indicate the total number of rows in the

cookieleveldatareport.

MetricsinCookieLevelData

Cookie level data typically matches four key metrics of data to the ad serverhierarchyentities;impressions,clicks,conversionsandrichmediainteractions.Onoccasion cookie level data is enhanced with geographical information orverificationdata.Analystsandmarketersthatplanaheadwouldthinkgettingalloftheavailabledata is thewisestdecisionbutofcourse thismeansmuch larger filesizes for storage. Some ad serving providers get around this issue by offeringmapping files (where names in a particular column or group of columns is justrepeatedoverandover).Insuchaninstanceamappingfileissuppliedtoallowtheanalysttotransformtherepeateddata,thismethodshrinksthefilesizeofthelargerdatasetandmakes thedatamoreportable.An importantconsiderationforcookieleveldataisusercookiedeletion.Ausercanbeexpectedtodeletecookieswithin90daysor receiving them,so thedata isnotuseful long termfor thepurposesofretargetingthesameusersagain(seeChapter7).Ratherthedataisusedtosurmounttrendsandpredictbehavioursforthepurposesofoptimization(seeChapter6).

Trackingimpulseconversions

Sincebasicaggregatedreportingpredominantlyreportsatadaybydaylevel,thosecampaignsrequiringafasterresponse,targetingandbuyingbythehourinsteadofby theday, require timestamps tooptimizeaccurately.Althoughdaypart analysiscan be revealing to an extent, it fails to answer questions about user actionsundertakensuchas“impulseconversions”whichmayonlyhappenmomentsafteruser exposer to an ad. Cookie level data can reveal this information whereaggregatedreportingcannot.

PathwaytoConversionReporting

The most common reason Advertisers and Agencies use cookie level data, is tounderstandmore about the cookie-levelPathway toConversion (P2C) also called“Exposure to Conversion” or “E2C”. This granular report shows how eachconvertingcookieIDencounteredany trackedor servedAdvertiser activitybeforegoingontotriggeraconversiontag.Therecordofthetotalnumberoftouch-pointsshown in this report for each cookieID, on its pathway to conversion before theconversionactivityistypicallylimited,sothatthefilesizeisnottoolarge.SmallerAdvertisersareusuallycomfortablewithatouch-pointlook-backoften“eventsinthepathway”butthelargerAdvertisers,withbiggermarketingteams,morebudgetfor data analysis and complex,wider reaching campaignsmay require a pathwaylengthofmorethanonehundredtouch-pointsorevents.

Figure5.5.2FullP2C(pathwaytoconversion)reportingatthecookielevel.Suchreportsare granular and very extensive. Diagram depicts just four touch points on the userpathwaytoconversionwhichinactualfactcouldcontainmanymoretouchpoints.EachtouchpointoreventismadeupofdataaboutthePublisherpageandthead.Thisincludesthetimestampsforseeingorclickingthoseadsaswellasadandplacementclassificationinformation.ThefinaltouchpointisfollowedineachrowbytheconversioninformationandextendeddatasothatforeachCookieIDitiscleartotheanalystallthedatapointsthatwerecapturedthatinsomewaycouldbeattributedtotheconversionevent.

Having ten events in thepathwaymeans that any events that occurredbefore thelasttenareremovedfromthedataset,sincetheseweresofarawayfromtheactualpointofconversionthattheycansafelybedisregarded.Ontheotherhand,ifsomeoftheseeventsloggedtheuserencounteringparticularlyimpactfulexecutions,thenitmaymakemoresensetolookbackfurtheror tohaveamoreintelligentwaytofiltertheevents.Thisiswhereattributionmodellingisuseful,sinceitcanfilteroutorplaydownpathwayeventsandprovideanindicationastothecontributorstotheconversionfromwithinaggregatedreporting(seeChapter4).

Advertisersthatrunalotofactivity,saturatingtheonlinespacewithadswillwantto look atmore events in the pathway to determinewhat impact all their ads arehaving acrossmultiple campaigns for all converting cookies. Such situations arehighly complex and require analysts to write additional programs to process thecookie level data, assigning statistical value not just to attribution that a singlepathwayeventwouldhadtoaconversiontag,buttoallconversiontagsacrossallcampaigns.

PointSolutionsforAdvancedCookieLevelAttribution

Beyond the conversion, someAdvertisers want to analyse what impactpathway events have when such events lead to non-conversionperformance events such as clicks, interactions or improvements inverification and viewability. Many technologies in the space are illequipped todealwith this level of complexitybut somepoint solutionshaveemergedtoaddresssuchanalyticaldemands.ThesesolutionsincludeClearSaleing, Encore Metrics and Visual IQ. All can be setup to re-processcookieleveldataoutoftheadserveratthemaximumgranularityorhavetheirownpixelsplacedinads,clickpathwaysandconversiontagsinordertoreplicatethedatasetintheseseparatesystems.Theissuewith

implementingadditionaltrackingforthispurposeisthattheycancomeatahighadditionalcosttoserveandtrackthecampaignforinsightswhichmay not provide the best ROI. In addition, the additional trackingtechnologywillbecountingtheimpressions,clicks,conversions,actionsand interactions at a very slightly different time from the ad server sotherewillbeaminordatadiscrepancywhichmustbetakenintoaccountwhen conducting an analysis. For these reasons it is recommended thatadvancedattributionpointsolutionstakeandprocesstheexistingcookieleveldata,returningtheinsightstotheAdvertiserorAgencydirectly.

ReportingonConversiontagActions

Granular reporting beyond aggregated reports are referred to by some ad servingproviders as “Raw Data Reports” or “Data Pass Back” reports, and offer agranularity such as an aggregated by conversion tag extended data customparameters (see Chapter 4) or a list of conversion tag “Actions”; showing thenumberoftimesthataconversiontaghasfiredinadditiontoconversions.Areportofthisnaturecanbecrucialtotroubleshootingthebehaviourofconversiontagsasitshowsthenumberoftimesaconversiontagisloadingandnotjustthetotalsalesorcountsattributedtotraffictotheconversiontag,comingfromanad.

Advertisers that look at this data are able to make a better inference about theimpactofonlinetrackedactivitybecauseofhowmanynon-conversionActionsaretakingplaceontheconversiontag.IfthenumberisverysmallthentheAdvertiser-side analyst can be assured that the majority of sales are being tracked andattributedtopaid-foractivityonline. If thenumberofActions isveryhigh thenitbecomes tougher to identify the sale attribution for untracked channels and mayindicateaneedtoextendtrackingtootherchannelsormarketingactivity.

Conversion-basedextendeddataattheCookieLevel

Cookielevelreportingisalsooneofthefewplaceswhereextendeddataparametersfromtheconversiontagscanbecaptured.ThiswouldincludetheOrderID,Quantityofitemspurchased,andrevenuegenerated,aswellascustomisedparameterssuchas the names of the products purchased or their properties. These properties arepresentedoneachrowofthecookie-leveldatawithinadditionalcolumnsfeaturedafter the conversion tag name and type, where a cookieID has performed aconversionactivity.

Figure5.5.3TheConversionCookieLevel data contains the extendeddata parameterspassed through by the Advertiser script. Apart from the default parameters such asOrderID and Revenue, this includes the custom parameters. It is worth the analystconsulting the trafficker to obtain the campaign setup information regarding theconversiontagssothatamappingtablecanbecreatedtointerprettheextendeddata.Inthis example a travel Advertiser collects data about flight bookings. Each customparametercontainsmoreinformationaboutthebooking.

Toenrichthecookieleveldatafurther,anAnalystattheAdvertiserorAgencysidecan marry the OrderID field between the third party ad serving data and theAdvertiserPOSorCRM.ThisisbestachievedbringingthetwodatasetstogetherinaDataManagement Platform (aDMP - seeChapter 8). Bringing these data setstogether allows anAdvertiser to build a fullmapof specific customer behaviour,sincethecookieIDbecomesenrichedwithdatathatmightofferthechancefortheanalysttoperformamethodofprofiling.

Forexample,theOrderIDcouldindicatefromthedataintheCRMthattheOrderbelongstoanaccountIDwhichhadundertakenpurchaseactionsoverthelastfewyearstoindicateacompactbuyingpatternintheleaduptheChristmasholidaywiththesameAdvertiser.Usingtheadservingdata,thecookieIDmayhaveundertakenacertain pattern of site visits displaying the Advertisers ads before purchase. Theanalyst could see if there was a trend between users of look-a-like profiles anddecidewhetherornot to inform theplannersandbuyers toconsiderbuyingmoreinventoryonsuchsitesclosertotheendoftheyeartoattractthebuyingbehaviourfrom these users and similar users. Alternatively creative messaging could beadapted using DCO to perform a targeting function across inventory that hadalreadybeenpurchasedonsuchsites.

Figure 5.5.4 the extended data from the Cookie level on the conversion tag can bemarriedwith theCRM information from theAdvertiser side insideaDataManagementPlatform (DMP). Such systems enable the data to be connected with a programmaticoptimization(seeChapter8)inordertoeitheradjustcreativemessagingortargetingforchannelslikeDCO,adjustbidpricesonkeywordsforpaidsearchorauctionsandbidsintheprogrammatic space throughaDSP (seeChapter8).Of course themarrieddata isalsohugelyinsightfulforthemarketer.

ImpressionandClickbasedtokendataattheCookieLevel

Publisherscanpassinformationintotheimpressioncallortheclickpathwayoftheadbypopulatingatokenfieldintheadtagcodewhentheadtagcodeisfired.ThisisexecutedbyaPublishercreatedscriptthatsitsonthePublisherpage.Anyextrainformationpassed throughineitherof thesedynamic tokenfieldscanbeused topopulate additional columns in cookie level data. Both of these fields are highlyfavoured bymarketers that want to get a better idea about exactly where the adappearedonthepage.SincethePublisheradopsteamistheoneimplementingthead,evenpagescrapingthedomainorpagelevelURLwheretheadwasdisplayedisnotenoughinformationtoconveywhereaboutonthepagetheadsat.Thesefieldscan be populated by placement position information such as ‘position 4’ so thatwhenanalysingthecookie-leveldata,theanalystcandetermineifadperformanceorverificationareaffectedbythepositionoftheadspot.

These advanced strategies take careful planning analysis. Some Agencies ormarketing teamshireDataAnalystswhoprocess such data sets full time, for thepurposes of campaign optimization. Analytics come with an important costconsiderationwhenpurchasingthemediainthefirstplace,sothefullbudgetshouldnotbeblownonadspacebutalsooninterpretingandpresentingtheresults.Evenboltingonothertechnologieslikethirdpartydataproviders(seeChapter8)insteadofphysicalanalysts toget thisdata insight,canaddsignificantcostand thereforeaffectthebottomline.

Whatismissingfromadserverreporting?

Adserverreportingisabsentonekeycomponentinbothaggregatedandgranularforms.Theeffortsandthecontrolsthatthetraffickerimplementsdonotalwayshavethosesettingfeddowntoreporting.Thiscanmeanthatifthetraffickerchangesaconversionwindow,arotation setting or the definition of an audience to omit a certain geography, these“optimizations” are not recorded in reporting. If the report displays a sudden spike inactivity, thecausemaybedownto theoptimizationbutonly thecampaignmanagement

sectionoftheadserverinterfacewillretainarecordoftheexistingsettingsandmaynoteventhavekeptanaccuratelogofchangestothosesettings.

It falls to the trafficker to maintain an accurate log of both campaign setup and anysubsequent changes to the settings, so that reporting can be coloured with thoseoptimizationsthatwereresponsiblefortheanalyticsoutput.

Thenextsectionfocussesonoptimization,aresponsibilityforthetraffickertoundertake,inresponsetotheadserverreportinginformationinordertoimproveexistingcampaigns,utilisingadserversettingstodoso.

Chapter6Optimization

Chapter6Optimization

• Questioningreportingdata

1. Wheredidthereportcomefrom?

2. Whatkindofreportisit?

3. Dothenumberslookright?Doesthedataaddup?

4. What timewas the reportpulled,what timezone is theaccount inandarethesethecorrectdateranges?

5. Arethesethecorrectmetrics,conversiontagsandhierarchyentities?

6. Do the names of the entities in the report (including the accountname)matchthoseinthecampaignmanagementinterface?

7. Whatisthemethodologybehindthesemetrics?

8. Istheshowngranularityprovidingenoughortoomuchinformation?

9. Isthereenoughdatainthereportorwerethetrafficvolumeslargeenoughtogaugedecentstatisticalaccuracy?

10. Is thereachance that thedata in the reportbeadjustedagain lateron?

11. Arealltherowsandcolumnsdisplaying?

12. Was the campaign setup in way that might affect a governedoutcome?

13. Didtheadserverencounteranykindofcountingfault?

• WhatisOptimization?

• OptimizingtoCampaignObjectives

• OptimizingtowardsDeliveryGoals(idealforBrandAdvertisers)

o UniqueReach

o TargetAudience

o Verification(qualitysafety)goalsandimpact

• Optimizations towards Performance Goals (ideal for DirectResponseAdvertisers)

o EngagementandInteraction

o TotalClicks

o TotalConversionsorE-Conversions

o ROI(&extendeddata)goals

• OptimizationSample

• RecordingchangesinOptimizationstrategy

Ananalysisof theadservingdataearlyandofteninacampaignprovidesaninvaluablefeedbackmechanismtotheAdvertiserorAgency.Byeithermakingchangesatacreativelevelormakingchangestotheplanningandbuyingstrategies,theoutcomeoftheoverallcampaign can be influenced before its final impression has been delivered. Reportingtherefore can act as the safety net to ensuring a campaign endsmeeting the campaigngoals or it can be used to take lessons from one campaign to apply to the next. TheapplicationofthislearningtoalivecampaignisreferredtoaseitheraMediaoraCreativeoptimization,dependingonwhetherthemediaplanner/buyerorthetraffickertakesaction.

Thefollowingsectionexplainshowtoquerythereportingdata,beforeexplaininghowtounderstandit,totakepositiveactiononit.

QuestioningReportingData

Once the relevant report has been run and exported from the third party ad servingreportinginterface.Theanalystneedstomakesenseofthedata.Thefollowingquestionsaddmuch needed context to the data and ensure that it is of sufficient quality tomakeoptimization decisions from. The analyst should know the answer to these questionsbefore proceeding. If the answers are not available from the report or the help sectionwithin the interface itself, the analyst can usually contact the support team at the thirdpartyadservingprovider.

1. Wheredidthereportcomefrom?

Having a colleague supply the data is not enough of a safeguard to ensure thenumbershavenotbeentamperedwithbyhumanhands.ThisiswhythirdpartyadserverreportinginterfacesgrantaccesstoaverywiderangeofusersincludingthePublishersandtheCreativeAgencies,sothattheycanensurethedataisgenuine.

Pulling a campaigndirectly from the thirdparty ad serving reporting interfaceorsecurely receiving the data from the third party ad serving provider’s deliverysystem (emailorFTP) is theonlyway toensure thevalidityof the sourceof thedata.

2. Whatkindofreportisit?

The report being pulled from the interface may not contain the right metrics toanswer questions about the delivery or performance of the campaign. Pulling areportandrunningit,onlytofinditcontainsthewrongdatacanbefrustratingandverytimeconsuming.

Checkthereportdescriptionandtitlebeforerunningordownloadingit.Ifitisnotclearifitistherightreport,adserverswillhavesupportingdocumentationorhepsections to define the metrics it contains. In addition it is possible to display apreviewofthereporttemplatewithoutanydatainittoseeifitlookslikeitmightbetherightone.

3. Dothenumberslookright?Doesthedataaddup?

Itpaystostartwithahypothesisbeforelookingatareport.Thissetstheanalysts’expectationsaboutthedatathatwillberevealedinthereport.Inadditionitisworth

tallying up the numbersmanually to check that the total columns add up. If thenumbersarewildlyoff,somethingmightbewrongwiththereport,thedataorthecampaign.

If the data does not look right, pull the report again to see if the data looks anybetter.A single glitch or bug in the reporting interface for a temporary period oftime could be affecting the way the numbers are outputted. If the data does notappearasexpected,thetraffickercangobackandtesttheads,retargetingtagsandconversion tagswithFiddler, checking that they rightcallsarebeingmade to theserver.Thesesamepiecesofcodecanalsobetestedintheliveenvironmentofthecampaigntoseeifthecodeisimplementedcorrectlyorifthetagsarebeingaffectedbyothercodeonthepage.Ifeverythingisperformingcorrectly,readtherestofthequestionsinsectionandperformthosechecksbeforeconsultingthesupportteamatthethirdpartyadservingprovider.Forreportsthatcontainconversiondata,checktheconversionwindowsettingsandattribution settings for conversions. These may be causing the total numbers ofconversionstobehigherorlowerthanexpected.

4. Whattimewasthereportpulled,whattimezoneistheaccountinandarethesethecorrectdateranges?

Dataprocessingoccursatdifferenttimesandindifferenttimezones.ProcessingcanalsobeaffectedbythetimezonethattheAdvertiser,AgencyorPublisheraccountissetto.Pullingareportinadifferenttimezonetothetimezoneontheaccountcanmeanthatthelatestdatamaynotbeavailableyet.Inadditiontothis,itiscommontopullareportwithadaterangethatdidnotmatchthedatesthatthecampaignwaslive.

To resolve this the trafficker can check the account settings in the campaignmanagementinterface.Inadditiontheanalystcancheckthatthedaterangesinthereport, match the ones required. It may sound obvious but a sizeable amount ofqueriesaboutdataaredowntoincorrectlymatchingdateranges.

5. Arethesethecorrectmetrics,conversiontagsandhierarchyentities?

In haste it is possible to pull reportswith thewrongmetrics inside or looking atconversionsforconversiontags thatarenotonthepages theanalyst is lookingtoanalyse.

Whencreatinganynamingconventions theseshouldbesharedwithallparties sothat it is clear which names match with what. Also analysts should be sure tounderstand ad server hierarchy levels, for new beginners it can be common toconfuseaplacementwithapackageforinstance.

6. Dothenamesof theentities inthereport (includingtheaccountname)matchthoseinthecampaignmanagementinterface?

Theanalystmaybeworkingfrominformationwhichisnottrue,beingaskedtopullconversiontagresultsforcompletelythewrongconversiontag.

Theanalystcanflipbetweenthecampaignmanagementinterfaceandthereportinginterface toensure that therightentitiesarebeingselected.Theanalystshouldbe

warythatiftheAdvertiserdoesnotuseanintegratedtagmanagementsystemitisnotpossible toconfirmwhichpageconversiontagsare implementedon.Analyststhatsuspectfaultsinconversiondatashouldasktraffickerstocheckthatconversiontagshavebeenimplementedcorrectly.

7. Whatisthemethodologybehindthesemetrics?

Thedefinitionofametricmayprovideaninsightintowhyitislow,highorzero.ForinstancethedefinitionofanimpressionaccordingtoIABistocountwhentheserverreceivestheadrequest.Animpressionelsewheremightcountwhentheadisdisplayed on the page. Comparing reports of the same activity but differentmethodologies isnot recommended, thedatawillneverbe thesame.Also,notallcounting systemscancount thingsat exactly the samepoint in time.Usuallyonesystem is counting things before another which at ad serving scale can result indifferentfigureswhensystemsarecompared.

Toresolvethis,theanalystshouldchecktoseeifaknownandaccepteddiscrepancyexists between a comparative report. The IAB confirm that an accepted level ofindustry discrepancy for reports that track the same things in different systems(usingthesamemethodology)shouldnotbeoverorbelowtenpercent.Therearethreetypesofpossiblereportingdiscrepancy.

a. Betweendatafromthesamesource(reportspulledfromthesameadserver)

b. BetweendatacomparingthirdpartyadserverwithPublisheradserver

c. Between data comparing the third party ad server with another technologyprovidersuchasa4thpartyorthirdpartyadserver

Although an ad serving provider can commit to investigating reportingdiscrepancies, typeb. ismuchmorechallengingtoresolve,sincethetwotypesofsystemreportinverydifferentwaysanditisinthebestinterestofthePublishertoreportinflatedfigures.InsteadPublisherscanbegrantedaccesstothethirdpartyadserverreportinginterface.

8. Istheshowngranularityprovidingenoughortoomuchinformation?

Themostgranularkindsofreportingaredifficulttoactionwithoutspendingmoretimeorresourceontryingtoextracttherightinsights.Ontheflipside,ifasingleadtag had been provided to a Publisher to run on all page types of their website,reportscouldnotgiveanindicationofperformancebysitesectionforinstance.

Gettingtoknowwhichreportswilldisplaywhatlevelofdatagranularitywillmaketheprocessofanalysingandpullingreportslessfrustrating.Displayingapreviewofthe report template before running it will give a good indication of the level ofgranularitythatwillbeavailable.Italsohelpstohaveastrongunderstandingofthedefinitionsofthevariouslevelsofreportinggranularity.Ifreportingdoesnotseemtoyieldgranularenoughdata,analysethecookieleveldatainstead.Bycomparingthe domain URL captured from the page level where the ad was delivered andamalgamate up the results by subdomain, the analyst can get section summarieswithout having to traffic more ad tags. To save costs on analytics however, thetraffickershouldcreatemoretags,replicatingthesameadassignmenttoplacement

setupwithsectionnames,forthePublisheradopsteamtoimplement.

9. Isthereenoughdatainthereportorwerethetrafficvolumeslargeenoughtogaugedecentstatisticalaccuracy?

When dealing with very big numbers, very small numbers in reporting becomeclose to insignificant. If a campaign was delivering one hundred millionimpressions, a placement or a single ad version that had incremented a thousandimpressions alongside versions that had incrementedmillions of impression, is aversionthatmayjustaswellhadincrementednoneatall.

Thetraffickerandcampaignplanningteamshouldhaveaccountedfordistributionofdataacrossthesetup.Thisisrelevant tothecreationofaudiencesegmentsandversions forDCO.Onsuchascale,wait foradserverentities to increment largeramountsofdatabeforeconductingananalysis.Thisiswhyitisrecommendedforthe analyst towait a day or two before pulling reports for the first timewhen acampaign goes live, ensuring that the data displayed are more than just thePublisher running a few tests on the ads while implementing them. If the datacontinuestobetoosmallthetraffickercantaketheexecutionofthecampaignsetupaltogetherorsimplifytheexperiment.

10. Isthereachancethatthedatainthereportbeadjustedagainlateron?

Theanalystwillwanttoconductananalysisacrossthefinalresultsratherthanhavetoundergotworoundsofanalysisonalmostthesamedatasetbecausethenumberschange. In particular, numbers in ad server reporting sometimes change forconversionsandforSearchchannelcostdata.Conversionscanchangewhentheadserver conducts several rounds of data processing owing to complex attributionmodelling and de-duplication. The Atlas ad server used to run one round ofconversion processing, delivery a near-accurate conversion figure to get earlyresults available in reporting, before conducting a more thorough sweep anddelivery within twenty four hours. AlsoGoogle Adwords sometimes recalculatessearchenginecostdataoverthecourseofaweek,whichispresentedinadserverreporting.

In all instances check with the support team at the ad serving provider that thereporting numbers displayed will not change and get confirmation about timingsandconditionsonwhichthedatacanchange.Forconversiondata,theanalystcanwait until conversion processing is complete before pulling reports for the moreaccurate numbers. For search cost data, the analyst can pull reports daily for theprevious week look back and compare those numbers to ensure cost data hasremainedfixedandisnotchanging.

11. Arealltherowsandcolumnsdisplaying?

Sometimes,particularlywithlargeAdvertisers,oldcampaignscangetarchivedbytheadservertospeedupthetimeittakestorunbigreports.Thisisalsosometimestrueofrowsorentitiesthatdisplaynodataorzeroes.Thisisusefultospeedinguptheperformanceofthereportinginterface,butisanincompleteviewforananalystthatmaywanttotakeoldorlowfiguresintoaccountwhenanalysingthedata.

Ifareportispulledandrowsorcolumnsexpectedseemtobemissing,theanalystcan run the report again, ensuring that the filters included are not filtering outcertainmetrics,entitiesorformsofdata.Alsoitisworththetraffickercheckingtheaccountsettingstoseeifandwhentheadserverisarchivinganyofthedata.Wherereportscannotberetrievedbacktoacertaindate,theanalystcanchecktoseehowlongtheadserverkeepsdatastoredforbyaskingthesupportteamattheadservingprovider. Inaddition there ismerit in asking the samesupport team tocheck thatnoneoftheolddataisbeingarchived.

12. Wasthecampaignsetupinwaythatmightaffectagovernedoutcome?

Carefulplanningisessentialinthecampaignsetupprocess.Ifthetraffickerandthecreativedeveloperarelessexperienced,theymaynothavethoughtaboutthereportoutcome as a result of the setup. Creating too few tags for instance won’t giveenoughgranularitytoprovidethemostinsightfulresults.InadditiontheremightbehighnumbersassociatedwithmorenegativemetricssuchasincompletevideoplaysforInstreamornoclickdataonahighreachDisplayexecution.

BeforestartinganyRichMediacampaign,thetraffickershouldhaveakickoffcallwiththecreativedeveloperandPublisher,makingsureinteractionsareintherightplace.Followingthis,theadimplementationshouldbethoroughlycheckedduringthe ad QA process. Once the campaign is live, verification reporting will reporterrors in targeting on the Publisher side such as ineffective brand safety. Bychecking delivery reports early and often, the analyst can avoid results thatdemonstrate an incorrect implementation by having the trafficker work withPublishersfixincorrectimplementations.Toreachcertaingoals,acampaignshouldbecorrectlyoptimized;thisiscoveredinfulldetailattheendofthesection.

13. Didtheadserverencounteranykindofcountingfault?

If the data reports significantly lower or higher volumes for delivery andperformance there is apossibility there couldbea counting fault in the ad serveritself.Thisisaveryrareoccurrenceandshouldonlybeconsideredasaconcernasalastresort.

Afteralltagshavebeendoublecheckedandtherestillappearstobebaddatainthereport, the analyst can contact the support team at the third party ad servingprovider.Thisteamwillbeabletoconductaninvestigation.Intherareoccurrenceof a fault, data can usually be reprocessed and re-published to the correct timeperiods inadserver reporting.Dependingon theseverityof the issuedependsonhowlongthisfixwilltake.Whenconsideringusinganadserverforthefirsttime,the provider can supply information about data security and data integrity. TheAdvertiser or Agency can expect certificates of accreditation from well-knowninternetandhardwaresecurityinstitutionswhovettheadservertoensurethatdatafaultsandserverdowntimearevirtuallyzero.

Once the analyst has established that the data is an accurate reflection of the live orcompletedcampaign,theprocessofoptimizationcanbegin.

WhatisOptimization?

Optimization is utilising ad serving data tomake an improved change in the campaignsetup. This change can either occur in the ad serving campaignmanagement screen oradjustmentscanbeundertakenbytheplannersandbuyers,changingwhichPublishersandaudiences the ads still to run would run against. These are referred to as CreativeOptimizations (in the ad server interface) or Media Optimizations (adjustments to themediaplanoutsidetheadserver).

Chapter3-DCOexplainedhowtheadservercanmakeuseoftheadservingdatawithoutanyoneneedingtopullreports.Theadserverwouldmakeuseofthedatainrealtimetoadjust the creative optimization of the ads using “auto-optimization”.The remainder ofthissectionwilllookinsteadatthealternative,manualoptimization.Thismeansexportingreportsfromthereportinginterfaceandhavingananalystlookatthedataandinstructthetraffickerorthemediaplannertomakechangestothecampaigntoimproveperformanceordelivery.InthissenseoptimizationmightallowanAdvertisertoreachacampaigngoalfasterorexceedit,ideallywithoutincreasingcosts,butreducingtheminordertodoso.

ThemoststringentofAdvertiserswillinsistthatoptimizationsoccurasoftenaspossible.Thismeansimplementinganoptimizationstrategyandthenwaitingfortheadjustmentstotake effect before optimizing again. It is for this reason that having real-time timeavailable data is not so useful for real-time optimization on small-scale campaigns.Optimizationswillusuallyneedaminimumofadayortwosworthofdatatoclarifyiftheadjustments have worked and daily aggregated reporting from the ad server is usuallysufficientforthisneed.

OptimizingtoCampaignObjectivesOptimizationmeansmaking themediabudgetgofurther.First, theanalystcanpullcostreporting from the reporting interface to see where themajority of themedia spend isgoing, then look more deeply at the data to see if those Publishers or the Creative isdrivingadspendto thosegoalseffectivelyorhavingastrongcontributionto it. Ifgoalsarenoteffectivelybeingmet(ortoimprovethingsfurther)theanalystcanconsiderhavingtraffickersandplannersmakesomethefollowingadjustmentsasoptimizations,dependingonthecampaigngoal.

OptimizingtowardsDeliveryGoals(idealforBrandAdvertisers)Ifanadcanbeseenbythemostrelevantaudienceitcanhavethemostimpact,eveniftheAdvertiserhasnotsetanyperformancegoals.Byensuringadsaredeliveredtothemostquality,relevantpages,attherightreachandfrequencyandtothetargetaudience,thereisabetterchancetocreateawarenesswiththoseusersthatencountertheads.

UniqueReach

TheoptimizationofUniquereachisagoaltofindanuntouchedaudienceorrecognisinghowmuchexposureauniqueaudiencerequirestocementtheimpacttheadcanhaveonthoseusers.Itisadvisedtofirstsetabaselinetoestablishhowmanyuniqueeyeballsorusers the campaign is reaching. To do this, pull a reach report by Publisher and byChannel.SomechannelsandPublishersmaybemoreeffectivethanothersatreachingtheintendedaudience.Thisdatacan thenbecombinedwith frequency information toshowthe levelsof frequency thateachPublisherandChannel is incrementingoneachunique

user. Combining this with cost reporting will showwhat it is costing to deliver at thebaseline.

Thenext step is to consider the importance of reaching a unique audience.Usually thegoal isharbouringagoalofperformance to reachacertainnumberof sales. If this isagoalthatcanbereachedonlinethentheanalystcanconsideroptimizingthecampaigntoconversionsas aprimarygoaloverunique reach.Byanalysing theattribution funnel incookie level data, the analyst might reveal that might show that frequency is moreeffectivethanuniquereachatdrivingthatconversiongoal,sofrequencycanbeadjustedtomatch thisneedeitheron the thirdpartyadserverorwith thePublisheradops teamdirectly.Sometimesthegoalforuniquereachdoesnottiebacktoonlineconversionsbutinstead might be designed to his an effective “GRP” goal. Here an Advertiser willcombine the online reach datawith reach data for othermedia channels such asTV tocalculate,projectandforecastsalesoffline.

UniqueReach-CreativeOptimization

• Toperformacreativeoptimizationonuniquereach(showadifferentmessagetoneweyeballs)requirestheuseofretargeting(SeeChapter7)suchthatoldusers are labelled and fall into a defined audience where new users are aseparateaudience.

• IfretargetingiscombinedwithDCOitwillnotcosttheAdvertisermuchmoretocreatepersonalisedmessagingfortheretargetingaudience,improvingthecampaignROI.

UniqueReach-MediaOptimization

• The unique site overlap report will show which sites are offering the bestunique reach, an analyst running this report can have planners and buyersadjusttheirbuyingstrategyaccordingly.

• A reach and frequency report will reveal the effective frequency for sites,placementsandadssothatthefrequencycapcanbeadjustedatthePublisheradserver.Lessbudgetcouldeffectivelybetransitionedawayfromfrequencytohitahigheruniquereachgoal.

• If a reach goal needs to be very high, it may be more effective to payPublishers based on alternative payment models such as sponsorship or toutiliseachannellikeAffiliate.

• Adevicereportwillshowhighdeliverytomobiledevicesthatdonotsupportthird party cookies. These deviceswill hinder a unique reach goal since itwouldnotbepossibletodetermineifthesameusersarereceivingadscross-device.

• Analysts can also conduct an analysis on conversion tags for total “actions”andcomparethisdatatositesideanalyticsreportsonreferralinformation.IfhighvolumesoftrafficarereachingtheAdvertisersitewithoutbeingtrackedbythethirdpartyadserverandtheseusersarearrivingtothewebsitefromPublishersitesdirectly,thereisagapofusertrafficthattheadserverisnotcapturing. The analyst can report these gaps to the planners and buyers to

reachouttothesePublisherstoimplementthirdpartytrackinganddelivery.Withoutthisinplace,trueoptimizationtouniquereachgoalsbecomesmuchmorechallenging.

Targetaudience

Usecontentclassificationverificationreporting toestablish thebaselineanddistributionof the audience on a non-specific audience buy acrossmultiple publishers.The affinitybetweenaudiencesandcontentoraudienceverificationreportingspecificallywillrevealwhat audience groups are seeing ads. Once it is clear which geographies and inferreddemographicstheusersbelongto,theoptimizationcanbegin.

TargetAudience-CreativeOptimization

• Where higher cost ad formats such as Rich Media are delivering to anineffective audience, the trafficker can add audience targeting to thecampaignsetupinthethirdpartyadserverinterface.

• Fluctuations of user traffic for various demographics are known to map totimesof theday.Rather thanpayforadditionalaudience targetingdata, thetrafficker can utilise ad controls to show different creative messages atdifferenttimesoftheday.

• To lower creative costs, the Advertiser can utilise DCO and retargeting forpersonalisationaswellascreativemessagesequencing.

TargetAudience-MediaOptimization

• SinceVerification reporting forcontentandaudienceverification,will revealtheaudiencetypesthatPublishersprovide,plannersandbuyerscanusethisdatatospendlesswithPublishersthatdonotattractthetargetaudience.Thissavesmoney by circumventing the need to increase spend on a guaranteedaudiencetargetingwiththatPublisher.

• Ifplannersandbuyersinvestininventorywithapricebasedguaranteeagainsta target audience goal, such as geography, there is an optimizationopportunity.When the activity goes live, the analyst can run a verificationgeo report to catch Publishers that fail to setup the geo targeting. Theremuneration of the failure can be spent on audience targeting with aPublisherthatdoesnotbreachtheaudiencetargetingconditions.

Verification(qualitysafety)goalsandimpact

Todelivermorequalityimpressionsanduserviewability,theanalystbeginsbycomparingcampaign verification and viewability results with the benchmark report for the sameAdvertiserverticalas theAdvertiser inquestion.Alternatively itmakessense to lookatverificationreportingfrompreviouscampaignswiththesameAdvertiseracrossthesamePublishers.Thiswill set abaseline fromwhich thePublishersandCreative inusemustmeet.

Verification-CreativeOptimization

• Viewability reporting or Rich Media reporting can be used to reveal the

averageviewduration thatusers lookat ads for.The traffickercanuse thisinformation to remove ads from rotation that have low view duration. ThesameoptimizationcanbemadewithInstreamreporting.

Verification-MediaOptimization

• AnanalystcanusethebrandsafetyanddomainURLfeaturesofverificationreporting to produce a blacklist of sites that breach the brand safetyrequirements. By issuing this blacklist to the Publishers, optimizes theinventory by ensuring ads are not shown again the blacklist in future.Thisproactively negates the risk of a high cost associated to bad PR around afailureofbrandsafety.Thiscostcanaddhindrancetothecostofrunningthecampaigninthefirstplace.

• Pre-emptiveverificationutilisesa similar strategy.Theunseenbenefit is thatinventorywhichadheres toverification requirements is consideredahigherqualityandlowerqualityinventorycouldbeboughtatalowercosttosatisfyasecondaryreachgoal.

• A buyer could introduce a new cost model to pay on viewable impressionsonly. This penalises poor viewability in the marketplace and meansPublishersmaymakemoreofanefforttooptimizemediatoappearinview.Thein-viewguaranteemayinflatemediacostswhichthebuyermustconsiderworthyofthein-viewgoalbeforeproposingthecostmodeladaption.

Optimizations towards Performance Goals (ideal for Direct ResponseAdvertisers)DeterminingtheperformanceoftheadsisseenbymostAdvertisersasthemostimportantconsideration when running digital campaigns. Any trends in reporting which can beshown to boost performance are worth carefully monitoring and applying to a wideraudiencereach.Itisimportantalsofortheanalysttoconsidernotjustoptimizingtooneperformance goal.High levels of engagementmight diminish clicks or conversions forinstance, so gathering a wide set of data before presenting the argument for theoptimization to the trafficker and marketing team is a best practice. As with Deliverygoals, storing old performance information or looking at publicly available data for thepurposes of benchmarking allows expectations about performance to be set from thebeginning of the campaign. Performance goals exist for engagement, clicks, number ofconversionsandvalueofconversionsasanROI.

EngagementandInteraction

Interactionandengagement isonlygoing tobeavailableonRichMediaandvideoads.Althoughthecosttoserveformatsinthese‘channels’willbehigher,engagementcanbeavaluable indicationof performance if creative andmedia fail to instigate users to click,convertandbuyfromads.

BrandAdvertisersthatrarelylookatperformancewoulddowelltomonitorengagementlevelsinreporting.Attheleast,theycanprovideanindicationofawareness.

Toestablishabaselineforengagementandinteractionasasuccesscriteria,theanalystcan

beginbycheckingthebenchmarkandperformancereportingforengagementonpreviouscampaigns.Ataminimum,formatperformanceforengagementbycountryandAdvertiserverticalwill set an expectation before optimization begins.The goal is then to beat thebenchmarkorimproveonthepreviouscampaign.Engagement-CreativeOptimization

• Sometimescreativedevelopersanddesignersrecycleoldcreativetemplatestosavetimebutuserengagementhasanaffinitytowardsnoveltyanddifference.TheCreativeAgencycanmaketheadmoreappealingtotheuserbyusingthelatest available components in the ad serving provider supplied componentpackforFlash.Inspirationfornewexecutionscanbeattainedfromadservingprovider showcases. Some providers also offer in-person support to inspireinnovation based on the available technology. More innovative executionswillboostengagementlevels.

• Creative designers can nowutiliseHTML5 to deliver executionswith betteropportunity for engagement through motion, touch and device-specificfeatures.TheAdvertisermaywishtoconsiderpayingtheCreativeAgencytodeliveranewexecution,ifthecampaignisalreadyrunning,ifthecampaignisnotperformingwellonengagementacrosstheboard.

• InstreamVideo,runsatahighermediacostandcosttoservebutisshowntoprovidesignificantopportunityforuserengagement.Inparticulartheanalystmayhave toconsiderswitching fromVASTformats toVPAIDorDCOforInstream. These formats are emerging and will attract higher levels ofengagementduetotheirnovelty.

• The analyst can look at ad server reporting by format cross-Publisher. Thereportsmayindicatethatsomeformatsattractbetterengagementthanothers.Theanalystcaninstructthetraffickertodroplowengagementformatsfromthe placements or up-weight those with higher levels of engagement. It ispossible that if the frequency and reach levels of badlyperforming formatswere adjusted that theymight improve to even higher levels that those adsthatappeartohavehighengagementatlowfrequencyrates.

• The analyst should also consider looking at cookie level data to extractinformation about what ads and formats may have influenced levels ofengagementineventsonthepathwaytoaspecificcustominteraction.

• In additionoptimizing to a refined audience targetingmay reveal an affinitybetweenaudiencetypesandengagementlevelstopresenttherichmediaadsto those audiences that aremore inclined to engage and display ads to therest.

Engagement-MediaOptimization

• Analysts may wish to look closely at engagement reports and interactionmetrics to seewhichNetworks, Publishers, Sites, Site sections, Placementsand whole channels are driving the optimum interactions. The analyst canadvisethemediaplannersandbuyerstoadjustthemediaplanaccordingthe

resultstoimprovethecampaignperformance.

• Traffickers can place retargeting tags (see Chapter 7) on the custominteractionswithinthecreativeandthenprogrammaticbuying(seeChapter8)couldbeusedtoseekoutthosesameusersagainbybuyingmediathattheyare actively on. By extending further media cost to those users above allothers,theoptimizationisbeingundertakenonengagement.

• By introducing a cost model based on engagement, such as cost perengagement or pay out on a video durationmetric, the campaignwill self-optimizetowardsthisgoal.

TotalClicks

The analyst could begin optimizing to clicks by establishing a baseline: checking thebenchmarkandperformancereportingforclicksonpreviouscampaigns.Ataminimum,formatperformanceforclicksbycountryandAdvertiserverticalwillsetanexpectationbeforeoptimizationbegins.Previouseffectiveclickcostdata(basedonaCPM)willalsoprovideanindicationastotheexpectednumberofclicksbasedonthespendalone.

TotalClicks-CreativeOptimization

• Standardreportingwillindicatewhichadsandcreativedrovethemostclicksfromwithintheplacements.Thetraffickercanadjusttheadcontrolssuchasweightingonplacements so that ads and creative attractingmore clicks areshownmoreoften.IfdifferentcreativeisbeingshowntodifferentPublishersites,considerswappingthecreativeoverifcreativeofthesamedimensionswith different messaging is performing better elsewhere as long as thePublisher can be ruled out as a reason for the improvement. Collating allhistorical reporting together can help form a benchmark to be clear if thePublisher is responsible for the improvement. Since the adjustment to thecreativehappens in theadserver, the livecampaignwillnotbeaffectedbythechange.

• Ifcreativemessagingisnotstandardisedacrossalladsizesandonemessageisfound to be responsible for improved performance, consider having theCreativeAgencybuildadditionadswiththeoptimummessage.Thetraffickercan then put these into rotation.To cut creative costs, consider usingDCOfromtheoutsettotakeadvantageofdynamictextfields.

• Inthisprocesstheanalystcananalysetheuserpathwaytotheclickevents.Apathwaytoconversionreportmayindicatethatclicksareinfluencedbyotheradsandcreativemessages.

• Someadserversfiltermultipleclickson thesamead. Ifmultipleclicksbyasingle user can still be classed as a contribution to the goal, considerconvertingthedisplayadtorichmediaandcountingadditionalclicksusingcustominteractions.

TotalClicks-MediaOptimization

• Pulling campaign reports will indicate which sites, sections, Publishers,

Networks, channels and formats provided the most clicks. Planners andbuyers can consider revising the proposed media spend, distributing theremaining budget to better performers where possible. This may includeexcludingsomePublishersfromtheremainderofthecampaign.

• Before removing anything from theplan, the analyst can lookat pathway toconversion and cross channel reporting to ensure that any changes to themediaplanwillnotinadvertentlyaffectclickperformanceelsewhere.

• Pathway to conversion may also provide stronger case to demonstrate thatsomePublishersareaddingnovalueacrosstheboard.

• Forsearchspecificallyplannersandbuyerscouldpaymoreforthekeywordsthatdrivethemostclicks.Keywordpathanalysiscanplayaroleinshowingwhichkeywordsandcombinationsofkeywordsaredrivingclicksacrossalltrackedchannels.

• The analyst can use cost data to determine howmuch it is costing to get aclick,evenifnoneoftheactivityisrunningonaCPCmodelbycalculatingbased on the impression to click ratio (CTR). Use this to calculate aneffectiveCPC.ProposepayingPublishersonaCPCisanunlikelyaskbutitwill ensure that themedia is totallygeared towards clickperformance.UsetheeffectiveCPCtoconvincethePublishertolowerthecostofthemediasothattheimpressioncostswillyieldtheclickcostsbasedontheperformancetodate.ACPCmodelmightbepossibleacrossaffiliateactivity.

• Channelreportingmayrevealthatsomechannelshaveagreaterimpressiontoclick ratio suchas email.So longas the reachvolumesarehighenough toreach theclickgoalwith theavailable inventory,use theadservingdata todemonstratethatplannerscanspendmoreacrossthosechannels.

TotalConversionsorE-Conversions

Theanalystcanbeginoptimizingtoconversionsbyrunningconversionreportsbyformatas well as by Publisher. This will indicate where andwhen conversions are occurring.Establishingabaselineforconversionsasasuccesscriteria,canbemorechallengingforconversions, as the amalgamated conversion data in the benchmark might includeconversioneventsthatarenotcomparabletoanAdvertisersgoal.Forthisreason,previouscampaigndataismorevaluabletosetexpectations.

Audiencequalityplaysanimportantroleinconversionoptimization.ThisdoesnotalwaysmeanmakinganadjustmenttoplanneddeliverybyPublisher.Itcanmeanacquiringmoredata to provide better audience intelligence. For instance, those users with insurancepolicies up for renewal may have made this information available to a data collectioncompany.MarketerscouldusethemarketingbudgettoborrowthedatatoestablishwhichcookieIDs are more valuable to the conversion goal. An advanced data managementstrategy could achieve this using a DMP and be followed by a creative or mediaoptimization. In addition it is important to decidewhat conversions are the establishedgoal.Selectingviewbasedconversionsorclickbasedconversionscomewith theirownconsiderations.Also,withconversionoptimization,whereclick-to-landispriorityfortheAdvertiser,clickbasedconversionswouldtakepriorityandtheattributionsettingsonthe

conversiontagcanbesetuptomeetthisneed.

TotalConversions-CreativeOptimization

• Just as with clicks, standard ad server reporting will show which ads andcreativedrovethemostconversions.Thetraffickercanadjusttheadcontrolsaccordinglysothattheseadsareshownmorefrequently.Aswithclicksandengagements,theanalystcanconsiderhowotheradsandcreativeinthesamecampaigncontributedtotheconversionbeforemakinganychanges.

• ConversionsarealsoaffectedbytheAdvertisersite,soanalystscanalsofocuson theclickthroughURLtoestablish ifconversionsweremore likely if theuserarrivedonaparticularlandingpage.

• AdvancedTimetoconvertreportingwillrevealiftheconversionwindowhasbeensettooliberallyornot.Havingthetraffickermaketheadjustmenttothewindowsontheconversiontagbasedontheanalysiswillensureconversionsarenotbeingmissed.

• ThetraffickercouldalsocarveoffsomeoftheaudiencetoseeifA/Btestingexposesthatoneaudienceperformsbetteronconversioneventsthatanother.Thetraffickerscanthenworkwiththeanalysttooptimizetothisaudience.

TotalConversions-MediaOptimization

• Onceagain,theanalystcanlookatwhichsites,sections,Publishers,Networksandchannelsprovidedthemostconversionsandconsiderremovingthosethatarenotdrivingasmanyconversionsoranyatallfromthemediaplan.

• For e-conversions, it will be essential for the analyst to look further up theconversionfunnelincookieleveldataatthepathwaytoconversion.Althoughengagementmappingwilltakeintoconsiderationtheconversioninfluencebyotherfactors, theanalystshouldkeepcheckingreportingtoensurethemosteffectiveattributionmodel is in place for the optimum representation of e-conversions.

• With regards to costmodel in display, themedia planners and buyers couldagreementtopayPublishersbasedonanagreedattributionmodelasaCPA.This will ensure the Advertiser’s unique conditions for conversionoptimizationarebeingmetinthebuyingcycle.

• For the search channel, the media could be optimized by the trafficker bypayingmoreforthekeywordsthatdrivethemostconversionsandbringinginP2Cdatatoenhancethedecision.

ROI(&extendeddata)goals

TheROIisrevealedintheconversiontagextendeddata,sotheanalystwillwanttolookat cookie level data on the conversion tag.Where the extended parameters have beenpopulatedthecookieandthepathwaytoconversionarethekeytoROI.Comparingcostsformediaandtechnologywiththerevenuefigure(realcostsforthepurchaseditem)willrevealtheROI.

ROICreativeOptimization

• Theanalystcandetermine if there isa linkbetweencreative, formatand thefinalrevenue.

• ThetraffickercouldimplementaDCOcampaigntolowercreativecostsacrosstheboard,improvingROIwhateverthefinalrevenuefigureturnsouttobe.

ROIMediaOptimization

• Theanalystcouldbuildprofilingaroundthosecookiesthatspendmoremoneyand target users which have similar behaviours with a media and creativestrategythatworkedtogettheprofileduserstobuy.

• The analyst could establish a baseline for the audience based on contentverification reporting and by Publisher to ensure that planners and buyersadjustthemediaplantobuyacrossmoreaffluentaudiences.

• As with other forms of media optimization, the planners and buyers couldchoose to pay Publishers based on anROImodel. Thiswould be articulareffectiveacrossperformancechannelssuchasaffiliates.

OptimizationSample

Theaudiencesizeforanyoptimization(bothmanualandautomatic)willalsodictatethesuccessfortheoptimizationstrategy.Averysmallaudiencesizedoesnotprovideenoughdatatogiveatrueindicationofperformancewhenscaledup(becauseinasmallergroupsomeusersdonotbehaveaccordingtothegeneralpopulationbutcanswingfavouriftheconsiderationsetislow).

RecordingchangesinOptimizationstrategy

Successfulstrategiescanbekeptonrecordandpotentiallyreappliedtocampaignsunderthesameorsimilarconditionslateron.Traffickerscanrecordwhereandwhenchangestoasetuparemade,whatdatadrovethechoicetomaketheoptimization,andtheoutcomeofthoseoptimizationssothatthosethatyieldpositiveresultscanbere-appliedlateronorinafuturecampaign.

We have looked generally at how reporting data can be applied to campaigns for thepurposes of optimization but one form of optimization has proven to be consistentlyeffective at reaching somegoals. In particular, it is cyclical nature of auto optimizationthathasdrivenhighadoptionforbehaviouraltargeting,theretargetingtousersthathaveencounteredtheAdvertiserbefore.

Chapter7RetargetingandPrivacy

Chapter7–RetargetingandPrivacy• RetargetingTags

• Figure7.1

• CookiePools

• RetargetingWindows

• RetargetingforCreativeOptimization

▪ Figure7.2

▪ DynamicRetargeting

▪ Figure7.3

▪ CarouselAds

▪ Figure7.4

• PrivacyandBehaviouraltargeting

▪ TheAdChoicesProgram

▪ Figure7.5

▪ DoNotTrack

• RetargetingforMediaOptimization

Retargeting is a way to reach leads and former customers over the noise of other adsonlineandisconsideredtobethemostevolvedandeffectiveoptimizationstrategy.Usingretargeting tags allows the trafficker to plan how creativemessaging andmedia buyingwillautomaticallybeadjustediftheadserverhasencounteredornotencounteredauserbefore. ThisChapterwill explain theworkings of retargeting and how it adds an extradimension to DCO enabling a programmatic media buying strategy. In this sense,retargeting can be classed as an automatic optimization as oppose to a manualoptimization.

RetargetingTags

For retargeting to work, the trafficker needs to generate a retargeting tag from thecampaignmanagementinterfaceofthethirdpartyadserver.Retargetingtagsbehaveinavery similar fashion to conversion tags. These are small pieces of code that, like theconversiontags,aresenttothesitemanagementteamattheAdvertisersitetoimplementthetagsintoplacesonthesitewhichwouldbenefitfromcapturinguserbehaviour.Suchlocationsincludeproductpages,landingpages,pagesinthebookingorpaymentprocessand the “thankyou”or confirmationpage.Strategicallyplaced, these tags canbe set totriggerwhentheuserbehaviourisperformed.Examplesincludewhenauserprematurelycloses their browser window before completing a transaction, or demonstrating someinterestinaproductofaparticularsizeandcolour.Theflexibilityofthesetagsmeansthatthestrategyfor theAdvertiserorAgencycanbecomplex;buildinggroupsofusers that

haveexhibitedaparticularon-sitebehaviour.

Figure7.1Retargetingtagsareflexibleandcanbeaddedtomultiplelocationstobuildupgroupsofusersthathaveperformedsimilaractionsorhavearrivedonsimilarpages.

Sinceretargetingtagscanalsobepiggybackedonconversiontagstheycancapturesomeofthesameextendeddataparametersthatasalesconversiontagcancapture.Thisallowsfor retargeting strategies that only show certain ads to users that, for instance, spent acertain amount ofmoneyon a purchase.Retargeting tags can also be placed inside adsthemselves during the development and setup stages of the campaign. Such retargetingtags can load when the ad loads or when the user triggers a certain interaction orbehaviour inside the ad. SomeAdvertisers adopt a policy of ensuring that users shouldneverbeallowedtoencounteranAdvertiserwithoutbeinglabelledusingretargeting.Thismethod would ensure that future efforts to message this user would become morepersonalised.Thisisbasedonknowingthatastrategyusedinthepasttotargetthisuserhadfailedtoinciteaconversionandthatadifferentstrategyshouldbeadopted.

ItisnotuncommonforPublisherstorestricttheuseofretargetingtagsinsideadsontheirsitesandnetworks.Building such retargetinggroupseffectivelydiminishes thevalueofthe audiences that the Publishers are attracting and trying to sell against. This wasreferenced in Chapter 2 as “Publisher Leakage”. Having the creative developer andtrafficker, checking Publisher specifications before implementing a retargeting strategyintoadsdirectly,willensurethatnoexpectationsarebeingbreachedoncetheadtagcodeistrafficked.

CookiePools

RetargetingTagscanbeimplementedduringacampaignforthepurposesofoptimization,butworkbest ifplacedbefore thecampaignbegins, toensure thatuserscanbetargetedbased on this information from the outset of the campaign.When a user encounters aretargetingtagtheircookieIDisplacedintoretargetinggroupor“cookiepool”.Thispoolisawayofcategorisingthatusersothattheycanbemessagedlateron.

SomeadserversallowAdvertisers toseehowmanyusersare inaspecificcookiepool.ThisallowstheAdvertisertoknowifitisworthsettingupacampaignoradspecificallytargetingthegroup.Verysmallpoolsarenotalwaysworthconsideringbecauseonabigscale the effort to target these users later on can costmore inman-hours to set up andmanage,thanthepotentialincreaseinearningbytargetingthoseusersagain.

RetargetingWindows

Theaimofretargetingistoshowanadtothatsameuseratalaterdatewiththehopethatbyseeingthead,theywillbecontributingtothecampaigngoal.Theopportunitytobuildup a pool of cookies and show that pool targeted ads at a later date is attractive tomarketersbutislimitedbyatimeframewhichisrarelyadjustable.Mostadserversoffera90daywindow inwhich thoseusers in thepoolcanbe retargeted.After thatpoint, thecookiesareremovedfromthepool(butaslongascookiesarealwaysbeingaddedtothepool,itwillnotdiminishentirely).

RetargetingforCreativeOptimization

RetargetingstrategiesforlargeAdvertisersarerarelysimpleaffairs.Ifausercanperforma number of actions on an Advertiser site or if a web-based product catalogue isparticularly large, then re-targeting rules in the ad server need to be structured andprioritised.Usersthatgotoaproductpageandthenpurchaseandpayviaagiftvoucher,presentsomeinterestingpossibilitiesfortheAdvertiserlateron.ShouldauserencounteranadfromtheAdvertiserlater,theadcanbedesignedtousetheretargetinginformationsothatthecreativedoesnotcontainareferencetothatproductagain.

Withintheadserverinterface,aretargetingaudienceistreatedasanyotheraudiencetype.InChapter4we explored how the ad server can be setup to consider if each user is asuitablematchforanaudiencetypebeforemovingontothenext‘rule’.Thetraffickerhastheoptiontosetupadsontheplacementlevelsothatwhenauserencountersanad,thecorrespondingcreativecanbeshown.ThisoptionworksmosteffectivelywithDCO.Thetrafficker can ensure that a corresponding versionmatcheswith the user behaviour thattriggeredtheretargetingtag.

Forinstanceifauservisitsthebasketpagebutneverreachestheconfirmationpage,tworetargetingtagscanbesetup,oneoneachpage.Thetraffickercansetupatargetaudienceintheontheplacementlevelthatdeclaresthatifauserisnotinthecookiepoolfortheconfirmationpageretargetingtag(butisintheoneforthebasketpage)thentheversiondisplayedshouldhaveamessagerelatedtotheuserabandoningtoshoppingprocess.Tomakethecreativemessagemorerelevant to theuser, itcouldbeadjustedtospecificallyreference the items that were not purchased. This requires a more complex form ofretargetingcalled“DynamicRetargeting”.

Figure7.2DynamicretargetingtagsallowAdvertisertopassdynamicvariablesintotheretargeting tag so that the variables can be used in the version or ad later on. In thisexampletheSKUnumber(auniqueproductidentifier)ispassedintheretargetingtagastheusercookieIDisplacedintotheretargetingpool.WhenthesameusernavigatestothePublisher page, the displayed ad or version is targeted based on the SKU number,featuringthesameproductagainordisplaysanadwhichcorrespondstotheproduct.Thedecisionismadebasedontherulessetupfortargetingbythetrafficker.

DynamicRetargeting

Retargetingtagshavetheabilitytocaptureauniqueidentifier,anexampleofwhichwould be a product ID. A script can be implemented by the Advertiser sitemanagementteam,allowingthepagethattheretargetingtagissittingon,topassavalueintothetag.AsauserarrivesontheAdvertiserspage(suchasane-commercepage featuring a jacket) the retargeting tag on that page receives the product IDpassedinviathescript.Theretargetingtagthenitsendsacallbacktoadservertoadd the user to the retargeting pool specific to the dynamic retargeting tagwhilewritingthespecificvariabletotheuser’scookie.Thisway,theuserbecomeslinkedtothatvariableandtheIDcanberecoveredwhentheAdvertisercomestoshowthatuseranadlateron.

Theuserisnowassociatedwithallotherusersthatencounterthissametagbutthedifferencewithadynamicretargetingtagisthedeploymentofthetag.AdynamicretargetingtagcanbedeployedtoallproductpagesontheAdvertiserecommercesite as a single piece of code. The unique identifier passed into the tag is storedagainsttheuser’sthirdpartycookie.

WhenanAdvertisercomestorunDCOactivitytheadservinginterfaceallowsthetrafficker to request that the version of the ad to be displayed depends on theproduct ID fed into the selected dynamic retargeting tag. Having connected theversions to a product feed for mass versioning, where all the product IDs aredisplayed, the result is that the user is per shown a version displaying the sameproduct again.When theuser encounters a placement, the code checks theuser’scookie, passing the dynamic variable through to the ad server. At the placementlevel, therulessetupbythetraffickerchecktoseeif theuseriswithinthecookiepoolforthedynamicretargetingtag.Iftheyare,theadservermatchesthedynamicvariable from theuser’scookiewithacorrespondingadversionanddisplays thisversiontotheuser.

Figure 7.3Dynamic retargeting connected with DCO. The versions of the ads for thecampaignarebuiltoutbasedontheAdvertiserproductXML.ThissameXMLisusedtopopulate the product pages of the Advertiser site. The same SKU identifier is used tomatch the pagewith the version. The user enters at Advertiser site at (1). The productSKU number is passed into the dynamic retargeting tag (R). The dynamic retargetingassociates this value with the user’s cookieID and passes the information into the adserver (2).Theuser thennavigates to thePublisher page (3) theplacement carries theuserscookieIDtotheadserver(4)wherethetargetingrulesseethattheuser’scookieIDis associated with an SKU number in the retargeting pool. The version with thecorrespondingSKUnumberisdisplayedonthepage(5).

Tomake thingsmore interesting, the traffickerandplanning teamcanplanaheadandchooseforthedynamictextfieldintheversiontodisplayacheaperofferthanthepriceshownonthesitefortheproduct.ThedynamicclickthroughURLwithintheDCOversioncanbeusedtocarryatokentotheAdvertiserpagethatcouldbeusedtodynamicallychangethephysicalpriceofthejacket.Anotherstrategyisnottodisplaythesameproduct,buttodisplayacomplimentaryproductasaversionofthead.

CarouselAds

Dynamiccreativecanbe takenastep furtherbynot justcapturing the last itemauserwasexposedto,buttostoreseveralitemsorvariablesintheuser’scookie.Thiscan correspond to a historical recordof the user’s journeyon theAdvertiser site,capturingalloftheproductID’sencounteredontheway.

WhensettingupCarouselorCatalogueads, the traffickerhas theoptiontoutilisethisdata todisplayallpreviousproducts seen,oronlycertainproducts inside thedeliveredadversion.Thesecouldjustbethemostexpensiveproductsorthosewiththehighestavailablestock.

Figure7.4CarouseladskeeparecordoftheusersdynamicvariableengagementsontheusercookieID.Theusernavigates theadvertiser site (1)where thedynamic retargetingtagcapturestheSKUnumberfortheproductinquestion.Theuserthennavigatestothesecondpagewherethesamethinghappens.ThecookieIDrecordsagainstbothvariables(2).When theusernavigates to thePublisher site (3) thead server servesanadwhichdisplays both previously seen products by looking at the variables on the matchingcookieID(4)andbringingversionstogetherasanewadversiondisplayingthecarouselofproductsintheplacementonthePublishersite.

PrivacyandBehaviouraltargeting

Governmental and legal bodies theworldover are in support of internet usershaving aright to control the information that is collected about them. Strictly speaking the thirdparty ad server is collecting information only about the cookie and not about the userthemselves(theinformationisnotpersonallyidentifiable“PII”).Itistheidentifiabledatathat users are keen to eradicate and the cookie is not storing any of that information.Despitethisfact,legislationisemergingtoprotectusers,sothatanybehaviouraltargetinginformationcollectedaboutthemcanbemadeanonymous(orpseudonymous)orremovedfromdatabases completely.This relates directly to the topic of retargeting, sinceoptingoutofretargetingwillmeanthatadsnolongertoappearto‘chase’usersacrosstodifferentpages,Publishersandchannels.

TheAdChoicesProgram

Theadchoicesprogramisasystemthatallowsuserstooptoutofbehaviouraltargetingvia a small icon in the cornerof the ad.The iconcanbe implementedby the traffickerinsidethethirdpartyadserverevenduringacampaign,andtheiconanditsfunctionalitywillbeaddedtoavisuallayerontopofthecreative.Whenclickedtheuseristakentoapagewhere they can choose to opt out of retargeting altogether or by each technologyproviderornetworkcollectinguserdata.

Point solutions by Evidon and Trustee have emerged to provide more complex

managementfacilitatesoftheuseropt-outincludingstatisticsforAdvertisersspecifically.Theseservicescomeatanadditionalcost.

Figure7.5Theadchoicesprogramrequiresthatallbehaviourallytargetedadscarryanad choices icon. When clicked, the icon provides the user with information aboutretargetingandallowstheusertonavigatetotheadchoicescookieoptoutpage.Onthispage the user can select or deselect cookie level technologies such as third party adserverstooptoutof.Ifanoptoutischosen,theadserverisinformed.InfuturewhenthesameusernavigatestoaPublishersite,theadserverwillcheckthecookieIDfirsttoseeiftheuserhasoptedoutofbehaviouraltargetingbeforedecidingwhichadtodisplay.Thoseusersthathaveoptedoutwillnolongerseetargetedads.

Incidentally,optingoutofthesepoolsusingtheoptoutmethodsprovidedbysuchprograms, requires the dropping of another cookie on that users machine toremember their settings.Deleting of these cookies, removes thememory of theirchoice to opt out and so usersmust opt out again after the cookie is expired ordeleted.

DoNotTrack

Togetroundthisbrowsersareattemptingtoimplementacookie-freecatch-allforthe opt-out or opt-in of cookieswhich is called theDNT functionality (“DoNotTrack”). Discussions between legal parties, browser operators, governmentalbodies, regulatory bodies and politicians continue to wage, as consensus aroundoptinginandoptingoutatdifferentlevelsofinformedconsentismismatched.Whatisimportanttonoteisthatoptingoutofbehaviouraltargetingviaacookieoptout,will not stop ads altogether, but rather stop the occurrence of targeted (and somewouldsayrelevant)advertisingwhichseemscounter-intuitive to thebest interestsofalladvertisingstakeholdersincludingconsumers.

DNTisaprivacystandardthatissendsthroughaflagtotheadserveronthehttpcallandifitisturnedon;serversshouldn’tsetuptrackingcookiesforthespecifichttpsession.Thismeansthatnocookieisdroppedonthedevice.Ifauserdisablesthe setting, the publisher ad server calls the third party ad server to add aparametertoallowcookiedropping.Iftheuserenablesthesetting,thethirdpartyiscalledtodisablecookiedropping.Thiswillalsodisablepiggybackedtechnologiesfromplacingtags.

More concerning for Agencies and Advertiser is that disabling cookies willeliminate ability to do attribution reporting and count unique audiences for the

purposesofde-duplication.Whilegivingtheusersachoicetooptoutofbehavioraltargetingandcookiesmayseemalarmingtosomemarketers,whengivenachoice,fewuserschoosetodisablecookiesaltogether,accordingtoIABresearch.However,some browsers are choosing to ignore consumer choice and are instead force thedisabling of some cookies, in particular third party cookies that would allow forretargeting. This directly damages the ability for third party systems to operateeffectivelyandfairlyandshouldbeaddressedwithconcernbygovernmentalbodiesagainstthebrowser’sabilitytoover-rideconsumerchoice.

RetargetingforMediaOptimization

Retargetingtechnologycanalsobeusedformediaoptimization,literallybuyingaudiencesbased on retargeting tag encounters. This is enabled using DSP and programmatictechnology in the ad exchange space covered in full detail in the Chapter 8 -Programmatic.Atahighlevel, theDSPtechnologyprovidersoffertheirownretargetingtag technology.This tagcanbepiggybackedalongsideanyotherretargeting tagsuchasinside an ad server container tag. The user encounters the retargeting tag from bothprovidersandassuchisaddedtotheretargetingpool.WithintheDSP,theretargetingpoolcanbeusedtoidentifypotentialimpressionsastheybecomeavailableontheadexchange.Ifthereisamatch,thentheAdvertiserorAgencycanbidontheimpressionthroughtheDSP.Oncetheimpressioniswon,theadserveriscalled,whereuponthetraffickersetupoftheadcheckstoseeiftheuserisinaretargetingpoolfromacreativeperspective.Theywillappeartobesoastheusertriggeredbothoftheretargetingtagsandtheuserwillbeshown the corresponding creative. Tech providers such as Criteo have made a verylucrative business out of this offering but tying together the creative and mediaoptimization capability of retargeting to buy and optimize against retargeted users atenormousscale.

This chapter has quietly offered a sneak peek into the topic of the next Chapter -Programmatic. Which, largely due to the success of retargeting campaigns, is seeingphenomenal growth and represents a true evolution in display advertising. Here, smartdecisionenginesbegintoreplacethemanualworkofoptimizationandbuyingstrategies.This replaces long operationalman hours by agency andmarketing teams to date so isbeingwelcomedasgame-changingbyindustrystakeholders.

Chapter8Programmatic

Chapter8–Programmatic• AdvertisingExchanges

• SupplySidePlatforms(SSPs)forthePublishers

• DemandSidePlatforms(DSPs)fortheAdvertiser

• Figure8.1

• ConnectingtheadservertotheDSP

• Figure8.2

• QualityInventorySources

• TheSecondHighestBidAuction

• RealTimeBidding(RTB)

• CostReportingintheadserver

• DSPStrategies

o DSPRetargetingwithDCO

o Figure8.3

o DSPKeywordTargeting

o NegativeRetargeting

o DSPdeliverycontrols

o BuildingPrivateExchanges

o Figure8.4

o StrategiesbasedonPublisherAudienceDescriptors

o CookieLevelDataProviders

o DataManagementPlatforms(DMP)

o Figure8.5

o PageLevelDataProviders

• Channel-SpecificComplicationsinRTB

• DSP-adserveroftheFuture?

Programmaticdescribesa fleetof technologieswith thevision toautomate theplanningandbuyingprocessesofdigitalcampaignsetupusingthepowerofreportingandaudiencedata. These technologies are still in their relative infancy and are evolving but theinfrastructuretheycreateisveryrealandrigid,foundedontheprinciplesofthefinancialexchanges.ThisChapterwillprovideanoverview to thedifferent technology types thatexistwithinthisspaceandexplainhowadserving‘plugsin’totheprogrammaticstack.

AdvertisingExchanges

In the traditional model of buying ad space, the Advertiser or Media Agency wouldcontactthePublisherdirectlyandmakeadealtopurchaseinventoryfortheproceedingadcampaign.ThiswasfollowedbytheemergenceofsaleshousesorNetworksthatworkedonbehalfof somePublishers, to sell the space toAdvertisers.Networksaggregated theinventorytogethersothattheycouldbenefitfromawiderpoolofavailableadspaceforthepurposesofdelivery fulfilment.On the relationship front,Advertisers andAgencieshadrelationshipswithbothNetworksandthePublishersthatsatoutsideofthem.

NetworkandPublishernegotiationspersontopersonareaslow,timeintensiveprocessesrelyingheavilyon relationships and to an extent; industrypolitics.Byhavingmachinesmakethenegotiations,theprocessofbuyingandsellinginventorybecomesmuchfaster,more efficient and more cost effective. It is now possible for Publishers to load theiravailableinventoryintoamarketplacecalledan“AdvertisingExchange”andAdvertisersandAgenciescomeintothemarketplacetobuyandbidontheavailablespace.

Thereisnotonemajorexchange,ratherafewthathavebeensetupandsomeexchangesare channel-specific so that exchanges for video inventory exist such as Tubemogulseparate toexchangesformobile inventory.Therearealsobroaderexchanges thatexist,attemptingtobeestablishedasthemainmarketplace.MajorNetworkshavetransformedtheir businesses from Network into an Exchange. This has given rise to theMicrosoftExchange, Google’s DoubleClick Exchange, Right Media (Yahoo’s offering) andmarketplacesfrommoretraditionalNetworkssuchasHiMedia.

SupplySidePlatforms(SSPs)forthePublishers

SupplySidePlatforms(SSPs)havebeendevelopedtobringPublisherinventoryintothemarketplaceassoonas itbecomesavailable.SSPstaketheavailable inventoryfromthePublisherandclassifyor label it inorder tosell it.SinceAdvertisersare lookingtobuymedia space based on a target audience, the SSP gathers this information from thePublisherasadatabasethatmapspagesandplacementsonthePublishersidetoaudiencetypesanddemographics.MajorSSPsincludeGoogle’sAdMeldandtheRubiconProject.

DemandSidePlatforms(DSPs)fortheAdvertiser

Advertisers and Agencies also have a gateway into the marketplace. This gateway isanother technology called aDSP (Demand Side Platform). Heavyweight players in theDSP environment include Turn,Mediamath, Appnexus andGoogle’s Invite Media andtherearemoreemergingofferingstoAdvertiserssuchasDataXuandX+1.Thesesystemsare operated by what are referred to in the industry as “trading specialists” and theenvironment of buying and selling in the ad exchanges is termed “trading”. Tradingspecialists will typically work either directly on the Advertiser marketing team; at theMediaAgencyorforanindependentoutsourcecompany.AsexplainedinChapter2,theseteams are referred to as “trading desks” and each trading deskmaywork to access theinventory in the exchange via one ormoreDSPs.Those that sitwithin theAgency arereferred to as ATD’s (Agency TradingDesks) and those that are outsourced are calledITD’s (Independent Trading Desks). By working continuously in the programmaticenvironment, trading teams canmaintain a history of programmatic buying activity andcostinformation,muchliketheplannersandbuyerswoulddowhenbuyingdigitalmedia

non-programmaticallywiththeassistanceofaplanningtool.

UsingaDSPistheoverheadcostofaccessingwhatisreferredtoastheOpenExchange.Inotherwords, theDSP is theAdvertiser’s ticket into themarketplace fromwhere theyarefreetobrowseandbuyalltheinventoryonoffer.

TheDSPtechnologyisverysimilarinitsworkingtoaSearchEngineMarketingplatform(SEM)suchasGoogle’sAdwords.TheDSPoperatorhasarecordoftheavailablemediabudgetforthecampaignandsetsgoalstobuytargetaudienceswithinacertainperiodoftime.Thecombinationofthesevariablesiscalleda“strategy”andthetraderwillsetupastrategyandputitlivetoallowtheDSPtoenterthemarketplacetoseekoutinventorythatmatches the variables. To set a strategy live requires the trading team to connect thevariables to an ad so that if the correct conditions aremet; 1) the Advertiser pays thePublisherthroughthemarketplace2)theadisreleasedfromtheDSPtotheSSP,where3)itispresentedtotheuseronthePublisherpageviathePublisheradserver.

Figure8.1Programmaticsimplified.Theopenexchange isamarketplacewheresellerspresenttheavailableadspaceviatheSupplySidePlatformandbuyersexamineandpayforthespaceviatheDemandSidePlatform.Whenthedealismet,theDSPprovidestheadtotheSSPfromtheadtagprovidedbythethirdpartyadserver.

ConnectingtheadservertotheDSP

To have this work requires that the trafficker, working in the third party ad servercampaignmanagement interface, set up a placement against a site entitywhich has thesamenameastheDSPorthetradingdesk.Forexample,ifthemediaplanreceivedbythetraffickerfromtheMediaAgencystatedononeof therowsthatanadtagneededtobetraffickedfortheGroupM tradingdesk–Xaxis; thetraffickerwouldsetupthecampaignwithintheadserverinterfacesuchthatthesitewasXaxis.ThetraffickerwouldthenassigntherightplacementandadtothatsiteandtraffictheadtagstotheXaxisteam.TheXaxistradingteamwouldthenloadtheadtagintotheDSPinterfaceandtheDSPwouldconnecttotheopenexchangetofindasuitableadspot.

Figure8.2Theprogrammaticspace fits into thebuyingandsellingprocessbetweentheAdvertiser and the Publisher. The direction of the arrows in this image represents thedirectionoftheadandofthemediabudget.Greyedoutisthetraditionalbuyingmodel:MediaAgencyorAdvertiserbuys the spacedirectly from theNetworkorPublisherandreservestheinventory.Thecreativeagencythenloadsthecreativeintotheadserver,thetagsarepassedtothenetworkandtheadisservedattherighttime.Withprogrammaticin the picture the ad is only passed to thePublisherwhen the agreement or strategy isfulfilledandtheprocessismanagedbythetradingdesk.

QualityInventorySources

DSPswilltypicallyconnecttomultipleexchangesinordertofulfilastrategybutthisisnotadefaultsettingandsomethinganAdvertisershouldcheckwhenusingatradingteamorDSPforthefirsttime.Thevariousmarketplacescarryinventoryofdifferingquality.

The ideaof the advertising exchange emerged to try and sell off all of the available adspace.AdvertisersandAgencieswouldtraditionallyusethePublishersalesteamstofindqualityorpremiuminventorytodisplaytheiradwithin.Theywouldbewillingtopayahighercost for thespace. Insuchanenvironment,bottom-of-the-barrelor“remnant”adinventorywentunsold.

The ad exchange provides an environment to sell ad space based on the price thatAdvertisersarewillingtobuyitfortoreachtheirgoals.Thishasledtosomeexchangescarryingmorevaluablead spotswhile somearehappy tocarry the remnant.ThemajordebateofcontentionarisesinthisspacewherethedefinitionofavailablequalityinventorymustbeveryconciseandclearsothatthegoalsaremetandtherightpriceispaidwithoutgivingeitherthePublisherortheAdvertiserabaddeal.

TheSecondHighestBidAuction

ProgrammaticcanworkbyhavinganSSPpresentadspace,availableinthefutureintheexchange. This inventory can be presented inmultiple exchanges so as to increase thenumberofDSPs that can see it.Once the inventory is presented, theDSP strategywillidentifytheinventoryasapplicabletothedefinedvariablesandwillmakeanofferforthead space.OtherDSPs are invited to do the same on behalf of otherAdvertiserswhosestrategymightalsobeagoodfitfortheavailableinventory.Anauctionthenbeginsforaset period of time in which DSPs make higher proposals for the ad space. When theauction ends, the second-highest bidderwill be awarded the space andbe asked to payaround a centmore than the bidding price agreed. Thismodel is used so that a singleTradingDeskcannotcontroltheauctionbybuyingalloftheinventorythatmatchestheirstrategy. If the highest bidder was allowed to win the auction, the biggest spendingAdvertiserswouldbeabletopickoffthebestqualityinventory.

Thefinalagreedpricepays the tradingdesk team, theuseof theDSP, theaccess to the

exchangeandthePublisherfeefortheadspacewhichwouldincludecostoftheuseofthetechnologyonthePublishersidesuchastheSSPandthePublisheradserver.Ontopofthis,theAdvertiserneedstoalsoconsiderthecostofad-serving.

RealTimeBidding(RTB)

ThePublisherdoesnothavetoloadvolumesofavailableinventoryviatheSSPatagiventime, but rather an available ad spot can be presented into the exchange as soon as itbecomes available. This is completely transformative in digital advertising because itmeansthatwhenauserarrivesonaPublisherpage,theavailableadspotcanloadatagbelongingtotheSSP,whichchannelstheavailabilitydowntotheexchangeinrealtime.This opportunity is literally to present one ad, to one user, on one placement, on onePublisherinrealtime.

Essentiallythisisayetun-fulfilledimpressionandasitloadsintheexchange,theDSPsleap on it like fresh meat in a hungry piranha tank. The auction begins and ends inmicroseconds,thespotispaidforandtheDSPconnectstheadserveradtagtotheSSPsothat the ad arrives on the page as it loads. The user notices no difference between thismethodandanyothermethodofaddelivery.ThisisreferredtoasRealTimeBiddingorRTB.

ToworkmosteffectivelytheDSPstrategymustalreadybesetliveandtheSSPtagsmustbesittingonthedormantpagesonthePublisherwebsite.TheseSSPtagscanbedesignedtocarryextrainformationaboutboththeuserandthepagedowntotheexchangeforrealtimebidding.IfauserisloggedintothePublisherpage,thesedetailscouldbefunnelleddown alongwith page-level contextual information or theURLof the page. If the userinformationisfunnelleddown,itneedstohavehadtheusersconsentbeforetheadspaceisofferedouttoathirdpartytoadheretoprivacycompliance.Suchcompliancehowever,mayhavebeengrantedwhentheusersignsuptogetaccesstothePublishersite.

Costreportingintheadserver

Apartfromthedeliveryofadsfromtheadservertothepage,theDSPcanfacilitatetheflowofreportingdatafromtheexchangebackintoadserverreporting.Thisensuresthatcost, delivery, verification and performance information is all still captured in a singlesystem of record at the ad server. Since the ad server is counting based on a succinctdefinition of impressions, the connected cost information must be based on thismethodologyforthesakeofdataaccuracy.ThisinformationcanbechannelledusingdataandAPIintegrationsbetweentheadserverandtheDSP.AdvertisersandMediaAgenciesshouldconnectwithadservingproviderstounderstandmoreaboutthisdataflow.

DSPStrategies

WhencomparedtobuyingadspacedirectlyfromPublishersandNetworks,DSPsofferagreater access to available inventory ad strategies basedon algorithmic optimization.ADSPwillensureitisnotmakingbidsonthesameinventorytwice.Itwillalsomakebidsforinventorybasedondifferentcostmodelsandcalculateifthegoalscanbemetwiththeavailablemediabudget,adjustingthestrategyasitworksitswaythroughthebudgetsoastomeetthegoal.

Thosebidswhicharelosttocompetitorscontaininsightswhichassistinwinningbidsfor

inventorywiththesamecompetitorsinthefuture.Thisisreferredtoas“winandlossleveldata”andisaproprietarydataofferingavailabletoAdvertisersfromtheDSPtechnologyprovider.

DSPRetargetingwithDCO

AsdescribedinChapter7,theDSPstrategycanincorporateretargetinginformation.For this to work the trading team can generate a retargeting tag within the DSPinterface and send this to the trafficker working in the ad serving interface. Thetraffickercanpiggyback theDSPretargeting tagontoacontainer tagoradserverretargetingtagsittingontheAdvertisersite.Whentheusertriggerstheretargetingtagstheiradserver,thecookieIDisaddedtotheadserverretargetingpoolbuttheDSPretargetingtagalsodropsacookieontheirbrowser.Whentheuserarrivesonaprogrammatic-enabledPublisherpage,theSSPplacementcanbeusedtocapturethecookieID for the DSP and pass it down into the consideration for the yet un-purchasedimpressioninrealtimebidding.AstheDSPrecognisesthecookieID,thecorresponding strategy can activate a bid on the impression. If theDSPwins theauction,theadserveradtagisdelivered,firstcheckingtheadservercookieIDontheuser’smachine.SincetherewillbearecordofthiscookieIDintheretargetingpooltheadserverwillservetheadorversionthatcorrespondstotheuser.

Figure 8.3 Dynamic retargeting with DCO purchased with programmatic. The userarrivesontheAdvertisersite,theversionsintheadserverasbuiltoutonmassversioningfrom theAdvertiserproductXML.The sameXMLupdates theAdvertiserproductpage.TheuniqueidentifierfortheproductasanSKUnumberispassedintotheretargetingtagfor the ad server (R) as is associated with the cookieID on the ad server (2). Theretargeting pixel for the DSP is piggybacked on the ad server retargeting pixel. Thisinformationat thecookielevel for theDSPcookieispassedontotheDSP(3).Theuserthen navigates to a Publisher page programmatically connected to an AdvertisingExchangeviaanSSP(4).TheplacementloadsintotheadexchangeviatheSSP(5)and(6).TheDSPseesthatthecookieIDmeetstheretargetingstrategy(7)andbidsontheasyet unserved impression. Winning the auction, the DSP passes the ad tag onto thePublisher page. The ad tag reads the users ad server cookieID (8) and sees that it ismatched to a dynamic variable as an SKU number. The ad server selects thecorrespondingadversion(9)andservesitintothePublisherplacement(10).

Such strategies are employed by technology providers like Criteo and Struq whospecialiseinthisspecifickindofaudiencechasing.Unfortunately,withsaturatedaudiencechasing, such technologieswill alwayswin the conversion under the traditional last adattribution model. Even if the user may have had every intention of buying the itemwithout clicking such an ad, the convenience of the placement acts as a shortcut toAdvertiser site. Marketers should think carefully about awarding full conversionattributiontosuchPublisherinventory.

DSPKeywordTargeting

An alternative strategy to retargeting would be that when the bid is won that the SSPpasses a keyword value from the page, through to the ad server via the DSP. Aswithkeyword targeting the ad server can use this value to display a corresponding ad or adversion.

NegativeRetargeting

TheDSPgives theoption tooffergreateraudience reachbyensuringusersare targetedthathaveneverseenanadorlandedonanAdvertisersitebefore.Thestrategysetupinthe

DSPcanstatethatiftheuserisnotinaretargetingpool,thattheyshouldbethetargetofthestrategy.Thisisenabledviathepresenceofaspecialistretargetingtechnologycalledan“exclusionpixel”.ExclusionpixelscanbegeneratedfromanyoftheDSPsandworkusingtheoppositemethodologytoretargetingpixels.

DSPdeliverycontrols

PopularfeaturesformerlyreservedforuseinPublisheradserversareavailableintheDSPsuchasgeo-targeting,strategyprioritisationandfrequencycapping.ThesettingsgetquitegranularsothattheDSPcaneffectivelydistributethemarketingbudget.

BuildingPrivateExchanges

AdvertisersandAgencies thathavemoremoney tospendand like the ideaof real timebiddingbutwouldpreferaccesstomorepremiumInventory,canworkwiththeDSPstosetuprestrictedor“PrivateExchanges”.Theseareinvitation-onlymarketplacesdesignedto giveAdvertiserswithmoremoney to spend, a “first look” at the available premiuminventory before it is offered out on a public exchange. Alternatively a set amount ofinventory can be procured at an agreed price between theAdvertiser and the Publisherfirst,andthemarketplacecanactasmanagementsystemtodecidewhichadstoshowtoaudiencesbasedondecisionsaboutthestateoftheinventory.IfanAdvertiserhasmultiplebrands,thereisnocommitmenttosupplyspecificcreativetotheinventoryuntiltheverylastmomentbeforeimpressiondelivery.Thismakesforamoreefficientuseoffirstaccesstoaprivatearsenalofinventory.

Figure8.4PrivateExchangessitalongsideopenexchangesasoptionsforAdvertisersandtradingteamswithenoughbudgettosupporttheavailabilityoftheprivatemarketplace.

StrategiesbasedonPublisherAudienceDescriptors

SincePublishers,NetworksandSSPsbringtheinventorytothemarketplaceandlabelitwhen it arrives; the Advertiser-side trading teams are not making decisions about themediaspacebasedonanindependentsystemofaudienceclassification.APublishermaywellhavedressedup the inventory tomake itmoreattractive toselland theAdvertisercanendupdisplayinganadonapagewhichhaslittleornomatchtotheuserseeingthead.TradingexperienceandhistoricalreportingonadperformancewilldemonstratewhichofthesePublishersareclassifyingtheiraudienceineffectively.Butthedemandforagoodmatchevery timehasseen theemergenceofsystemsthatbringmore information to thetrade. The collection and presentation of this information can be facilitated by a thirdparty,whichcanestablishmorevaluetotheinventoryinthemarketplace.

CookieLevelDataProviders

Trading teams can buy audience data from specialist third party data providers such asVisualDNA, and BlueKai. Alternatively data is available from collectors of traditionaldigitaldatapointssuchasemailaddressesfromproviderssuchasAxiom.Theseproviders

havedatasetsaboutaudiencetypesbasedoncookielevelinformation.TheAdvertisercanusethisdatasettolookupmoreinformationabouttheuserofanun-purchasedimpressionbeforethedecisiontobidontheimpressionismade.

In addition to using third party data, the Advertiser can use their own datasets fromsystemsliketheirCRMtogetabetterunderstandingabouttheun-purchasedimpression.This isreferred toasfirstpartydataandisconsideredbymanytobe themostvaluabledata that an Advertiser can collect and activate for use in the programmatic space.ProvidingthereisacookieIDattachedtothedataset,thedatacanbeused.Todothis,thedataset is synchronised with the available cookies on a user’s machine in a complexprocess called “cookie matching” and is facilitated by an intelligent data warehousesystemcalledaDMPorDataManagementPlatform.

DataManagementPlatforms(DMP)

DMPs are easily plugged into theDSPs to inform the bidding strategy and someDSPscomewiththeirownDMPfacilitates.PlayersincludeplatformssuchasAudienceScienceand eXelate. These platforms are growing in popularity as they providemore effectivedecisionmakingonthebuyingofinventoryinprogrammaticbyimprovingtheROI.DataactivatedviaaDMPmightshowthatauserisonthecuspofmakingapurchase.WiththeDMP activated, programmatic facilitates the opportunity for an Advertiser to pay arelativelyhighprice(oncedatacostsarefactoredin)togettheadinfrontoftheuser.Ifthe purchase that this triggers offsets costs then the DMP data enablement strategybecomesaparticularlylucrativeoneandthedatacanactsasacompetitiveadvantageoverotherbiddingAdvertisers in themarketplace.Examples includeproviders thatspecialiseinoptimizationsuchasRocketFuel.

Figure8.5DMPscanoffervalueinmultiplepositions in thebuyingandplanningcyclewithinprogrammatic.DatathatcanbematchedtoexistingcookieIDsattheuserleveladdmore information about the user to the Advertiser buying decision and will allow thetradingteamtoadjustthebuyingstrategyaccordingly.Withoutadditionaldatapluggedinvia the DMP, the Advertiser is almost buying blind, taking the existing audienceclassificationastruthwithoutindependentverification.

DMPs are data storage and activation facilities for data that Publishers can alsoutilise.Inbuyingorhiringthirdpartydata,thePublishercanindependentlyverifythequalityof their inventoryas it is loaded into themarketplace tomake itmoreattractivetoAdvertisers.ThecostfordoingsocanbesetintheminimumbidpricethatthePublisheriswillingtosetastheauctionbegins.

PageLevelDataProviders

Outsideof thecookie space,datacanalsobeplugged into theDSP thatdoesnotdependonuserbasedinformationbutinsteadonpagelevelinformation.Examples

includeGroupM’sCrystalSemanticsorPeer39.

AsexplainedinChapter5-thesetechnologiescanbeusedtopre-emptivelyverifythat thepageinquestionisasuitableenvironment inwhichtodisplayanad.TheDSPcan incorporate this information into thebuyingstrategy to target towardsoraway from the defined variables. DSPs will also typically enable traders to loadblacklistsofsitesaspartoftheirstrategiestoensurethatbannedsitesorpagesareneverconsideredwithinthemarketplace.

RegulatorybodiesencourageDSPs,exchangesandcompanies inprogrammatic tofacilitate the safe purchase of inventory such that universally recognised bannedsites and blacklists are adhered to as part of a crackdown against fraud andcybercrime.

Channel-SpecificComplicationsinRTB

Inthecurrentlandscape,inventorycanbeboughtforalmosteverymediachannelthroughthe exchanges. There are specific issues related to RTB forVideo andMobile.Mobilesuffersbecauseanimpressionloadedintoanexchangeinrealtimemaynotnecessarilybeabletodelivertheadbeforethemobilesignaldropsout.VideoinventorysuffersfromaproblemofsupplywherethereishighdemandforInstreamvideoadsbutlimitedsourcesofqualitycontent.Thisiswhyadserververificationandfraudreportingisinhighdemandwithin programmatic.The allure of highpricingon Instream inventoryhas led to somelong tail Publishers engaging in fraudulent activity to buryAdvertiser inventory amongpoorcontentorintospotsthatweremisrepresented.

DSP-adserveroftheFuture?

DSPs and ad server have not merged as technology providers due to the impartialityofferedbytheadserverinitstotalseparationfromtheinterestsofthePublisher.Sincetheadserver independentlydeclareshowmanyadsweretrackedordelivered, theadserverreporting shows theAdvertiser howmuch to pay the Publishers and it is of paramountimportancetokeepthesetechnologiesseparated.

Thirdpartyadserverscancontinuetoofferde-duplicationatthecookielevelandproviderichchanneladservingforDCOwhileencumberingSearch,Mobile, Instream,AffiliateandDisplay.Thereportingaloneinpathwaytoconversion,cookieleveldata,adjustmentsforcookiedeletion,post-deliveryverificationandviewableimpressionsmeansadditionalfirstpartydatafortheAdvertisertoactivateinsidetheDSP.However,thethirdpartyadserverremainsthehubofcreativeoptimizationwheretheDSPisemergingasthehubofthemediaoptimizationinthefutureofdigitaladvertising.

Chapter9TheBigPicture

Chapter9TheBigPicture• Figure9.1-DisplayandProgrammatic

• Figure9.2-LayeringinPaidSearch

• Figure9.3-LayeringinAffiliates

• Figure9.4-LayeringinUsers,AdvertiserProperties,AdsandTrafficking

• Figure9.5-LayeringinRetargeting

• Figure9.6-ThirdPartyAdServing-theDiagram

Having covered the aspects of ad serving at a high level, it is possible to piece themanagementofthechannelstogethertogetathirdpartyadserverperspectiveofdigitaladvertising. The first diagram picks up from Chapter 8 by incorporating Display andProgrammatic.

Figure9.1-DisplayandProgrammatic

TheMediaAgencyorAdvertiser canprocuremedia inventory either through thedirectbuychannel;speakingtoPublishersandnetworksdirectlytomakeadealontheadspaceorrulescanbesetupintheDSPalongwithabudgettobidoninventoryintheexchangeenvironment. The DMPs ad context to the impressions and ad spaces being purchasedwith the option to trade in real time. This samemethod can be applied to all Displaychannels (Standard Display, RichMedia, DCO, Instream video and mobile). The thirdpartyadserverplaysaroleinsupplyingthecreativeasanadtagtotheDSPordirecttoPublishertobeimplementedintothePublisheradserver.

Figure9.2-LayeringinPaidSearch

SearchhasmaturedprogrammaticallyfasterthanDisplay.TheestablishedtechnologiessitwiththeSearchenginesasSEMsystemslikeAdwordsandasbidmanagementsolutions.The third party ad server offers tracking integration to each to have eyes on thesechannels,whileofferingcrosschanneldatathroughapathwaytoconversionfeedtoadd

intelligencetothebidsonkeywords.

Figure9.3-LayeringinAffiliates

SinceaffiliatesalsoutilisesearchthereisalineintothesameSEMtools.Thethirdpartyadserveralsooffersclickbasedtrackingtotheaffiliate.Affiliatereportingisfedbackintotheadserver.

Figure9.4-LayeringinUsers,AdvertiserProperties,AdsandTrafficking

IntroducingbasicviewsontheuserperspectiveandAdvertiserperspective;TheuservisitsPublisherpages,properties,SearchEnginesoraffiliatestoutilisecontentviathebrowserand device, carrying a publicised IP address and, providing they have not opted out ofthirdpartyadvertising,athirdpartycookie.TheuserisdeliveredanadviaaplacementonthePublishersite.Theadservercountsthedeliveryandthecreativeisdisplayed.Shouldtheuser interact or click, the ad server counts the actions and redirects theuservia theclickthroughURL.Theuserarrivesat thedestinationpagesuchasa landingpage.Thelandingpageandsubsequentpagescancarryretargetingtagsandconversionorcontainertags. The conversion tag feeds the user conversion event into ad server reporting. TheanalystcanreadthereportsandprovideinsightsbasedonbehalfoftheMediaAgencyortheAdvertiser

Figure9.5-LayeringinRetargeting

Theretargetingtagcanbeusedtosegmentanaudienceorcookiepool.ThedatacanbeusedforcreativeoptimizationintheadserverorpushedtotheDemandSidePlatformasadatasettotargetaudiencesformediaoptimization.

Figure9.6-ThirdPartyAdServing-theDiagram

The full picture is significantly simplified but the established channels (bar social) arecapturedinadservingdata,allowingforde-duplication,attributingthevalueofpaidformediatothebuys.Diagramsofthisnaturedonotrepresentthetrueclickpath,theflowofcost,value,anddataorspend.Theyalsodonotrepresent thecomplexityofanindustrythatsupportshundredsofthousandsofbusinesses.Eachentitydepictedcanbemultipliedand entities intertwine where data adds value to all nodes. The image depicted on thecover of this book is an abstraction of this view, a model I have struggled to presentvisuallyinmoresimplifiedterms.

ClosingChapterThirdpartyadservingtechnologyisamarketingrevelation.

Farfrombeinga trackingandservingtechnology, thedata that the thirdpartyadserverproducesisanindustrybackedindependentrecordforAdvertiserstoestablishasecurityaroundthepurchaseofdigitalmediaspace.

Itisrelevanttodrawattentiontotheapparentparallelwhichexistsbetweenbankingandadserving.Theattitudeisoneoftrust.HousingthedatawhichtellstheAdvertiserswhatto pay the pay Publishers; this highlights the value of account-keeping across a singlemethodology with global scaling, a level of commercialisation required to support theglobalcommunicationssystemwecalltheInternet.

Takingastepbackfromthetechnicalitiesandevolvingfeatures,thefoundationlaidbythethirdpartyadserverbringseconomicalwealthtothedigitalstakeholdergroups:CreativeAgencies,MediaAgencies,AdvertisersandPublishers.Andasthistechnologydiversifiesandchanges,itbecomesevermoreimportantthatasanindustry,weremaineducatedandinformed;challengingthefoundationsofAdvertisingsothatit’sbenefitsalwaysnourishtheenduseraboveallelse.AdvertisingasawholeisnotalwaysembracedbyconsumersbutitsvalueinprovidingafreeinternetforallmustmeanthatusersandAdvertiserscanreachamutualpartnership.

Thisbookisdesignedtomakethetechnologysideofdigitaladvertisingmoreaccessibletoeveryone.Theconceptscoveredheremayseemfreshandnewbutdigitalevolvesbytheday.Toextendthevalueofadvertisingtechnologyfurtherinvolvesreachingforwardtowards thegoalofbetteradvertisingeffectiveness.AndAdvertisingrightfullymust layits claim in ameaningful and scientific way on the outcome of a purchase as a directinfluenceoverthebehaviouroftheconsumer.

Businessefficiencydemands that thegeneral artof indirectly influencingapurchasebemadefullyaccountabletoprovethatgeneraladvertisingisnotawasteofmoney.Aswehave seen, ad serving technologies andnewadditions to the scene likeDSPs andotherpoint solutions are gradually filling in the data gaps.However if advertisingwere trulyeffectivetherewouldbenodoubtastoitscapabilitiesandallprofitswouldbedivertedtothecause,togenerateevermoresalesuntilbrandsandtheirproductsandservicessaturatetheconsumerspace.

Saturationisagenuineproblem,farreachingcampaignsthatoverlaponeanotherhaveamazeofdata toprocess to realiseeffectivenessandcement trueattribution.The race totrackallchannelsreachesoutintotheTVandvideospaceasmediumsconvergeandtheuserconsumptionofadvertisingevolves innewways.AndwhereTVandOnline teamsbothmeet, thegoal remains thesame.The technology, it isperceived, ifusedcorrectly,candowhatadvertisinghasattemptedtodosinceitsadvent:correctlypredictanddriveuseractiontopurchaseindefinitely.

Today’s third party ad servers are gathering the data thatmarketers need to realise thisdreambut the information remains in silos. In theeconomiccompetitionofadvertising,thedatasitswiththeAdvertiser;theirsuccessesaretheirsecretsandthebeststoriestoldin the best performing data, feed cleverly into the sales algorithms of marketers. Ad

servingtechnologyistheenablertothisendandisthemetaldetectorrevealingthehiddentreasureofthissuccess.

For it is the long labour, cost and learningof themarketer, thatbring the fruitsof suchefforts to the possibility of such a future. But we are only at the start of the journey.Hopefully thisbook invitesgreateropportunity to thoseeager toshape thebright futurethatliesaheadfordigitaladvertising.

Thankyoufortakingthetimetoreadthisbook.OnceagainIwelcomeanyfeedbackfromyou,andIwillendeavourtoassistinanywayIcantoimprovingtheindustryforeveryonethroughmywork.

FurtherReadingandStudyThis book has been carefully compiled of ad serving and advertising technologyfundamentals. The detail contained within these pages could be classed at an‘intermediate’levelofknowledge.Includingthewholeswathofupcomingtrends,customtechnologies and start-upswould havemade this book too extensive to focus on thosefundamentals.Thisistheframeworkfromwhichtobuildnewplatforms,conceptsandseehowtofindevolvingopportunitiesinthesetechnologies.AsIwaswritingthisbookIwasstill learning new things and I have actively omitted some concepts which are moreadvanced.TrulyIcouldhavebeenwritingforeverandneverpublishedifIhadincludedeverythingIknowaboutDigitalAdvertising.

Tohelpyourlearningbeyondthefundamentalsortobetrainedonanyoftheconceptsinthisbook inpersonpleasevisitmywebsite:http://cristalconsultancygroup.comhereyouwillbeabletoattendtrainingcourses,receiveprivatetuitionandsignupforconsultationsforyourbusiness.AdvertisedtrainingcoursestendtobebasedoutofLondon,UKbutIdotravelfrequentlysogetincontactifyouwouldlikemetoarrangetrainingoverseas.

..andtostayontopoftheverycuttingedgeinOnlineAdvertisingTechnology,beyondthefundamentals,pleasesubscribetomyblog,publishedonthesamewebsite:

http://cristalconsultancygroup.com

AcknowledgementsThisbooktookmealongtimetowrite,IwasnotawareoftheenormousamountoftimeIwouldhave todedicate to finallyget itout ‘into thewild’.Although therearevery fewreferences to outside academicmaterial (mostly due to the infancy of the industry andtechnology) Iwant toverysincerely thank those thathavehelpededucatemeup to thislevelofknowledge.Thelistofnamesexceedsahundredpeople,butatriskofmissingoffanynamesIjustwanttosendahugethankyou.

Inadditiontherehavebeenwholeraftersofproofreaders,editors,minorcontributorsandsupporters thathavemyunrelenting thanks.Theeffortsweundertakeeverydayassuresthe free, public availability of the information technologywe call the internet. For thistherecanbenogreaterrewardthantokeepourindustryknowledgeuptodateandspendthe days of our lives giving birth to inspiring new ideas, building new intelligenttechnologiesandcontributingdirectlyto theevolutionofourcivilisationforgenerationsofmadmenandwomentocome.Iwishyoualleternalprosperityinheartandmind.

FinallyIwouldliketothankmyfamily,whohavealwaysbelievedinmycapabilitiesandhavesupportedmypassionforwritingandtechnologyateveryturn.

RubyHirsh,AshleyCristal,JamesCristal,NickCristal,MichelleMiller

_______________

Glossary

Term Definition

302AnHTTP code to tell the browser that the page in questionhad beenmoved. This is called a server side redirect and isusedbyadserversasamethodtocountandtrackadactivity.

thirdpartycode

Codebelonging toanexternal system to theonebeingused.Thirdpartycodeistypicallyusedinafirstpartysystem.SuchasexampleiswhereaPublisherusesaPublisheradserverandbrings third party ad server tags into the system. The tagscontainthirdpartycode.

404

AnHTTPcodetotellthebrowserthattherequestedpagehadnot been found. 404 errors typically come up in ad servingwhen a click through URL lands the user on a page thatdoesn’texist.

4thpartycode

Code belongs to another technology provider from the thirdpartyprovider.Anexampleiswherecodeisgeneratedfromasurveyvendorandprovidedtothetraffickertoimplementonanadsuchthatwhentheadclosesthesurveyislaunchedfromthefourthpartysurveyvendorsystem.

Actiontag

Anothername foraconversion tag from theAtlasadserver.Action tag might be a more accurate description sinceconversiontagsdonot justmeasureconversionsbutcanalsomeasurenon-conversionevents:alsocalled“actions”.

Actions

A conversion tag loads whenever the code is rendered by abrowser. If the user did not encounter an ad previous to theconversion the ad server does not register a conversion butmakesalogofthecodeload.ThisiscalledanActionandanActionisnottiedtoauniquecookieID.

Adassignment Theprocessundertakenbythetraffickerinsidethethirdpartyadservinginterfacetoassignadstoplacements.

Adbrokering

A Publisher, network or exchange may be selling inventorywith the claim that it falls under their ownership but in facttheyareactingasamiddleman toanotherPublishernetworkorexchangetoprocuretheinventoryatacertaincost.Bewaryof ad brokering. More middlemen mean more technologies,greater discrepancies, larger margin for delivery error andhidden costs. Verification can provide some level of controlbut the industry is rife with the practice given the ease ofpassingadtagsontoanadditionalparty.

Adclutter

APublisher that swampsweb pages in ads clutters the pageandreducesthevalueof thecontentandsubsequentlytheadspace.GetacommitmentfromthePublisher thatadswillbe

delivered to clutter-free pages and utilize advancedverification technology to detect for ineffective impressionsundertheguiseofadfraud.

Adcontrols

Thenamegiventhesettingsontheplacementwhichallowadsto be rotated, sequenced, targeted and delivered at a givenfrequencygovernedbyasetofrulessetupbythetrafficker.Ifthereisonlyoneadassignedtotheplacementthenadcontrolsarelimited.

Adlocation

OnceimplementedonaPublishersitetheadappearstoliveina fixed location on the web page when it loads into thebrowser. Although intrinsic to standard display ads, RichMediaadscanstretchawayfromtheoriginaladlocationintoanexpandedstateornothaveafixedlocationatall (suchaswith a Floating ad). The ad is essential anchored in the adlocation which from a Publisher’s perspective is sometimeswhatisreferredtoasthe“placement”.

Adservermacro

Macrosaresmalltokensortinypiecesofcodeusuallyasinglewordlongsurroundedbysymbols.Themacroisaninstructionto theadserver to loadadditiondata into thepositionof themacro.ThiswilltypicallybeanidentifiersuchasaPlacementID.

AdservingHierarchy The tree-like folder system within an ad server in whichcampaignsaresetup.

Adspot (seeadlocation)

AdTags

Once ads have been assigned to placements inside the adserver campaignmanagement interface by the trafficker, thecodecanbesenttothePublishers.ThecodeforeachspecificplacementiscalledanAdTag.

Adblocking

Therearetwotypesofadblocking.Oneisperformedlocallyon a client deviceusuallywithin thebrowser toblock an adloading.Theotheriswithinaverificationtooltopreventanaddelivery based on the ineffectiveness of the site orenvironmentinwhichitisbeingrequestedtoload.

AffiliateAuditing

Affiliates within a network or subscribed to a program arechecked regularly by the team managing the program toensurethattheAffiliatesiteisstillsuitabletobeincluded.

Affiliateprogram

An ad campaign extended across multiple Affiliate sitesusually managed by an Affiliate network. The campaigncreative and click-through for the affiliate program areavailable from the affiliate network or the programmanagerworkingattheAgencyorAdvertiser.

AffiliateScreeningAprocessundertakenbytheAffiliateprogrammanagerortheAffiliatenetwork tocheckthatanAffiliate issuitable to joinanAffiliateprogram.

Aggregatedreports

Reportingtemplatesconstructedfromadserveranalyticsandcounting data, amalgamated across ad server entities so thatreporting is a summary of activity and can be analyzed andunderstoodeasilywithoutthespecialistskillofananalyst.

API ApplicationProgrammingInterfaceATD AgencyTradingDeskBehavioralTargeting AnothernamegiventoRetargeting.

BlackBoxTechnology

Not disclosing the workings of a technology owing to thecomplexityorimplicationofdoingso.Anexamplewouldberevealing Google’s search algorithm. Google as a searchengine is a black box technology. The workings remain amystery but it is agreed that it does the job. Revealing theworkingswouldallowotherstodefraudthesystem.

BlindNetworks

To reduce the cost of inventory, planners and buyers canpurchaseinventorybasedpurelyonaudienceinformationandnot on the name of the site where the final ad will bedelivered.ThisgivesPublishersandNetworksmoreflexibilityto optimize delivery of ads across Networks based oninventory availability, rather than having to commit to thefinal location earlier on.To ensure the audiencematches theAdvertiserrequirement,verificationreportingandadblockingcanensurethepageisbrandsafeonaddelivery.

BlindNetworks

A blind network is Publisher network where inventory islabeledonlybytheaudiencetypesuchthatthefinallocationof the delivered ad is not known to the Advertiser. BlindNetworkstypicallysell inventoryatalowercostsoasnottohavetocommittotheexactpagelocationoftheadbeforeitisknownifthatspotwillbeavailable.

Booking pathway (on anAdvertisersite)

Thedesignofthesitetodriveusersfromthepointofarrival(the landing page) to the point of conversion (the thank youpage).

BrandAdvertisers

Advertisers whose main goal is deliver a creative message.Thismightbe tocreateawareness rather that todirect trafficfor the precise goal of making a sale. The sale event mayhappenalongtimeafterthebrandactivityoftheadcampaign.This Advertiser is focused on delivery goals and GRP.CommonBrandAdvertiserverticalsareFMCGandAutos.

Cache

Amachinewillstorecopiesoffilesandinformationtomaketheretrievalofthatinformationfasterifitisrequestedagain.Acacheshouldberefreshedoftensothattheuserdoesnotgetoutofdateinformation.Thisappliestoadsandwebsitesandacachecanoccurlocallyonauser’sdeviceorinthecloud.

Call-to-action Apartofanadorwebpagedesignedtobeengagedwithbytheuser.Thiscouldbeasubmitbuttonortrigger.

CDN ContentDeliveryNetwork

Channel

Achannelwithin thecontextofdigitalmarketingconsistsofany sub-channel within Digital. This also spreads itself toformat complexity such that the most popularly referencedchannels are: Standard Display, RichMedia, Search (SEM),Natural Search (SEO), Dynamic Creative (DCO), Mobile,InstreamVideo,Affiliate,SocialandEmail.

CLD CookieLevelData

CleanData

Aconsistentnamingmethodologyproduceslargesetsofdatathatcaneasilybesorted,concatenatedornormalised.This isclean data and is used in reference to utilises consistentnamingconventionsasad,campaignorconversiontagnames.

CMS ContentManagementSystem

CodeWrapper Codewhichsitseithersideof themainsubject toeffectivelywrapthesubjectup.

CompanionAds AnadthatsitswithinvisiblerangeofanInstreamvideoadtoaccompanyit.

ContainerTags

Conversiontagswhicharecapableofhavingotherthirdor4thparty tags or pixels piggybacked on top.The conversion tagtherefore acts as container to the other tags. In theDoubleClickadserverthesearereferredtoasFloodlighttags.

Contextualtargeting

Also called Semantic targeting.The placement code extractsphrasesandwordsonapagetounderstandthesubjectmatterofthepageinquestionbeforedecidingtodeliveranadbasedontheinterpretationofthatinformation.

ContractedState Anexpandableadthatisdormantorhasalreadycompleteditsexpansioneffect.

ConversionEvent

Aconversioneventisjustanotherwayofdescribingtheeventthat is the conversion.Conversion event is amore universaltermthan“conversion”whichhasaconnotationthattheeventinquestionwasacompletedsale.

Conversiontag

Generatedwithin theadserving interface, theconversion tagisapieceofcodethatcanbeplacedinanadoronAdvertisersite.Ifauserhaspreviouslybeenencounteredbytheadservertheywilltriggeraconversioneventinadserverreporting.

Conversions

A record of an event performed by the user, such as theloading of a specific page or ad which signifies thesatisfactory completion of a marketing goal. This is mostcommonlyasale.

CookieMatching

If two separate datasets both use cookies as the uniqueidentifierfortheuser,thecookiescanbematchedtogethertoseewhereauseriscommontobothdatabases.ForinstanceanAdvertiser could produce a script to site on the Advertiserpage that can read the users first party cookie ID for the

website.Thesamescriptcould take thecookieIDandpass itinto the extended data of an ad server conversion tag. Thecookie level data would reveal a match between the twocookieID’s.

CPACost Per Acquisition. A common model to pay Publishersbased on their effectiveness at driving users to convert orpurchasebasedonthesubjectofthead.

CPCCostPerClick.AcommonmodeltopayPublishersbasedontheuserclickingonthead.CPCismosteffectivelyusedinthePaidSearchchannel.

CPL CostperLead.AnothernameforCPA.Theconversioneventinquestionisthecollectionofuserdata.

CPM Cost per “mille”. Cost per thousand impressions measured.Themostcommoncostmodelindigitaladvertising.

CPQ Costperquote.AnothernameforCPA.Crawlers (Search engineCrawlers) See“Spiders”

CreativeAgencyAn external provider of design and development services tobuildcodeanddeliverthecreativewhicheventuallybecomesthead.

CreativeFatigue

Mostcommonlyafailuretoimplementfrequencycappingsothatthecreativemessageisshownsooftenthattheusersighsindisappointmentuponseeingtheadagainandagainwithnocontroltostopitsrepeatappearancearoundtheweb.

Creativeshop

AnumbrellatermtomeanCreativeWorkshopandwillmeanaCreative Agency or an internal creative team at a MediaAgency.TheCreativeShop is home to the creative designeranddeveloperroles.

Creativeteam

A reference to the creativedeveloper anddesigner roles thatcansitwithinaMediaAgencyorataCreativeAgency.Thecreativeteamisresponsibleforthedesign,development,buildanddeliveryoftheadcreativeandmayalsobeinvolvedinthedesignandmaintenanceoftheAdvertiserwebsite.

CRM ClientRelationshipManagement system.An examplewouldbeSalesforceorMicrosoftDynamics.

CTR Click-ThroughRate.Apercentageofthenumberofuserthatsaw(incrementedanimpression)andthenclickedonthead.

CVTools ContentVerificationTools.

DataLeakage

Publisher audiences can be harvested using retargeting tags.The harvesting of these details for further use later on isconsideredaleakofthatPublishersaudience.TheclaimmaybethattheaudienceisthepropertyofthePublisher.

DCO DynamicCreativeOptimization.Always making sure that the entire audience for all ads

De-duplication

belongingtoanAdvertiseristrackedinasinglesystemsuchthat users can be recognized as unique and their uniquepathways to events such as conversions can be described intheadserverwithoutanyquestion thatsomedatabelongs tothesameusermorethanonce.

Delivery

Areferencetoadsbeingdeliveredbytheadserver.Adeliveryismeasured in impressions so delivery relates to impressionbasedgoalsanddata.Deliveryofadsonlinecanbecomparedto TV ads being delivered to TV viewers. The ad is simplysenttotheendrecipientwithnoconfirmationthatitwaseverseen.

DesktopDeviceAdesktopdeviceisanamegiventoacomputerthatremainsina fixedpositionsuchasaPC.A tablet isnotclassedasadesktopdevicebutalaptopmayfallintothiscategory.

DeviceFingerprinting

For mobile advertising, devices do not carry a universaluniqueidentifiersuchasathirdpartycookieineverycase.Adevicefingerprintisanidentifierusedinplaceofacookieforthe purposes of attribution and themeasure of unique reachandfrequency.

Direct ResponseAdvertisers

Advertisers more focused on driving conversion events orsales.ThisAdvertiserisfocusedonperformancechannelslikeSearchandAffiliateaswellasperformancegoals likeclicksandROI.Commondirect response verticals are online retailandtravel.

DisplayAds Also called Standard Display ads. These are the mostpredominantvisualadsasflashseenacrosstheweb.

Domain Themainfolderinwhichawebsiteresides.Thedomainsitsatthetoplevelandsubdomainssitwithinit.

DPB Datapassback-theAtlasadserverversionofCookieLevelData.

Dropoffrate:Thediminishingpercentageofusersarrivingtopages.Thosethatarriveonthelandingpagebutfailtoconvertonthethankyoupagearesaidtohavedroppedoutofthebookingpathway.

DSP DemandSidePlatform

E2C Exposure to conversion reporting - the DoubleClickequivalentofP2C.SeeP2Creporting.

E-conversion

Dividingtheattributedconversionamongtouch-pointsonthepathway toconversionallowseach touch-point tobevalued.ThisassignedvalueisrarelyawholenumberandisreferredintheAtlasadserverasane-conversion.

EffectiveImpressions Impressions that met the Advertiser criteria for qualityverification.

Entities

Adserverentitiesarenamedpartsof theadserverhierarchythat will typically have a corresponding entity ID such as

Advertiser, Placement, Site, Channel, Ad, Creative andVersion.

Events in thepathway (toconversion)

Also called “touch points”.User see, click and interactwithads for an Advertiser before finally converting. These adengagementsare referred toasnon-conversioneventson thepathwaytoconversion.

ExclusionPixelA retargeting pixel specifically used to negatively retargetagainstthecookiepoolitcollects.PredominantlyusedbytheDSP.

ExecutionThe Creative is sometimes referred to as a “creativeexecution”orjustan“execution”.Anotherwayofsayingthefinaldeliveredad.

Expandedstate

A Rich media expandable ad, at its most full dimensionsmaintained for a period of time or until a user engages toreducethesize,changingthestateoftheadtothecontractedstate.

Experiment (forusertargetingtestinglikeABtesting);

FactorsReportA report that can recalculate attributed conversionsbasedondifferentattributionmodelsfromtheoneusedtocalculatetheconversionsinthefirstinstance.

Fileweight Thesizeofthefile.Foradsthisismeasuredinkilobytes(kb)andmegabytes(mb).

FlashCookies

AstoragefacilitywithinFlashtorememberinformationabouttheuserandtheirrelationshipwiththeadfortheinformationto be used later on. Flash cookies are not favored inDigitaladvertisingbecauseflashcookiesarehiddenfromtheuserasadefaultandaremoredifficulttoremovefromauser’sdevice.

Floodlighttags DoubleClick’snameforacontainertag.Seecontainertags.

FLV Flash Video. A reference to the file type that creativedeveloperssupplytoplayvideowithinanad.

Frequency Thenumberof times that auniqueuser is exposed toanad.Canbecappedbyfrequencycaps.

FTP

File transfer protocol. Referencedwith regards to large datasets that can be posted to an Advertiser’s FTP site for thepurposes of large file delivery and storage from an externalsystemsuchasthethirdpartyadserverreportingsystem.

GDN TheGoogleDisplayNetwork

GeoPerformance

Also calledGeoReporting. The name given to a report thatshowsuserIPaddressmappedtoalocationfromcountryleveldown to postal and zip codes in some regions with varyingaccuracy. Geo performance is used to ensure Publishers arecorrectlytargetingadstousersintheagreedgeography.Gross Rating Point. A measure from TV advertising toextrapolateeffectivereachandfrequency.AGRPnumbercan

GRP be so refined and calculated so as to accurately point at agivennumberofin-storesalesifthegoalisreached.

Hierarchy

Thetreelikefolderstructurehousingtheadsintheadserverinterface. Advertisers, campaigns/media plans, networks,Publishers,sites,sections,packages,

HTML5 The latest edition of the mark-up language HTML withimprovementstosupportvideoandJavaScript.

IAB InternetAdvertisingBureau

Impulseconversions Auser conversionwhichoccurswithin a single ad exposureandwithinashorttimeframe.

In-App Adsthatappearwithinmobileapplications.

Ineffectiveimpressions Impressions that failed to meet the Advertiser criteria forqualityverification.

Inventory Theamountofadvertisingspaceavailable.ISP InternetServiceProviderITD IndependentTradingDesk

JumpTag ADFAclickcommandthatcanbeusedtocarrytheusertoanover-ridepageURL.

KPI KeyPointIndicator.Amarketingtermusedtodescribeagoalsetbythemarketingteamseenasapriorityforsuccess.

MediaBuy Thepurchasedadspace.Media Plan (ad- serverhierarchyentity)

Within the thirdpartyadserving interface, themediaplan isthenamegiventhecampaignlevel.

Media Plan (Agencydocument) Thedocumentationthatlistsallofthemediabuys.

NegativeRetargeting SeeExclusionPixel.

Optimization Using the reporting data to make decisions to improvecampaignperformance,deliveryandverification.

OrderIDThe unique identifier for the conversion event generated bythe Advertiser and passed into a conversion tag on theAdvertiserpage.

P2Creporting.

Apathwaytoconversionreport.TypicallyacookieleveldatafeedshowinghoweachcookieIDencounterseventsortouch-points as ads before reaching a conversion tag withtimestamps.AP2C report allows theAdvertiser to constructattribution models. P2C is the acronym given to theDG/Mediamindcookielevelpathwaytoconversionreport.

Parallax

A creative effect common to eastern cartoon wherebackground and foreground move at different speeds. Theeffect requires a high capacity of user RAM and so is onlybeingadoptedrecently.

PathwaytoConversion See“P2CReporting”.

Payload The payload is the content of a tokenwhich could be an adservermacro.Forexample:“?rtu=payload”.

PCP

A Publisher custom parameter. A token which can bepopulated in the ad call or click on the Publisher side topopulateadserverreportingwithmoreinformationsuchasaplayerID.

Performance

A reference to non-delivery data about the relationshipbetween ads and users received and processed by the adserver. Performance is measured in clicks, conversions andengagements/interactionssoperformancerelates togoalsanddata based on these metrics. Performance is a feedbackmechanism, allowing the analyst to determine if adseffectively met goals related clicks, conversions orengagement.

PII PersonallyIdentifiableInformation

Placement

Justlikewiththeword“tag”,theword“placement”hasdualmeanings. From the perspective of an Advertiser and thirdpartyadserver,placementmeansthecodepackagethatissentto thePublisherandfromwhich theadsorcreativecalled toserve fromwithin it, areplayed.Placement in thiscontext istheportablehousinginwhichtheadsinrotationorsequencereside.FromaPublisherperspective“placement”referstothefixedadspotonapagewhichnevermoves.TraffickerssupplytothePublisheradoperationsteamcodeasan“adtag”whichwhen inputted into the Publisher ad server are eventuallyservedintothePublisher’s‘placements’allaroundthePublisher’s‘placements’allaroundthePublishersite.

Pointsolutions

Specialist technology solutions typically with advancedfeatures in the field of their expertise. Utilizing a pointsolutioninadditiontoanadserverwillincurhighercostsandintegrations between point solutions and the ad server areparamounttoensuringcleandata.Manysuchtechnologiesarestart-ups.Questiontheabilitytoscaletomultiplemarkets,andtheextentoftheavailablesupport.

POS PointofSalesystem.

PostClickConversion Aconversionwheretheattributionmodelinplaceawardsthisconversiontotheuserclicking.

Post ImpressionConversion

Aconversionwheretheattributionmodelinplaceawardsthisconversiontotheuserbeingexposedtoanad.

PPC PayPerClick.AnothernameforCPC

Pre-emptiveVerification

Technology which checks the yet-un-purchased userimpressionatthepageleveltodetermineiftheenvironmentissuitablefortheAdvertiserusingtheDSPtoproceedwiththe

auction.

PremiumInventory Adspacewhich is inhighdemandandcomesatahighcostrelativetoremnantinventory.

PublisherAudiences

Publishers will sell their ad space based on the marketersrequest foramatchingaudience.Theadspace is labelledupandsoldbydescribingtheaudiencethatwillbeexposedtotheadsthatwilleventuallyoccupythespace.

PublisherPropertyAPublisherwebsiteispartofthePublisher’spropertyasisaPublisher’s app. This phrase is most commonly used as analternativetodescribingasinglepartofthePublishers

PublisherSpecsPublisherrequirementsandlimitationsforthedevelopmentofads that will be permitted to be delivered on their site orproperties.

QA QualityAssurance.TheperiodoftestingandadtocheckthatitfitswithadserverorPublisherspecifications.

QueryStringAtokenappendedtoad tagoneither theclickor impressionportiontocarryoutanadditionalfunction.Queryrelatestoarequestforinformationfromadatabase.

R&F AshortenedforminreferencetoReachandFrequencyReach Thetotalnumberofuniqueimpressions.

Recency

This describes the order bywhich things happenedwith themost recent in theprimaryposition.The termismostlyusedwhendescribingcustomattributionmodelinginordertoplaceavalueoneventsthatoccurredmorerecently.

Recordabilityrate SeeVTR.

ReferralURL

The URL of the page before the page the user is on. ThereferralURLiscountedastheURLresponsiblefordrivingtheusertothepresentURL.ThereferralURLcanbethepagetheadwasshownon.

Re-messaging AnothernamegiventoRetargeting.

RemnantInventory AdspacewhichisnotpopularwithbuyersandisavailableathighvolumeforlowcostfromaPublisher.

RichmediaspecsAcommontermusedbycreativeagenciesanddevelopers toask what coding requirements and restrictions the ad serverwillenforcetoensureadspassQA.

ROI

ReturnonInvestment.Amarketing/businesstermtoestablishwhat the Advertiser gets in return for the marketing spend.Intelligent marketers can use ad serving technology toestablish ROI goals and see if advertising efforts aregeneratingincrementalsalesorreach.

RTB Realtimebidding.Inreferencetoprogrammaticbuying.

SDK SoftwareDevelopmentKit.Inreferencetomobileapplicationdevelopment.

SecondScreen

The usermay be using one device to consumemedia at thesametimeasanother.Theprimaryscreenattractsmostoftheuserattention, the secondhas less.User trends suggestusersconsumesocialmediaonthesecondscreenwhileconsumingpremiumvideocontentontheprimaryscreen.

Seethroughrate(STR) ThepercentageofuserimpressionsthatcouldbecheckedforVerification.

SEM

Search Engine Marketing or Paid Search. Also usedinterchangeable todescribe thekeywordcreationandbuyingtoolsets (suchasGoogleAdwords) or used to describe a bidmanagementtool(suchasMarin)

SEOSearchEngineOptimizationorNaturalSearch.Anon-paidforadvertising channelwhere awebsite is optimizedby the sitedevelopertoattractmoresearchengineusertraffic.

Single-counting Vs.Double-counting

Asettingatanaccount levelwhichallowsconversions tobede-de-duplicated (single counting) or duplicated (doublecounting).PredominantlyusedtocontrolattributionreportingforNaturalSearch.

SKUnumber AlsocalledaProductID.UsedinreferencetoDCO.

Spiders (Search EngineSpiders)

Spiders are programs that crawl the web, simulating a user,loading pages, digesting the information those pages andpassing the information back to the source database forcataloguing. In Search Engine Optimization thesetechnologiesarecalledSpiders.

Spotlighttag AtermfromtheDoubleClickadservertomeanaconversiontagthatdoesnothavecontainerfunctionality.

SSP SupplySidePlatform.Inreferencetoprogrammatic.

Stagingenvironment

A testpageorseriesofpageson thePublisherorAdvertisersidewhichtendstoreceivetrafficbyinvitationtotestadsandcontent before putting the live on the main website. Thestaging environment is controlledby theCMSand is a termcommontotagmanagementandadQA.

Subdomain

The prefix before the domain or directly after it. Thesubdomain is designed to depict an area of a site or domainthatmightindicateasitesection.Subdomainisreferencedinad serving as part of domainURL verification, reportingsubdomains where page level transparency is unavailable tocall out ineffective impressions thatmay impedeverificationsuchasbrandsafety.

SWF The file type of flash animations. Most non-mobile displayadsareSWFfiles.

TMS TagManagementSystemorUniversalTaggingSystem.

TheCrossDomainFileAsecurityfeatureofwebsitessuchthattheimplementationofthefileallowsforthereadingandwritingtoremotedomains.

InreferencetoDCO.

TheFold

The fold of the user’s browser is the horizontal or verticalcutoff where the remainder of the ad or web page is notvisible. Scrolling the page reveals content or ads previouslyhidden by the fold. A term commonly used to describe thefunctionalityofViewability.

Token Anaddition toanad tag toperforma functionsuchascarryadditionalinformationtoanothersystem.See“Payload”.

Trafficker

The rolewhichmanages the campaign setup inside the thirdpartyadservinginterface.Traffickersareusuallyemployedata Media Agency or on-site at the ad serving provider.Traffickerscanalsobehiredtoworkremotelyviaatraffickingoutsourcecompany.

UniversalTagging SeeTagManagementSystemURL AwebaddressforHTTP

UserAgentThe technical term for a user’s browser on any device. Theuser agent is a code passed to the ad server to identify thebrowsertype.

Version

A reference to a variant of an ad for DCO. Each version isessentially an ad in its own right but would have beengenerated by a machine rather than created by a creativedeveloper.

Viewthroughrate(VTR) ThepercentageofimpressionsforwhichViewabilitycouldbemeasured.

VOD An overused incorrect industry reference to the InstreamVideochannel.

Wastage

A reference to impressions delivered ineffectively such thatthedeliveryoftheadswillbeirrelevanttotheuserandcouldnot possibly be used to attribute the ad to amarketing goal.Themarketingbudgetiswasted.

WhiteLabelingDisguising technology ownership such that a technologyappearstobeownedbyonepartybutinfactislicensedfromanothervendor.

XMLfileInreferencetoDCO.Anadditionalfiletoaccompanyanadtoallowtheadtobeupdatedwithoutrequiringaflashdevelopertodoso.

FIN