DMA - Art of Targeting and Personalization

Data | Domain | Delivery

Presented to:

Art of Targeting & Personalization

Stephen H. YuAssociate Principle, Analytics and Insights

1Confidential eClerx – An ISO/IEC 27001:2005 Certified Company March 16

Stay ahead of the competition – gain access to advanced analytics techniques

and strategies with the 2015 DMA Analytics Journal

Grow your brand –reserve advertising space for the 2016 Analytics Journal

Learn and network – attend DMA’s 2016 Marketing Analytics Conference in

Austin, Texas | June 23-24

Increase your visibility – sponsorship opportunities for every company and

budget

Get published – submit an article as part of the Analytics Advantage Blog Series

Analytics Community

http://thedma.org/wp-content/uploads/DMA-Analytics-Journal-Final1.pdf

http://thedma.org/membership/member-groups-communities/analytics-community/analytics-journal/

http://mac.thedma.org/

http://thedma.org/wp-content/uploads/Sponsorships_9.21.15.pdf

http://thedma.org/membership/member-groups-communities/analytics-community/analytics-advantage-blog-series/

http://thedma.org/membership/member-groups-communities/analytics-community/analytics-advantage-blog-series/


What we will cover

DATA LANDSCAPE INSIGHTS VIA ANALYTICS WHY MODEL?

ART OF TARGETING "ANALYTICS-READY"

ENVIRONMENT

PROPER PERSONALIZATION

VIA DATA AND ANALYTICS


Where the Data Movement is Going

No guessing game – You MUST know your target

Vast amount of online & offline data collected

But are they being used properly?

Analytics play a huge role in prospecting & CRM

Short paced marketing cycle getting shorter

Marketers must stay relevant to their target to cut through the noise

Huge difference between advanced marketers and those who are falling behind

And it’s all about proper “Personalization”

Winners are the ones who know howto stay relevant with their target bywielding the power of data faster.


It is NOT about Channels or Technologies

There is no such thing as an “online person”

– It is almost offensive

– Channel-centric view confuses buyers

– New channels and technologies in the future

» What then?

This data business should be about “People”

– No one is one-dimensional

– “Buyer-centric” point of view

» Should NOT be channel-, product-, division- or brand-centric

» But most marketers are

» Buyer-centric data structure leading to proper “Personalization”

Never about the technology, but about the people who are looking

at the new device (or even thin air)

– And they are in control, not marketers!

“The Future ofOnline is Offline”– Stephen H. Yu, 2002


Take the data seriously, not just your

gut feelings

Define the goals first, then control

the flow of data

Don’t blindly trust machine based

solutions

Be logical, as there are no toolsets

that read minds

Set specific goals for small

successes

It is about the Data Users, too

Don’t be a “Data Plumber”, but a

businessman

Don’t be technology oriented, but

solution oriented

Don’t do things just because you

can

For Decision Makers For Data Scientists


Leading with Analytics

“A database is

not just a sum of

all data sources,

and Analytics is

not just an array

of statistical

techniques”

1.Business Goals

2.Answers via Analytics / Modeling

3.Databases Optimized for Analytics

Solution design based on business goals, not around capabilities


What 1:1 Marketing is about

Marketers must know:

– Whom to contact, and

– What to say, if they decided to contact

someone

» What to offer through what channel and

when

Analytics help marketers with both:

– Targeting, and

– Personalization

Everyone is being

bombarded with

marketing

messages through

multiple channels


Different Types of Analytics

“ANALYTICS” MEANS DIFFERENT THINGS…

BI (Business

Intelligence) Reporting:

Display of success metrics,

dashboard reporting

Descriptive Analytics:

Profiling, segmentation,

clustering

Plus, “Prescriptive Analytics” for All Stages

Predictive Modeling:

Response models, cloning

models, value models,

revenue models, etc

Optimization:Channel optimization,

marketing spending analysis,

econometrics models


Insights through Data Refinement

Insights are derived from data through the refinement process

Data Collection by

Channel

Rapid Data Retrieval

Basic Dashboard

Data Hygiene and

Standardization

Consolidation and

Summarization

Advanced Analytics

including Statistical

Modeling

Comprehensive

Dashboard and BI

Reporting

Ad hoc Reports

Campaign Targeting

and Management

Personalization

DATA PLAYERS MUST EXCEL IN:

Collection Refinement Delivery


Why Model?

Models summarize complex data into simple-to-use “scores", and fill in the gaps by converting “unknowns” to “potentials”

Increase targeting accuracy

Reduce costs by contacting less/smart

Stay relevant with target customers

Achieve consistent results

Reveal hidden patterns in data

Reach marketing automation faster

Expand the target universe

“Supposedly” save time and effort


Why NOT Model?

Universe too small

Predictable data not available

1:1 marketing channels not in

plan

Tight budget

Lack of resources

Really? Remember 1:1 Marketing

is about:

Knowing whom to engage

Knowing what to offer if you

decided to engage someone

Models provide answers for both


What is a Model?

Target vs. Non-Target, such as

Buyer vs. Non-Buyer

Responder vs. Non-Responder

Loyal vs. Attrition

High Value vs. Low Value

Defining target and non-target is equally critical

“A model is a

mathematical expression

of differences between two

dichotomous groups”


Art of Targeting

Remember T, C, M

1.Target

2.Comparison Universe

3.Methodology

Defining the proper target is most critical even

more than the methodology

Marketers must get involved in Target

Definition

– State the goals and usages clearly

– Don’t be a bad patient demanding specific

prescriptions

“Some targets are not what they seem…”

Start by hanging the target in the right place


Defining the Target (or Targets) (1/4)

How frequent is frequent enough?

How much is high enough value?

How big is the size of the ideal target?

Continuous Target



Multiple distinctive segments in

the target universe

For example,

Infrequent Big Spenders

Frequent Small Spenders

New Customers

Dormant Customers

Geographic targets

Demographic segments

Multiple Targets



Multi-step approach for multi-step

sales/marketing

‒ Sales pipeline (various stages of

lead qualification)

‒ Open-Click-Browse-Convert-Repeat

cycle

Very narrow target in a big universe

Sub-targets within major segments

Target within a Target



Inversely Related Targets

For example,

Frequent shoppers with

low average spending

Responsive prospects

with bad credit

Build multiple models

and find cross-sections


Any Pain Implementing Models?

Modelers are fixing data all the time

Repeatedly rely on a few popular variables

Always need more variables

Takes too long to build models and deploy them

Inconsistencies shown when scored

Disappointing results!


Predictive Modelling is all about “Ranking”

12 3

Relational or unstructured

databases won’t cut it

Must create “Descriptors”

that fit the level that needs to

be ranked

Households

Individuals

Companies

Email Addresses

Products

Ultimately, models must

properly “Rank”

Define the level of

data accordingly


Single View of Customer

Web

CRM

Email

POS

Social

Media

Call

Centre

Mobile App

Mobile App

Call

Centre

CRM

Social

Media

POS

WebEmail


Marketing Database Supporting Analytics

Database Optimized for Analytics

– Analytics supporting efficient Targeting and Personalization

– “Buyer-Centric” Portrait

» Transform Channel-, Product-, Division-, or Brand-Centric data to “Descriptors” of the

Target

The Solution – “Analytical Sandbox”

– Additional table(s) without overhauling existing DB structure

– Ideal environment for:

» Analysts and analytical toolsets

» Model maintenance/scoring

– Finished groundwork for level-playing field

» Data Hygiene/Standardization

» Categorization/Tagging/Binning

» Data Consolidation

» Variable Creation

– End-to-end run – from data collection/refinement to campaign execution/backend

analysis


“Analytical Sandbox” Conceptual Flow

Analytics is not just about statistical techniquesRealign data to achieve accurate, consistent and speedy results

Hygiene / Edit

Categorization

Consolidation

Summarization

Variable

Creation

Model

Development

Model

Application

Reporting

Knowledge

Sharing

Results Analysis

Attribution

Analytical Sandbox


Customer-Centric Environment

MASTER

CUSTOMER

TABLE

W/ CONSISTENT

ID

TRANSACTION

HISTORY

PRODUCT &

AFFINITIES

OTHER

ACTIVITIES

DIRECT

MARKETING

PROMOTION

HISTORY

EMAIL

PROMOTION

HISTORY

OTHER

MARKETING

PROMOTION

HISTORY

DM

RESP

EM

RESP

Descriptors of Customers by

Product, Time-series, Amount,

Activity, Status, Etc.

Descriptors of Customers by

Promo/Response

(Adequate, Over & Under)

BEHAVIOR CHANNEL

Summarization &

Attribution

Summarization &

Variable Creation

Customer-Centric View


Monetary

Frequency

Recency

Sample Variables after Summarization

Weeks since last online purchase

Years since member sign up

Days since last delinquent date

Months since last response date

Orders by offer type

Orders by product/service type

Payments by pay method

Average days between transactions

Total $ past 24 months

Life-to-date spending

Average dollars by channel

Average dollars by product type

BEFORE AFTER SUMMARIZATION


Addressing your customers by their first

names?

Suggesting more of the same products

that they just purchased through

collaborative filtering?

Collecting explicitly expressed

preferences and reacting to them?

Keeping in touch with your customers

all the time?

Customizing emails and landing pages

based on customer preference?

Knowing when to contact through what

channel?

About Personalization

But, maybe you are

‒ “Personally” annoying your

customers and prospects

‒ Personalizing a fraction of the base

and completely ignoring the others

‒ Personalizing sporadically only

when obvious trigger data become

available

Personalization is the big

buzzword now, but what does it

mean?

They are all better than copying and

pasting the same content to

everyone…


“Personalization is about the Person”

Look at it from the customer’s point of view…

Transform transaction, event, or product level data to:

‒ Describe people, not product

‒ Create 360-degree Single Customer View

(“Analytical Sandbox”)

Develop “Personas”, then match products to them

‒ Not the other way around

‒ Fill in the gap with modeling

Personalization is about the Person

No one is just an “online” or an “offline” person

“Personalization Engines” are often overrated

(especially product level collaborative filtering is on

auto-drive mode)

Raw SKU level data are utterly inadequate for

personalization


Verified “known” explicit data are

scarce even in this age.

Data are often missing for targets who are:

‒ New to business

‒ Dormant

‒ New to channel

‒ Hiding their tracks

But, most personalization efforts are

done based only on “known” explicit

data!

Need to maximize the value of available

data, even implicit or anonymous data

Even now, real data are hard to come by

DATA COVERAGE

RIC

HN

ES

S O

F D

AT

A


Sample PersonasRaw Data

Demographic / Firmographic

Transaction Data / RFM Data

Products & Services Used

Promotion / Response History

Channel Usage Data

Lifestyle / Survey Responses

Delinquent history

Call / Communication Log

Movement Data

Survey / Social Media / Sentiments

Data to Answers via Modelling

Likely to buy a luxury car

Likely to take a foreign vacation

Likely to be a wine enthusiast

Likely to have a home office

Likely to be a risk averse investor

Likely to respond to free shipping offer

Likely to be a high value customer

Likely to be qualified for credit

Likely to upgrade/leave/come back

Formulate the answers through modeling


1.Not even considering personalization yet; still spraying the same HTML to everyone

2.Some personalization is considered, but do not know where to start

3. Identified basic toolsets for personalization, but do not have specific data or technology roadmap

4.Created the data roadmap, but did not start thorough data inventory

5. Identified required data sources, but datasets are not cleaned up or consolidated for 360-degree

view of customers

6.Datasets are ready for personalization, but only with “known” explicit data; statistical modeling to

fill in the gaps is not considered yet

7.Tested personalization engines through major marketing channels of choice, employing collected

“known” explicit data

8.Creating model-based “personas” with all available data, filling in the gaps with statistical

techniques

9.Personalizing most messages and offers through every touch point, employing explicit data (“hot”

data) and implicit/inferred data (“personas”)

10.Collecting and utilizing results data to enhance targeting models, personas and personalization

engines continuously, leading to full automation

Data & Analytics Steps towards Proper Personalization


Holistic Approach to Personalization

Data, analytics, content management and delivery working in conjunction

INSTALL

PERSONALIZATION

ENGINE

TEST THE

ENGINE WITH

SIMPLE

SEGMENTS

PERSONALIZE

EVERY

MESSAGE

Deployment of interactive

display capabilities

– Web

– Email

Ground work for next steps

Data-driven

personalization - secure

access to deeper data

Employ all available

“known” explicit data

Employ model-based “personas”

Convert “unknown” to

“potentials”

Extend personalization to

customers with little or no history

1 2 3PHASE PHASE PHASE


Multiple Dimensions of a Person

Frequent Traveller

Early Adapter

Family Oriented

Bargain Seeker

Examples of Personas:No one is one-dimensional


More about messaging than targeting

Pin a target individual into one

segment at a time

Hard to update with reliable

consistency

Group them first, describe them later

‒ End up calling everyone in a

segment the same way

Segments vs. Personas

Built for 1 attribute at a time

Describe an individual with multiple

attributes

Identifies dominant characteristics of

a person via side-by-side comparison

Each persona represents diverse

array of data

Easy to update, 1 persona (i.e., one

model) at a time with consistency

Ready for multi-channel marketing

Clustering/Segmentation/Cohorts Personas


“Personas” built for specific attributes

Project small “known” attributes to large universes of “unknowns” in form of

model scores

Fill in data gaps leaving no missing value – Scores for every record using all

available data

Enable side-by-side comparison of attributes – Quickly find dominant

characteristics of an individual

Simplify matching process between individuals and best suitable

products/services

Support message “rotation” for an individual customer using multiple personas

Lead to marketing automation – Simple scores are no burden to personalization

engines

Model-based Personas in Action


Examples of Model-based Personas

Frequent Flyer

Foreign Traveler

Luxury Hotel

Gourmet

Wine Enthusiast

Adventure Seeker

Young Family

Budget Conscious

Family Oriented

Romantic

High-end / Luxury

Seasonal

Frequent Small Gifts

Pre-packages

Bargain Seekers

Specialty Items

Corporate Purchase

Home Office

Consumables /

Repeat Purchase

Big Ticket Items

Technology Buyers

Early Adopters

Trend Followers

For Hospitality Industry For Gift Industry For Business Solutions


Modern 1:1 marketing is about proper targeting and personalization with the buyer-

centric view

Business goals first; it is not about data or technology

Invest in analytics – models can pack large amount data into simple answers to questions

Databases must be optimized for analytics and modeling – maintaining consistency is the

key

Add “Analytical Sandbox” to the existing data environment for end-to-end efficiency

Personalization is about the person, not channel

Expand the horizon: Personalize all the time for everyone through every touch point

Move from simple segments to personas for constant and effective personalization

Key Takeaways


Questions?

DMA - Art of Targeting and Personalization

Data & Analytics

Transcript of DMA - Art of Targeting and Personalization