Optimizing your A/B Test Installation for Speed

26
Optimizing Your A/B Tests Installation for Speed & Conversion Rate Brought to you by:

Transcript of Optimizing your A/B Test Installation for Speed

Optimizing Your A/B TestsInstallation for Speed & Conversion Rate

Brought to you by:

Dennis van der HeijdenFounder & CEO of Convert.com

@Convert #ConvertAcademy

Learn How Speed Impacts A/B Tests and Overall Conversions&

Minimize the Risk of Impact

Purpose of Webinar

Packed Presentation, Let’s Get StartedThe Trust About Async/Sync Speed and How to Test Well

● How Tools Increase Load○ Any Script Lowers Speed○ Lower Speed Lowers CR○ When Scripts are loaded○ Async/Sync○ What to Do?

● Inner workings of A/B testing tools○ Async and sync codes○ Avoid seeing the blink, what

hack do A/B testing tools use?

www.convert.com

Script in Header

Being the first to load allows A/B testing tool to manipulate the content following and make the changes without blinking.

It is blocking and yes it is hurting site-speed, later on why that hardly matters...

www.convert.com

Manage Expectations of ClientsAt moment new JS is live baseline changes slightly

www.convert.com

What is really winning?● Are tools wrong?● What is common sense testing?

Manage Expectations of ClientsAt moment new JS is live baseline changes slightly

www.convert.com

Installing Anything Extra Lowers Conversion Rate● Installing any Javascript increases speed

and lowers conversion rate● Test results should offset that loss

Workings of A/B Testing ScriptsLayers of testing

www.convert.com

Where and when do we make changes?

Manage Expectations of ClientsAt moment new JS is live baseline changes slightly

www.convert.com

Expect Average Drop 1%

1. Installing A/B testing tool lowers load speedadding ~200 ms / ~500 ms on first load per CDN hub

2. + ~100 ms / 200 ms first load page 3. + ~50 ms / ~100 ms on follow up loads

Average CR drop will be in scenario 1/2/32% / 1% / 0.5%

Workings of A/B Testing ScriptsLayers of testing

www.convert.com

Where and when do we make changes?

The Inner Workings of A/B Testing Tools

psst, please don’t tell...

“The Dirty Secrets and Misdirections”

Understanding the terminology is super important when selecting a new A/B testing tool. There is a reason smart agencies go deep on sync/async parts of the script and how tools hide their true technology from you?

Let's talk about:

● Do you really understand sync and async?● Rethink your current A/B testing tools impact...● How do we trick your eyes and make SmartInsert© possible?

Don’t Copy Low Quality A/B Testing ArticlesHow do we really affect site speed and why to some tools “blink”?

www.convert.com

Understand How Page LoadsThanx New Relic, for image

www.convert.com

A/B Testing Tools TechnologyHow does Convert Patent Pending SmartInsert© Work?

www.convert.com

● Click to add text.○ Click to add text.

■ Click to add text.● Click to add text.

● 0.30 Sec.Nothing...

● 0.35 Sec.Convert Starts○ Hide Body for 1 Sec.○ Check Every 0.05 Sec.○ Unhide Body

● 1.50 Sec.Convert Long Done

● 0.30 Sec.DOMloading

● 0.35 Sec.DOMcontent Loaded Start

● 1.50 Sec.DOMcontent Loaded End

● 1.80 Sec.DOMcomplete

Fast Website + Convert Experiments

A/B Testing Tools TechnologyHow other A/B testing Tools Work… example #1

www.convert.com

● Click to add text.○ Click to add text.

■ Click to add text.● Click to add text.

● 0.30 Sec.Nothing…

● 0.35 Sec.Tool Starts○ Makes All Changes○ Visible Changes

Swapping / Blinking● 1.50 Sec.

Tool #1 Done

● 0.30 Sec.DOMloading

● 0.35 Sec.DOMcontent Loaded Start

● 1.50 Sec.DOMcontent Loaded End

● 1.80 Sec.DOMcomplete

Fast Website + Other A/B Testing Tool #1

A/B Testing Tools TechnologyHow other A/B testing Tools Work… example #2

www.convert.com

● Click to add text.○ Click to add text.

■ Click to add text.● Click to add text.

● 0.30 Sec.Nothing...

● 0.35 Sec.Nothing…

● 1.50 Sec.A/B Testing Tool Loads○ Hide Body for 1 Sec.○ Make All Changes.

● 1.80 Sec.Body Still Hidden

● 2.50 Sec. + 0.7 Sec. WhiteBody Unhide

● 0.30 Sec.DOMloading

● 0.35 Sec.DOMcontent Loaded Start

● 1.50 Sec.DOMcontent Loaded End

● 1.80 Sec.DOMcomplete

Fast Website + Other A/B Testing Tool #2

Synchronous and Asynchronous Scripts

Calling #BS on 35kb Code Size

Comparing Sync and AsyncSo ehh … so how much load time does a 35kb script take in seconds?

Does it matter how big the initial script is?

www.convert.com

Client Small Script http://cdn.fast.com

CDN.Fast.com35kb

jQuery X280kb

Geo-Check120ms

New Image Variation400kb

715kb+

120ms

Comparing Sync and AsyncNow what is it really?

www.convert.com

100ms 200ms 400ms

In a Sync-World ≠ Convert Experiments

100ms

200ms

In a Async-World ≠ Convert Experiments

400ms

Total Time = ~700ms

Total Time = ~450ms

Convert Experiments ScriptWhat do we really do and when?

www.convert.com

100ms 100ms 400ms

Sync and Async = + 100ms extra

200ms

200ms

100ms

Async Goals & Cookie SettingSmartInsert© Sync

Competitor #2 Async ScriptOthers might do it different… and why that matters

www.convert.com

100ms

100ms

400ms

Sync and Async

200ms

200ms 100ms

Async Goals & Cookie Setting

Async Script Variations = Blinking*

*When changes are loaded in areas A or B

A

B

C

Where can you find most important speed and debug settings in Convert?

● jQuery reduction● Measure jQuery Code Send● Chrome Plugin

Demo Where to Find Settings

Q&ADennis van der HeijdenFounder & CEO of Convert.com

@Convert #[email protected]

Convert ExperimentsAuthentic A/B Testing Software

http://www.convert.com

Brought to you by: