Upgrade to Pro — share decks privately, control downloads, hide ads and more …

AB Testing - Why? What? How? - Web2Day 2017

AB Testing - Why? What? How? - Web2Day 2017

How often do you change the colour of a button on your website? How often do you change the layout of a part of the website? How do you measure the effect of these changes? Do you just make an assumption that your changes will be liked by your users/customers? I will be talking about a data-driven approach – A/B testing. I plan to cover these aspects –

– Why should you use A/B testing in your business?
– What all can you test using A/B testing?
– How can you do efficient A/B testing of change in your product?

I will also cover some scenarios where A/B testing may not work or may fail and how you can overcome those limitations by being careful.

Sahil Dua

June 09, 2017
Tweet

More Decks by Sahil Dua

Other Decks in Technology

Transcript

  1. @sahildua2305 #web2day #nantes #abtesting Booking.com Graduate Software Developer Git Pandas

    Open Source Contributor Open Source Contributor DuckDuckGo Community Leader Sahil Dua @sahildua2305
  2. @sahildua2305 #web2day #nantes #abtesting Agenda 1. A/B Testing - Quick

    Introduction 2. Why A/B Testing is important? 3. What all can you test? 4. A/B Testing process 5. Why do tests fail? 6. Conclusion
  3. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact
  4. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact 3. Not dependent on opinion or assumption
  5. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact 3. Not dependent on opinion or assumption 4. Cuts down on functionality bloat
  6. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact 3. Not dependent on opinion or assumption 4. Cuts down on functionality bloat 5. General product deployment is easier
  7. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact 3. Not dependent on opinion or assumption 4. Cuts down on functionality bloat 5. General product deployment is easier 6. and finally ...
  8. @sahildua2305 #web2day #nantes #abtesting The key to learning is feedback.

    It is nearly impossible to learn anything without it. “ - Stephen D. Levitt
  9. @sahildua2305 #web2day #nantes #abtesting Why A/B Testing? 1. Your customers

    decide what stays 2. Easy to measure the business impact 3. Not dependent on opinion or assumption 4. Cuts down on functionality bloat 5. General product deployment is easier 6. Learn from your customers
  10. @sahildua2305 #web2day #nantes #abtesting What to A/B test? Webpage 1.

    Call-to-Action 2. Content 3. Copy 4. Visual media 5. Site navigation 6. Personalization
  11. @sahildua2305 #web2day #nantes #abtesting What to A/B test? Webpage 1.

    Call-to-Action 2. Content 3. Copy 4. Visual media 5. Site navigation 6. Personalization Social 1. Social icons 2. Profile images 3. Social ads 4. Social messaging
  12. @sahildua2305 #web2day #nantes #abtesting What to A/B test? Email 1.

    “From” name 2. Greeting 3. Subject length 4. Images 5. Personalization 6. Optimal time
  13. @sahildua2305 #web2day #nantes #abtesting What to A/B test? Email 1.

    “From” name 2. Greeting 3. Subject length 4. Images 5. Personalization 6. Optimal time Mobile 1. Navigation 2. Display 3. Push notifications 4. Deep linking
  14. @sahildua2305 #web2day #nantes #abtesting A/B Testing is a tool to

    collect data so that you can make accurate decisions. “
  15. @sahildua2305 #web2day #nantes #abtesting A/B Testing Process STEP 1 Observe

    a problem STEP 2 Write a hypothesis STEP 3 Brainstorm solutions STEP 4 Run the experiment STEP 5 Analyze the results, make a decision
  16. @sahildua2305 #web2day #nantes #abtesting Qualitative Data 1. User Testing 2.

    Survey Data 3. Heat Maps 4. Customer-facing Teams
  17. @sahildua2305 #web2day #nantes #abtesting Quantitative Data 1. Web traffic 2.

    Email marketing 3. Purchase history 4. Customer support tickets
  18. @sahildua2305 #web2day #nantes #abtesting What are we looking for? 1.

    Confidence Interval 2. Statistical Significance 95% Statistical Significance = 5% probability that the observed difference was due to chance
  19. @sahildua2305 #web2day #nantes #abtesting What are we looking for? High

    statistical confidence Lower risk that it was due to chance
  20. @sahildua2305 #web2day #nantes #abtesting Metrics to measure Webpage 1. Page

    views 2. Sign ups 3. Clicks on CTAs 4. Conversion Social 1. Images uploaded 2. Sign ups 3. Active users 4. Session duration
  21. @sahildua2305 #web2day #nantes #abtesting Metrics to measure Email 1. Open

    ratio 2. Links clicked 3. Conversion Mobile 1. Notifications clicked 2. Clicks on CTAs 3. Conversion
  22. @sahildua2305 #web2day #nantes #abtesting Technical Flaws - Increase in page

    load time? - New errors or warnings? - Cross-browser issues? - Cross-device issues?
  23. @sahildua2305 #web2day #nantes #abtesting Missing Context - Is something less

    visible now? - Is it in the wrong spot? - Did you remove something to “make space”?
  24. @sahildua2305 #web2day #nantes #abtesting Untargeted Audience - Did they type

    the URL directly? - Are they returning or new visitors? - Did they come from an email campaign? - Are they regular customers?
  25. @sahildua2305 #web2day #nantes #abtesting Timing Issues - Is your change

    seasonal? - Showing the feature too early? - Does the tooltip disappear too fast?
  26. @sahildua2305 #web2day #nantes #abtesting Low Traffic Do you have enough

    traffic? https://www.optimizely.com/resources/sample-si ze-calculator/
  27. @sahildua2305 #web2day #nantes #abtesting Why do tests fail? - Technical

    flaws - Missing context - Untargeted audience - Timing issues - Low traffic
  28. @sahildua2305 #web2day #nantes #abtesting Conclusion 1. Start testing now 2.

    Confirm your instincts 3. Learn and take actions