Multivariate DIY - Daigneault
Download as PPT, PDF1 like775 views
The document provides an overview of multivariate optimization, detailing how to use Google's free optimization tool to improve conversion rates through systematic testing of various webpage elements. It explains the process of setting up tests, creating variations, and analyzing results, emphasizing the advantages of multivariate testing over traditional A/B testing. Additionally, it includes resources and next steps for running personal tests and highlights recent test results and their implications.
![Multivariate Optimization Do-It-Yourself Edition [150] April 9, 2010 Steve Daigneault](https://image.slidesharecdn.com/mvtdoityourself-100408190743-phpapp01/85/Multivariate-DIY-Daigneault-1-320.jpg)
Multivariate DIY - Daigneault
- 1. Multivariate Optimization Do-It-Yourself Edition [150] April 9, 2010 Steve Daigneault
- 2. Goals Solid understanding of multivariate optimization and why it’s so great Overview of how Google’s free optimization tool works Examples of recent tests and results Resources and suggested next steps for you to run your own tests
- 3. What is multivariate optimization? We all know A/B tests Create two versions Split traffic Watch response rates
- 4. Your landing page submit Trust Mrs. Peacock. She knows how to light candlesticks without killing anyone. Give to the Library
- 5. Things you want to test submit HEADLINE: 2 alternatives Save the Conservatory Colonel Mustard Needs You PICTURE: 2 alternatives FIELDS: 1 alternative Remove 1 field Donate now! BUTTON: 1 alternative Bigger and orange
- 6. How would you A/B test this? Set-up different landing pages for: Headline: 2 alternatives + 1 control = 3 Picture: 2 alternatives + 1 control = 3 Fields: 1 alternative + 1 control = 2 Button: 1 alternative + 1 control = 2 3 x 3 x 2 x 2 = 36 pages 36 – 1 control = 35 new pages to build
- 7. Test everything at once Input all elements at once Software creates many page variations Traffic split among page variations Takes into account interactions between all the elements Fancy math tracks the effect each individual item has on conversions Multivariate > A/B testing
- 8. HOW DOES OPTIMIZER WORK?
- 9. Actually builds all page variations Some software will build fewer pages + use predictive modeling Optimizer = free + great for basic tests Statistically significant results using the same traffic but using it more wisely Optimizes conversions (not average gift*) Google Optimizer
- 10. Add Google Optimizer code around elements you want to test.* Upload variations of those elements into Google Optimizer. Optimizer then builds page variations, splits traffic and calculates changes in behavior. *requires you have direct access to page code Overview
- 11. Choose the page you want to improve.* EX: donation page Identify your conversion page.* EX: confirmation page Identify page elements you want to test. EX: button and header image *these pages will get tracking code. 7 Steps to Test
- 12. Create variations. EX: new graphic buttons and headers Add Optimizer code around the elements you’re testing on your test page. Upload the variations to those elements into Google Optimizer. Review and launch. 7 Steps to Test
- 13. SHOW ME RESULTS
- 14. Sample results: Page variations
- 15. Estimated conversion rate range provides the most immediate insight into overall performance. This column shows you how well each combination is performing relative to your original content.
- 16. Chance to beat original the probability that a particular combination will be more successful than your original content. Above 95% (or below 5%) is considered significant. Above 99% (or below 1%) is considered “high confidence”.
- 17. Chance to beat all the probability that a combination will be more the successful than all the other combinations running in the experiment. Usually a lower number because it’s competing against many page variations vs. just the original.
- 18. Observed improvement displays the percent improvement over the original combination.
- 19. Conversions/impressions the raw data of how many conversions and page views a particular combination generated. (You want 200+)
- 20. Sample results: Page elements
- 21. Relevance rating shows how much impact a particular section has on your experiment. 0 means it had little to no impact. 5 means it was a major distinguishing factor.
- 23. vs.
- 24. vs.
- 25. vs.
- 26. #6: new nav and new button original gift string #5: new nav and new gift string original button Winner:
- 27. MORE ANSWERS
- 28. How much traffic do I need? How many things can I test? www.google.com/analytics/siteopt/siteopt/help/calculator.html Traffic Calculator
- 29. How much could improvements help? Value Calculator www.bplans.com/business_calculators/conversion_rate.cfm
- 32. Optimizer Engineer Blog www.gwotricks.com/default.htm
- 33. Frequently Asked Questions www.google.com/analytics/siteopt/siteopt/help/FAQ.html Technical overview gives you a deeper understanding of the tool www.google.com/analytics/siteopt/siteopt/help/techoverview.html Comparisons of different tools to help you decide if Google Optimizer is right for you www.whichmvt.com / Resources
- 34. How Was this Session? Call In Text Online Call 404.939.4909 Enter Code 150 Text 150 to 69866 Visit nten.org/ntc-eval Enter Code 150 Tell Us and You Could Win a Free 2011 NTC Registration!