Next Era of SEO: A Guide to SEO Split-Testing
10 likes3,776 views
The document discusses the importance of split-testing in SEO, highlighting how traditional A/B testing can be adapted for SEO strategies. It emphasizes the complexity of Google's algorithms and the impact of machine learning, indicating that not all sites respond similarly to changes. The guide outlines steps for conducting DIY SEO split-tests and concludes that split-testing should be a more integral part of SEO practices.
Next Era of SEO: A Guide to SEO Split-Testing
- 3. Risk Aversity Performance Cautious Fantastic. Recommend for awards. Cautious Terrible. Close to getting fired. FIREFIGHTER TRAITS Frank George
- 4. Students were now told: We made it up. Entirely fictional! !
- 5. How should a firefight treat risk? They should be cautious! They should embrace risk. It is a dangerous job! STUDENTS NOW ASKED: Frank George
- 6. People cannot let go of old beliefs. See what they want to see. !
- 7. Have you recommended ALT attributes on images for SEO reasons? ?
- 8. ALT ATTRIBUTES GOOD FOR SEO
- 9. TEST RESULT: ALT ATTRIBUTES HAD NO IMPACT
- 10. Google algorithm is too complex. We are too biased. !
- 11. Blog Post
- 12. Blog Post
- 14. HOW YOU CAN TEST — FREE TOOL
- 15. ‘TRADITIONAL’ CRO A/B TESTING IS USER ORIENTED Visitors A & B: control version of a page Visitors C & D: test version of the same page
- 16. SEO A/B TESTING IS PAGE ORIENTED Pages 1, 2, 3: Control Layout Pages 4, 5, 6: Test Layout All users see the same as each other
- 17. SEO A/B TESTING INCLUDES GOOGLEBOT Googlebot sees the same as users Pages 1, 2, 3: Control Layout Pages 4, 5, 6: Test Layout
- 18. OVERVIEW
- 19. SEO SHOULD BE INFORMED BY DATA & TESTS 1
- 20. 2009: ALGORITHM IS PRETTY WELL UNDERSTOOD No Panda No Penguin No RankBrain No Machine Learning We know what to do.
- 21. AMIT SINGHAL - HEAD OF SEARCH, GOOGLE
- 22. See this fascinating comment from a user called Kevin Lacker (@lacker):
- 23. I was thinking about it like it was a math puzzle and if I just thought really hard it would all make sense. Kevin Lacker
- 24. Hey why don't you take the square root? Amit Singhal according to Kevin Lacker
- 25. I was like, why a square root, it doesn't make much sense Kevin Lacker
- 26. Multiply by 2 if it helps, add 5, whatever, just make things work and we can make it make sense later. Amit Singhal according to Kevin Lacker
- 27. THE ALGORITHM WAS MOSTLY HANDMADE
- 28. ALGORITHM IS TOO COMPLEX TO VISUALISE
- 29. EARLY 2016: AMIT SINGHAL RETIRES
- 30. JOHN GIANNANDREA TAKES OVER
- 31. PREVIOUSLY HEAD OF MACHINE LEARNING
- 32. IMPACT OF MACHINE LEARNING
- 33. IMPACT OF MACHINE LEARNING Translate 10 YEARS RESEARCH 2 MONTHS MACHINE LEARNING Imagine the same step change in the search algo…
- 34. IMPACT OF MACHINE LEARNING
- 35. It will be extremely hard to understand many rankings. Even for Google. !
- 36. No, but I’m still pretty good at this You’re thinking this to yourself right now.
- 37. How good are we? ?
- 38. Google Information Needs Study (NBED LINK) http://adage.com/article/media/future-tv/303508/ SEARCH: future of tv advertising
- 39. Google Information Needs Study (NBED LINK) http://contentmarketinginstitute.com/2016/04/television-advertising-native/ SEARCH: future of tv advertising
- 40. DA PA LRDs Rank http://dis.tl/rankq-1a 91 47 2 ? http://dis.tl/rankq-1b 86 57 9 ? SEARCH: future of tv advertising
- 41. DA PA LRDs Rank http://dis.tl/rankq-1a 91 47 2 3 http://dis.tl/rankq-1b 86 57 9 8 SEARCH: future of tv advertising
- 42. https://www.jensinkler.com/fix-your-kettlebell-swing/ SEARCH: kettlebell technique
- 44. DA PA LRDs Rank http://dis.tl/rankq-2a 41 42 12 ? http://dis.tl/rankq-2b 76 44 20 ? SEARCH: kettlebell technique
- 45. DA PA LRDs Rank http://dis.tl/rankq-2a 41 42 12 4 http://dis.tl/rankq-2b 76 44 20 6 SEARCH: kettlebell technique
- 46. http://www.macworld.co.uk/how-to/iphone/how-to-back-up-iphone-ipad-backup-3595444/ SEARCH: do I need to backup my iphone
- 47. http://www.idownloadblog.com/2013/08/28/not-enough-icloud-storage-iphone-cannot-be-backed-up/ SEARCH: do I need to backup my iphone
- 48. DA PA LRDs Rank http://dis.tl/rankq-3a 77 59 16 ? http://dis.tl/rankq-3b 67 34 3 ? SEARCH: do I need to backup my iphone
- 49. DA PA LRDs Rank http://dis.tl/rankq-3a 77 59 16 7 http://dis.tl/rankq-3b 67 34 3 5 SEARCH: do I need to backup my iphone
- 50. How do we compare? ?
- 51. WE SURVEYED 2000 PEOPLE FOR 10 QUERIES
- 52. % OF UK SEARCH RESULT PAIRS PREDICTED CORRECTLY 30% 40% 50% 60% 70% Laypeople
- 53. % OF UK SEARCH RESULT PAIRS PREDICTED CORRECTLY 30% 40% 50% 60% 70% Laypeople <= 3 yrs SEO
- 54. % OF UK SEARCH RESULT PAIRS PREDICTED CORRECTLY 30% 40% 50% 60% 70% Laypeople <= 3 yrs SEO > 3 yrs SEO
- 55. % OF UK SEARCH RESULT PAIRS PREDICTED CORRECTLY 30% 40% 50% 60% 70% Laypeople <= 3 yrs SEO > 3 yrs SEO Coin flip
- 56. % OF UK SEARCH RESULT PAIRS PREDICTED CORRECTLY 30% 40% 50% 60% 70% Laypeople <= 3 yrs SEO > 3 yrs SEO Coin flip DeepRank
- 57. TESTING - STOP GUESSING WHAT WORKED Testing the same type of change on different sites can have differing effects.
- 58. THE CASE FOR TESTING 1) Predicting effect in algo is really hard (no best practice)
- 59. THE CASE FOR TESTING 1) Predicting effect in algo is really hard (no best practice) 2) Some sites respond differently to others for same change
- 60. THE CASE FOR TESTING 1) Predicting effect in algo is really hard (no best practice) 2) Some sites respond differently to others for same change 3) Machine Learning impact on algo will make this worse
- 61. THE CASE FOR TESTING 1) Predicting effect in algo is really hard (no best practice) 2) Some sites respond differently to others for same change 3) Machine Learning impact on algo will make this worse 4) Optimise further than previously possible
- 62. 2 SOME RESULTS FROM OUR SEO TESTS
- 63. DistilledODN Your Server Your Home Page CDN Distilled ODN Your Optimized Home Page
- 64. 2 IMPORTANT GRAPHS. THIS IS THE FIRST.
- 65. BLACK IS THE VARIANT, TREAT BLUE AS THE CONTROL
- 66. BLACK (ACTUAL) HIGHER THAN BLUE (CONTROL) - POSITIVE
- 67. THIS IS THE START DATE
- 68. SECOND GRAPH: TOTAL ADDITIONAL SESSIONS
- 69. SHADED AREA SHOWS CONFIDENCE INTERVAL
- 70. TEST IS SIGNIFICANT WHEN THE FAN CROSSES 0
- 71. IN THIS CASE IT HAPPENED HERE:
- 72. HERE THEY ARE TOGETHER.
- 73. ALT ATTRIBUTES 1
- 74. HYPOTHESIS: ALT attributes on images are good for SEO
- 75. RESULT: ALT attributes had no effect
- 76. SEO ‘CONTENT’ 2
- 77. HYPOTHESIS: Text content on category pages is sort of good?
- 78. RESULT: No real impact when removed
- 79. RESULT: A 3.1% improvement in organic performance
- 80. Not every site is the same. Identical changes can have different results. !
- 82. HYPOTHESIS: Javascript impacts Google’s crawling / indexation MAY 2014 “our indexing system has been rendering a substantial number of web pages more like an average user’s browser with JavaScript turned on. OCT 2015 “We are generally able to render and understand your web pages like modern browsers.”
- 83. Javascript EnabledJavascript Disabled SITUATION: Javascript needed to render product list
- 84. RESULT: CSS fallback led to a 6.2% uplift
- 86. Title tag before: Which TV should I buy? - Argos Title tag after: Which TV to buy? - Argos HYPOTHESIS: Title tags should match greatest search volume term
- 87. RESULT: Organic sessions decreased by average 8%
- 88. STRUCTURED MARKUP ON CATEGORY PAGES 5
- 89. SITUATION: Category pages have lots of images but little text
- 90. HYPOTHESIS: Structured Markup per product will help
- 91. RESULT: Organic sessions increase by 11%
- 92. 3 HOW YOU CAN RUN TESTS YOURSELVES (+ A FREE TOOL)
- 93. 3 STEPS TO DIY SEO SPLIT-TESTS 1. Create two buckets of pages. 2. Make a change to all pages in one bucket. 3. Analyse which bucket performs better.
- 94. 1. CREATE TWO BUCKETS OF PAGES. A good DIY approach is using GA segments. Easy to make (see link). Suggestion: Create Segments by category (e.g. blog tag, product category).
- 95. At least 1,000 organic sessions per day to each bucket !
- 96. 2. MAKE A CHANGE TO ALL PAGES IN ONE BUCKET. Control Variant This will be your variant bucket. Utilise your CMS where possible. Ask Dev for scalable approach.
- 97. 3. ANALYSE WHICH BUCKET PERFORMS BETTER. distilled.net/diy-splittester
- 98. Typically takes 2 - 3 weeks to reach significance. !
- 100. THEN READ THIS: Blog Post
- 101. 4 SUMMARY
- 102. SPLIT-TESTING IS BECOMING MORE NECESSARY 30% 40% 50% 60% 70% Laypeople <= 3 yrs SEO > 3 yrs SEO Coin flip
- 103. SPLIT-TESTING IS BECOMING MORE ESTABLISHED
- 104. SPLIT-TESTING MAKES YOU MORE CREDIBLE
- 105. SPLIT-TESTING IS ALREADY ACCESSIBLE
- 106. Step change in your SEO approach. Start now. !
- 107. WWW.DISTILLEDODN.COM CHECK OUT OUR SEO A/B TESTING PLATFORM: @TomAnthonySEO @Dom_Woodman @WillCritchlow DIS.TL/SMXDE17 GET THE DECK: