Googlebot and Search Console lessons from SEO testing

My presentation at BrightonSEO Brighton April 2024.

When we run SEO experiments, we are always on the lookout for more data to help us analyse them better, get more confident results, and detect smaller uplifts. Googlebot crawl data and Search Console data are two promising data sources, but each comes with its quirks.

In this talk, I cover the interesting and at times counter-intuitive lessons we have learned trying to glean insights from huge amounts of previously unavailable data.

Transcript

1. Googlebot & Search Console data Will Critchlow SearchPilot @willcritchlow Speakerdeck.com/willcritchlow
2. Biggest channel Organic Search
3. Biggest channel Biggest opportunity Organic Search
4. Biggest channel Biggest opportunity Least understood channel Organic Search
15. Can we improve our test analysis?
16. Googlebot crawl data Can we improve our test analysis?
21. Expected benefit: better test analysis
22. Expected benefit: better test analysis Less than half the effect can possibly be visible in the data
23. Expected benefit: better test analysis Up to ¾ of the effect could be visible in the data
24. Expected benefit: better test analysis Up to 90% of the effect could be visible in the data
25. Surprise benefit: we can run more tests, faster 90%
26. Surprise benefit: we can run more tests, faster 90% 4 days
27. Surprise benefit: we can run more tests, faster 90% 4 days Sometimes even faster!
28. Two big benefits to fast recrawl 1. End inconclusive tests sooner
29. Two big benefits to fast recrawl 1. End inconclusive tests sooner 2. Shorter “cooldown” windows between tests
30. GSC data
31. OK, it won, but WHY?
32. There are only three SEO hypotheses:
33. There are only three SEO hypotheses:
34. There are only three SEO hypotheses:
35. There are only three SEO hypotheses:
40. Challenges with Search Console data
41. Sampling #1
42. Sampling (sometimes this is really bad) #1
43. Sampling #1 Averaging averages #2
44. Sampling #1 Averaging averages #2 New and lost rankings #3
45. Summary
46. Summary Googlebot data lets us run more tests, faster
47. GSC data can explain results Summary Googlebot data lets us run more tests, faster
48. GSC is not good for declaring winners Summary Googlebot data lets us run more tests, faster
49. Thank you! Will Critchlow SearchPilot @willcritchlow Speakerdeck.com/willcritchlow