Ten findings, each with its method and sample size
Each row opens its own write-up: the method, the window, and what the finding does not show.
| No | Finding | Sample | Result |
|---|---|---|---|
| 01 | Ranking better cost the clicks | 1.4M impr | 0.17% |
| 02 | Impressions that cannot become a visit | 2.1M impr | 77% |
| 03 | Ad revenue spread, ad slots held constant | 21 pages | 4.6x |
| 04 | Ad serving in one market | 102 countries | 0.072 |
| 05 | AI assistants as a share of traffic | 2,728 sess | 1.1% |
| 06 | A cluster nobody had looked at | 184,133 | 0.14% |
| 07 | Schema kept after Google dropped it | 153 sess | 26% |
| 08 | Position did not predict clicks | 382 pages | 8.7x |
| 09 | Biggest page, not the best page | 382 URLs | 55x |
| 10 | Long tail advantage, and its limits | 3,328 queries | 1.57x |