How we measure
Each fix follows a recorded measurement method. Results show what changed, how it was measured, and what remains uncertain.
- The expected effect and the smallest result worth keeping are recorded before a fix goes live, with the method that will judge it.
- An estimated window is not a promised date; an early read follows the recorded method.
- A comparison alone doesn't prove a cause.
- Can't tell is a legitimate result, not a failure.
- Applied and checked is not the same as Worked.
One fix, three possible results
The fix: restore the links to four category pages. Before it goes live, the method is recorded — search clicks over the 28 days after Google re-fetches the pages, against the pace before the fix, season-adjusted; it counts as Worked only from +8%.
- Applied
- The links are live on all four pages and checked. That says the change happened — not that it helped.
- Worked
- 28 days later, clicks are up 31%, past the +8% recorded before the fix went live, and nothing else that changed explains it.
- Can't tell
- Clicks rose 5%, or a Google update landed in the same weeks: the result can't be told apart from the noise, and the verdict says so.