Method

Everything published here is measured on properties we built and operate. There is no modelled data, no industry estimate and no third-party benchmark presented as a finding.

What every figure carries

Three things, stated next to the number rather than in a footnote: the instrument and the cut, the observation window as absolute dates, and the sample size. A figure missing any of the three is not published.

This is not a stylistic preference. Most claims about search performance are unfalsifiable by construction, because the audience cannot check them. Stating the method is what makes a claim checkable, and a checkable claim is the only kind worth making.

Instruments

• Google Search Console, pulled through the API rather than read from the interface
• GA4, session and revenue metrics joined to page path
• AdSense reporting, read by country and by page rather than by ad unit

Where a cut is known to be unreliable we say so on the finding. Some dimensions do not reconcile once a filter is applied, and a figure that has not been reconciled against a second cut is reported as a hypothesis, not a result.

What we will not claim

• That a rate computed from a handful of events is a trend. Small samples are labelled as small, in the argument rather than a footnote.
• That a correlation observed on one property generalises. What is published is what was measured, on what, and when.
• That any of this predicts your result. The diagnostic measures your data; it does not assume ours applies to you.

Corrections

Findings are corrected in place, with the change noted and dated. A measurement that turns out to be wrong is more useful published with its correction than quietly deleted, and a site arguing for stated method does not get to hide its own revisions.

Your data

The diagnostic runs on files you export and send. We do not ask for access to your Search Console or analytics accounts, and we are not a processor of your live data. Files are used for the report and are deleted on request at any time.