How to build a query corpus worth measuring
A measurement is only as good as the list it runs on. Here is how ours is built, and why it is generated rather than curated.
6 Aug 2026
The temptation with a keyword corpus is to export ten thousand rows from a keyword tool and start measuring. It produces a large number quickly and a useless series slowly, because you cannot say what the list was last month.
Generated, not curated
Ours expands from 24 subject areas and a list of named products and concepts, through a fixed set of intent templates, into 4,185 queries. The generator is deterministic: run it twice, get the same list. That single property is what makes week 2026-W25 comparable to week 2026-W36 by construction rather than by hope.
Intent assigned on entry
Every query carries its intent — commercial, informational, comparison, navigational, support — assigned when it enters the corpus and never re-derived afterwards. That is why the intent splits on this site stay stable while the results underneath them move. Re-classify intent from the results and you have built a mirror, not a measurement.
The audience test
A corpus of consumer queries would attract traffic that cannot buy a search API. Ours is deliberately confined to the subject matter its readers work in. The narrowness is the point: coverage figures here describe developer and marketing queries and should not be read as coverage of Google overall, and we would rather say so than imply a generality we did not measure.
The threshold
Not every query gets a page. A query needs enough organic results to be worth describing and at least two weekly observations, because a page with a single measurement offers nothing a free rank checker does not. The uniqueness lives in the series, so publishing before there is a series is publishing a thin page on purpose.
Full method: how we measure.
Run the same measurements yourself
Free tier, no card. Every figure above came from one endpoint on a weekly schedule.