Column · 2026-10-19
A number is not evidence by itself. It becomes evidence when you can say what it would be if you were wrong. Without that sentence, a number is decoration.
Before arguing about a figure, ask what its unit is. Most disagreements about market size are actually disagreements about units that nobody stated.
Each of those pairs looks similar in a slide and behaves completely differently in a decision.
For each of 132 areas of a life, we record how many distinct phrasings people use to ask about it on a given day. A figure is a breadth of language on a date.
That is all. It is a real measurement, taken the same way every day, and comparable across areas on the same day.
They are not search volume. Suggestion breadth does not convert to traffic, and every conversion factor is invented. We do not publish one.
They are not a forecast. A change needs two days; we publish both days and nothing beyond them.
They are not representative. The endpoint we read is weighted toward one country and one language, and suggestions are personalised, so a reader elsewhere may see a different set the same day.
For any figure you plan to act on, write this:
> *This number would tell me I am wrong if it were below ___ , and I would change course.*
If you cannot fill the blank, the number is not yet a measurement — it is a mood with digits.
Because the flattering version is the one that gets quoted, and quoted numbers acquire authority they did not earn. A figure that survives being quoted only because nobody checked its unit is a liability, not an asset.
The discipline is simple and unpopular: state the unit, state the date, state what it does not cover, and let the reader decide whether it is enough.
---
*The method, the limits, and the full record are on the data page. Free CSV, no account.*
See the data this is built on
Open the snapshot