Things I measured, including the ones that did not work

What counts as worth publishing

There is an enormous amount of SEO writing and very little of it contains a number that anybody checked. I try not to add to the pile. A piece gets published here when it has at least one of three things in it: something I counted, something I got wrong and corrected, or a method somebody else can run to get their own answer.

That rule has a cost. It means there are weeks with nothing, and it means some of what is here is duller than the industry standard. It also means that when I tell a client the median business in their market has fifteen reviews, I can show them the count.

Corrections

When a figure here turns out to be wrong, the correction goes on the page with a date on it, not into a silent edit. One of the pieces below documents a measurement I had to throw away within two hours of making it, because I checked it again and it did not hold.

Reusing any of it

Quote it, chart it, republish the figures, with credit and a link. If you want the underlying data for something here, ask and I will send it. If you re-run one of the measurements and get a different answer, I would genuinely like to know, because that is more useful to me than agreement.

The longer studies live on the agency site: the Collingwood Review Index and the Canada AI Visibility Index are both free and both carry their method.