A mechanism, or nothing
TraceableEvery claim traces to something the product measures or something the code does. If we cannot point at what produces the number, the post does not run.
Writing
Mechanism, or we don’t publish it.
Answer-engine optimisation from the instrument’s side. Nothing is published yet.
Nothing to list
No post is published yet.
An empty archive is a cheap thing to fix — you write four posts of opinion and the page looks alive. The four rules below are what stop that being the plan.
Instead
There is already something to read.
The material that would have been a first post is already written, as product documentation rather than as an essay. Drier, and more useful.
Published today
Three pages that exist.
Rather than three posts that do not.
- Methodology 0.4 / 0.6
- See the engines 5 engines
- What AEO is Primer
- Help center Guides
The bar
Four conditions, all of them awkward.
The sample travels with the number
10 minimumThe app refuses a confidence band under ten samples. Writing is held to the same rule: window, engines and sample size come attached, or the figure stays out.
No causal language
CorrelationAttribution is correlation-based, not a randomised controlled trial. That sentence is in the published methodology and does not get quietly softened.
Disagreements get printed
Both shownWhere our own sources disagree, a post says so and shows both, rather than quoting whichever number happens to flatter us.
Readiness is weighted 0.4 and visibility 0.6, over a rolling 21-day window. A post that cannot say which of those it is talking about is not ready to run.
Start here
Your own store is the first data point.
The free audit reads your store the way the product does and shows you the answers it got back.