Methodology
The numbers are only as good as how you measure them
When you tell your board 'AI recommends us twice as often', someone will ask how you know. This page is that answer. We publish how we measure — so every number we hand you can survive the question.
Numbers you can defend to your board
A visibility score with no n is a guess dressed as a fact. We report the denominator behind every number, so you can tell a real signal from noise.
±CI shown with every θ · Illustrative
We test what each market actually sees
Who's asking, where the service is, the simulated vantage, the actual egress node, and the UI/answer language are five different things. Collapsing them hides the truth. We keep them separate.
five layers, never collapsed
Both versions of every AI, checked
An engine's API answer and its app answer draw on different source pools. We sample both and calibrate the difference — a distinction competitors don't even acknowledge.
Δ sampled on both channels, difference calibrated
Your data gets more valuable every year
We store the raw answer with an integrity hash, versioned by extractor. As extraction improves, the entire history can be re-analyzed — so your data appreciates rather than decays.
history re-analyzed as extraction improves
Every number has a receipt
Every claim replays to the source answer. Auditability is a first-class feature: when a client's procurement asks 'where did this number come from', we put the chain on the table.
every claim replays to its answer — the chain goes on the table
The method is public. The years of tracked answers are not.
Measurement is where an engagement starts, not where it ends — our team takes the findings and does the optimization work. The method is open; the accumulated answer archive behind it is earned, not for sale.
Get a free diagnostic