Every finding has to survive someone trying to refute it.
The framework, the 100-point scale, and the two steps most audits leave out: hostile verification, and controlling the instrument before quoting a number.
The framework, the 100-point scale, and the two steps most audits leave out: hostile verification, and controlling the instrument before quoting a number.
Five letters, and every one of them is a list of things that can be measured. The framework exists so that a diagnosis has an address: a finding belongs to a channel, and a channel has a defined scope.
The question it answersCan these systems read the site at all — and does the site say the same thing twice?
The question it answersDoes the catalogue structure match how people shop it — and does the feed agree with the site?
The question it answersDoes each market get a page written for its own demand, or a translation of another market’s?
The question it answersWhen a system answers instead of linking, is the answer yours?
The question it answersCan a generative system state a fact about this brand and defend it?
The weights are fixed and published so that a score means the same thing between engagements. Technical foundation carries the largest single weight because it is the only layer whose failure is total.
These are the expensive parts and the invisible ones. They are also the reason the figures in an Atlas report can be quoted back to us — read five ways to measure Core Web Vitals wrong for what happens without them.
Atlas reads the site before the first conversation.