Content coverage

A second, separate measurement: which page roles a site has, and which it lacks.

Not part of the total

The coverage score does not feed the Agent-Readiness total. It is reported alongside it and it generates fixes, but it moves no points. Folding it in would move every historical customer number at once, which is a product decision rather than a refactor.

Formula

score = round(earned_weight / total_expected_weight × 100)

Role weights

Role	Weight
`pricing`	10
`catalog`	10
`comparison`	9
`faq`	8
`about`	7
`proof`	7
`editorial`	6
`how_it_works`	6
`contact`	5
`policies`	4

Expected roles by archetype

Archetype	Expected roles	Total weight
`service`, `saas`	pricing, comparison, faq, about, proof, editorial, how_it_works, contact, policies	62
`ecommerce`	catalog, faq, about, proof, editorial, contact, policies	47
`unknown`	about, contact, faq, policies	24

unknown expects the least, because claiming a site lacks a comparison page when you cannot tell what kind of site it is would be a guess dressed as a measurement.

Confidence and withholding

Confidence is medium when pages were found only via homepage links, high otherwise.

The score is withheld (null) in two cases:

Discovery method was home_only — nothing but the homepage was reachable.
The discovered-URL list was truncated. Here the roles found are still reported; only the claim about what is missing is withheld.

Both cases are the same principle as the evidence gates: you cannot report an absence you were never in a position to observe.

Fixes

Each missing expected role becomes a fix on the content_estate channel, with an expected gain equal to that role's weight.