Publication model

How it works

Publication is decided twice: automated rules establish a ceiling, then an owner decision may demote it. A separate content-bound fingerprint is required before any product route can be indexed.

Four effective states

Capabilities belong to fields

A product can remain useful while a particular value is pending, absent, ambiguous, conflicting, noncanonical, stale, or not standards-attributed. Display, filter, compare, chart, and current-market capabilities are evaluated separately.

Market comparisons answer buying questions

Product pages lead with the reported value and a plain-language comparison with “Typical.” An open-on-tap explanation defines Typical as the middle value, states the data rule, and lists every current pad included. Each comparison names its product count and appears only when at least five pads can be compared.

Paired facts use intersection

A chart point or deterministic difference exists only when every constituent supports that operation and their variant scopes are compatible. Pairing a valid field with an invalid one cannot make the invalid field canonical.

Owner feedback stays separate

The pilot records only a rating aggregate, review count, exact-model scope, source, and observation date from a brand-owned product page. It cannot affect field capabilities, sorting, filtering, market comparisons, recommendations, or publication.

Build-time aging

Current price functions use a 30-calendar-day evidence window and the actual UTC build clock. Static production artifacts must be regenerated at least every 14 days so stale values can age out.