Methodology

A source and its evidence trail are retained first; a fact or observation is then established and attached to the subject it describes. Only a record that passes publication rules becomes a public page. AI may assist extraction, classification, consistency review and translation, but it is not a factual source and does not decide truth on its own.

The register describes businesses from publicly verifiable business sources. Verified records with sufficient public value may receive a standalone page; probable records remain compact listings; unresolved records stay in the research record only. The research database is therefore not the same as the public register. Every public entry states why it is in scope.

Structured records become localized public copy through defined presentation templates; internal notes and technical locators are not published as prose. Claims link to the sources actually used. A reported material error is checked against evidence, corrected in the main record and, when an address changes, the old address is permanently redirected.

Offer and availability observations are retained as timestamped evidence in the acquisition layer. They are not published as business attributes, seller lists or current-stock claims.

Automated collection checks robots.txt, terms and source policy before fetching. It never bypasses login, CAPTCHA, WAF, rate limits or other restrictions. The site does not mirror whole third-party pages.

At launch, public pages do not publish a market index or per-kilogram price. If that changes, every value must carry source, date, price type, unit, normalisation and publication status; an advertised price will not be called a transaction price.