Data integrity
How this directory stays true.
Sources before claims.
Every record in the ecosystem directory is re-checked on a schedule against authoritative sources. Machines can confirm facts and flag change. They cannot rewrite what a record says about an organisation — a person does that.
— records under continuous verification
Sources
Authoritative sources only, and only where automated access is permitted.
An organisation's own website is read only after its robots.txt has been fetched and checked. If robots.txt disallows automated access, or cannot be retrieved, the site is never requested. No source is scraped around a paywall, login or rate limit.
What updates automatically
Auto-approved (low risk)
Website URL
Auto-applies only when the change is a redirect within the same registered domain (scheme, host prefix or trailing path).
Confidence required: 90%
ROR identifier
Identifier reconciliation against the Research Organization Registry. Adds an identifier, never changes prose.
Confidence required: 95%
Wikidata identifier
Identifier only. Used to cross-check other sources, never displayed as a claim.
Confidence required: 95%
Always reviewed by a person
Website domain
A different domain can mean acquisition, rebrand or a dead link. Always reviewed.
Organisation name
Renames are reviewed by a person.
Trading status
Dissolution, liquidation or closure is always reviewed.
Location
Location drives cluster placement on the map.
Cluster
Changes where an organisation appears on the map.
Industries
Changes directory classification and matching.
Company stage
Interpretive judgement, never automated.
Funding
Never automated under any confidence.
Description
Editorial copy is written by a person.
Each detected difference is stored with its source URL, the value found, the value currently published, the time it was checked, a confidence score and a review status. Nothing is discarded, so every published change can be traced back to its evidence.
Moderation queue