Entity resolution
Same-name companies are the main source of diligence error, and the problem is worse across scripts. How we match, what we exclude, and what we record.
Keys first
A dossier is anchored on a registry identifier: UEN, Delaware file number and CIK, Companies Registry number, 統一編號, 法人番号, Companies House number, ACN, NZBN, corporation number, MBR number. Names are attributes of the key, not the key.
Scripts and romanisation
Chinese and Japanese names are matched in their original script; romanised forms are recorded as the company's own usage or an official transliteration, never as a separate entity. Officer names are matched across filings the same way.
Look-alikes recorded
Every dossier records the same-name or similar-name entities that were found and excluded, with the reason: different jurisdiction, different officers, different address, different incorporation date. If the exclusion is uncertain, it is an open question, not a silent choice.
Measured
Entity mismatch rate is one of five operating metrics we track; it is reported on the roadmap page.