Methodology
How Fylings sources, verifies and scores company data
Every record on Fylings comes from an official public source and carries its provenance: the source, the date we last checked it, and a confidence rating. This page explains exactly what that rating means and how each part of a company page is produced. Fylings never fabricates data — where a registrar does not publish a field, we leave it blank rather than guess.
Where the data comes from
Core company records are read from the official company registrars of each country we cover (for example Nigeria's CAC, Tanzania's BRELA, Zambia's PACRA, Mauritius's CBRD, and the OHADA regional RCCM). Records are then enriched from other official systems where a match is found: the Global LEI Foundation (GLEIF), stock-exchange disclosures, intellectual-property registers (TMView: OAPI and ARIPO), and national open-contracting (OCDS) procurement data. Each enrichment names its own source and links to it.
The confidence rating
Confidence is a signal-weighted rating of how much we know about a record and how well it is corroborated — not a judgment of the company itself. A dissolved company can carry a High-confidence record; a genuine company can carry a Baseline one if its registry publishes little. Every record starts from the same base — an official registry record exists — and earns additional weight from four groups of signals:
- Source freshness. How recently we re-verified the record against its registry: checks within the last week count most, within the last month less, within the last year less again.
- Field completeness. Whether the registry publishes — and we hold — the record's activity, registered address, incorporation date, tax ID, and a meaningful status. A fuller record is a more trustworthy one.
- Cross-source corroboration (the strongest signals). Whether a second official system agrees the entity exists: a GLEIF-verified LEI, a stock exchange listing (with its disclosure obligations), a match in an official trademark register, or a match to disclosed government procurement awards.
- Registered charges. Where a registry publishes security interests, holding that data adds a small weight.
The combined score is capped: it never reaches 100% — public-registry data is never perfect — and never falls below a floor that reflects the fact an official record exists at all. On company pages we show the rating qualitatively:
- High — at least one other official system corroborates the record: a GLEIF-verified LEI, a stock-exchange listing, a match in an official trademark register, a match to disclosed government procurement awards — or the company itself has claimed and verified the profile.
- Good — a complete core registry record: a meaningful status, an incorporation date, and a registered address or activity — but no external corroboration yet.
- Baseline — the registry publishes few fields for this record (often just the name, number and type). The company is real; the source is thin.
The exact numeric score is available through the Fylings API on every record, where it belongs in an automated pipeline. The ⓘ popover on each company page itemizes exactly which of these components that record earns.
Source tiers
Each record is labelled with where it comes from. Official registry verified means the record was read from the country's company registrar. Registry + LEI verified means the registry record is additionally cross-confirmed by GLEIF. Official regulator source and official identity source (LEI) mark records compiled from regulators, exchanges, or GLEIF where the company register itself is not publicly readable (for example Kenya). Vendor stubs from procurement data — suppliers named in government contract awards but not yet matched to a registry record — are labelled as such and deliberately carry low confidence.
Verification dates
“Last verified” is the date we last read the record from its official source — not the date the source last changed. Live-fetched fields carry their own dates: a Nigerian tax ID fetched from CAC is stamped with the date it was retrieved, and sanctions screening results carry the date the name was screened.
Sanctions screening
Company names are screened against the OFAC SDN, OFAC Consolidated, and UN Security Council lists, refreshed daily from the official publications. Matching is by name similarity: on a company's own page we only surface near-exact matches (so a page never flags on a weak, generic-word collision), while the screening tool shows weaker possible matches for active investigation. A possible match is never an accusation — name collisions are common, and results are framed for review against the company's full details.
Inferences are labelled
Anything Fylings derives rather than reads from a source is labelled as an inference — for example the sector we classify from a company's registered name and activity. Where an inference disagrees with the registry's own data, we show the discrepancy instead of silently picking one.
Corrections
Spot an error? Email hello@fylings.com with the company's registration number and what's wrong, and we'll re-verify against the source.