Match, dedupe, and enrich every customer, supplier, and prospect record — resolved to its official registration number, with the filing behind each field and the date it was read. Stewards decide on evidence, not on a score.
0M+
Company profiles, resolvable to a registry entry0+
Government registries read directly at source0+
Countries and territories, including thin-file marketsDaily
Refresh cadence, with change events pushed to your systemsIt degrades quietly — a near-duplicate here, a stale status there, a field nobody can trace. You feel it when finance can't roll up spend to a parent, or an auditor asks where a number came from and the file has no answer.
"Acme Ltd", "ACME LIMITED", and "Acme Mfg." are one company to a human and three to a database. Keying identity on a string guarantees drift.
Sales, procurement, billing, and compliance each created their own record with its own spelling and ID. Without a shared key, one customer is counted four times and governed zero times.
Companies change name, address, directors, and status. A record that was right at onboarding quietly becomes wrong — and re-cleansing projects only reset the clock.
A vendor appends revenue and headcount, but not the filing it came from or when it was read. The field can inform an opinion but cannot support an audit.
Spend, exposure, and risk sit at the legal entity but are managed at the group. When parent-child linkage is missing or stale, the ultimate parent's true exposure never rolls up.
A black-box score says 87 and nothing more. Stewards can't see which attributes agreed, so they review everything by hand — which is where most match automation quietly dies.
Matching is a pipeline rather than a single score. Every stage exposes what it did, so a low-confidence result tells a steward what was missing instead of leaving them to guess.
Legal-form suffixes, punctuation, diacritics, transliteration, and address formats are normalised using rules specific to the country — because "Ltd", "GmbH", and "S.A." carry different meanings and different collision risks.
The normalised record is blocked against 600M+ registry profiles on name, address, domain, officer names, and any identifier you supplied. Supplying a registration number or postcode is what moves a record from the review band into auto-merge.
Each candidate is scored attribute by attribute, and the agreeing attributes are returned alongside the number. A score of 0.93 arrives with the reason it is 0.93 — and with the runner-up it was chosen over.
Above your auto-merge line, records merge and write back. Below it, they queue for a steward with the evidence already attached. Decisions are retained, so the same ambiguity isn't reviewed twice at the next refresh.
Up to 40 fields per company, grouped into six categories — each read from the registry it was filed with and carrying its source. Modelled values are labelled as modelled.
Up to 40 fields per record. See the full data dictionary →
Collapsing four records into one customer is only half the job. If the golden record has no parent, finance still can't see that six separate vendors belong to one group — so negotiated terms leak, concentration looks lower than it is, and credit limits are set against the subsidiary instead of the group.
Group structure is built from ownership filings and officer registers across jurisdictions, so parents, subsidiaries, and shared beneficial owners resolve to the same identifiers as the rest of your master data.
Registries are read daily. When something material changes on a company you hold, the delta is pushed into your systems as an event — so the golden record is corrected before the next quarter's reporting, not during it.
Most enrichment suppliers resell a chain of upstream vendors, then wrap the result in an identifier you license. We read the registry, keep the trail, and hand you keys you already own.
Company attributes are read directly from the registry of record and 400+ official sources. You see the filing behind the field, rather than a value that has passed through three resellers on its way to you.
Resolution keys on registration numbers and LEIs — public identifiers. There is no proprietary ID embedded in your estate to license annually or unpick if you move.
Provenance is attached at field level as data is written, not reconstructed on request. When governance asks how a value was derived, the lineage is already in the record.
If you embed company data in a product your own customers use, licensing can include reseller and redistribution rights — a conversation that ends most enterprise procurement cycles early elsewhere.
Match rates hold up outside the easy jurisdictions. 200+ countries including the thin-file and emerging markets where supplier and customer records are hardest to resolve.
Statista scales its own B2B company intelligence product on data sourced directly from official government registries — a master dataset it can trust as the single source of truth downstream.
Read the case studyMatch and enrichment run where your master data already lives — through the API, as scheduled batches, in the warehouse, or directly inside the CRM.
Resolve a record on write. Send a name and country, receive the entity, identifier, confidence, and matched attributes.
Match an existing file at volume, or land resolved records and change events straight into your warehouse on a schedule.
Dedupe accounts, fix parent-child structures, and enrich on create — inside Salesforce, HubSpot, or Dynamics 365, with the source written alongside every field.
Connect Global Database to any MCP-compatible client or agent through the Model Context Protocol — company data queried in natural language, with sources attached.
Ask master-data questions in plain language — duplicates by group, unresolved records, what changed this week — with sources attached.
Global Database's built-in analyst. Ask for duplicate groups, unresolved records, or what changed this week — answered in-platform, traceable to the filing.
Send us a sample of your account or vendor file. We'll return it resolved, deduplicated, and enriched — with the source and timestamp on every field, so you can judge the match rate on evidence.