UK Companies House Company Number

Detects UK Companies House company registration numbers (CRNs). A company number is always 8 characters: either 8 digits (England & Wales incorporations), a 2-letter jurisdiction/entity-type prefix followed by 6 digits (e.g. SC Scottish company, NI Northern Ireland company, OC/SO/NC LLPs, RC/SR/NR royal-charter bodies, FC overseas companies, OE overseas entities), or R0 + 6 digits (pre-partition Northern Ireland companies, registered under the R0 prefix). The prefix allow-list is taken from the Companies House Uniform Resource Identifiers (URI) Customer Guide plus the post-2011 additions CE/CS (charitable incorporated organisations), SG (Scottish qualifying partnerships), OE (Register of Overseas Entities, 2022), RS (registered societies, e.g. the live entry RS008375) and FE (further education / sixth form college corporations). Bare 8-digit runs collide heavily with invoice, order and account numbers, so the all-numeric form is only ever matched when directly attached to a company-number label; the more distinctive prefixed form additionally gets a keyword-gated tier without a directly-attached label. Two allow-list prefixes are excluded from that bare prefixed form because they collide with common document-number abbreviations — NO ("INVOICE NO123456") and SO (ERP sales-order numbers such as SO123456) — and are matched only via the label-attached regex.

Type: regex
Engine: boost_regex
Confidence: medium
Confidence justification: Medium confidence: the 8-character structure is rigid and the prefix allow-list is verified against Companies House documentation, but there is no check digit, and the all-numeric form is indistinguishable from any other 8-digit reference without its label. Every exported tier is evidence-gated — the all-numeric form requires a directly-attached company-number label (label-embedded-in-regex counts as label gating per the D1 decision record), and the bare prefixed form additionally requires Companies House keywords in proximity. The bare prefixed form also drops the collision-prone NO and SO prefixes (document-number abbreviations that appear near generic keywords like "registered office" in UK letter footers); those two prefixes are covered only when label-attached. There is deliberately no bare unlabelled tier.
Jurisdictions: uk
Regulations: Companies Act 2006, UK GDPR
Frameworks: ISO 27001, ISO 27701, SOC 2
Data categories: business-identifier
Scope: narrow
Risk rating: 5
Platform compatibility: Purview: Compatible, GCP DLP: Unsupported, Macie: Unsupported, Zscaler: Compatible, Palo Alto: Unsupported, Netskope: Unsupported

Pattern

(?i)\b(?:company\s+(?:registration\s+|reg(?:istration)?\.?\s+)?(?:number|no\.?)|(?:companies\s+house|registered)\s+(?:company\s+)?(?:number|no\.?)|companies\s+house\s+registration|incorporation\s+number|CRN)\s{0,3}[:#.-]{0,3}\s{0,3}(?:(?:AC|CE|CS|EN|ES|FC|FE|GE|GN|GS|IC|IP|LP|NA|NC|NF|NI|NL|NO|NP|NR|NV|NZ|OC|OE|RC|RS|SA|SC|SE|SF|SG|SI|SL|SO|SP|SR|SZ|ZC)\d{6}|R0\d{6}|\d{8})(?![A-Za-z0-9])

Corroborative evidence keywords

Companies House, company registration number, company number, CRN, certificate of incorporation, registered office, registered in England and Wales, data record, database record, record set, data extract, data export, database table, spreadsheet, data registry, registry entry, master data, bulk export, data dump

Proximity: 300 characters

Should match

Company registration number: 04352950 — Labelled all-numeric England & Wales company number
Company No. SC123456 — Labelled Scottish company number with SC prefix
CRN: OC334521 — CRN-labelled England & Wales LLP number with OC prefix
Registered number 12345678 — Registered-number label with 8-digit company number
Companies House number NI070546 — Companies House label with Northern Ireland prefix
Incorporation number: R0123456 — Pre-partition Northern Ireland company number (R0 prefix + 6 digits)
OE000123 — Bare Register of Overseas Entities number (prefixed structural regex, keyword-gated tier)
RS008375 — Bare registered-society number — live Companies House entry (Nurture Housing Association Limited), RS prefix

Should not match

Invoice number: 12345678 — 8-digit run attached to a non-company label — the all-numeric form is label-gated
Company number: 1234567 — Only 7 digits, too short for a company number
Company number: 123456789 — 9 digits — the trailing alphanumeric guard rejects longer digit runs
Company number: AB123456 — AB is not a Companies House prefix — fails the verified prefix allow-list
Company number: R1234567 — R followed by a non-zero digit — only the documented R0 + 6 digit form is a valid pre-partition Northern Ireland number
INVOICE NO123456 — NO as the "number" abbreviation in an invoice header — NO is excluded from the bare prefixed regex and only matched label-attached
SALES ORDER SO123456 — SO as an ERP sales-order document prefix — SO is excluded from the bare prefixed regex and only matched label-attached

Known false positives

Unlabelled 8-digit sequences (invoices, order numbers, phone fragments) share the all-numeric company number shape exactly, and sentinel values such as 11111111 or 00000000 satisfy it too. Mitigation: The all-numeric form is only matched when directly attached to a company-number label; AllDigitsSameFilter drops repeated-digit sentinels. There is no bare unlabelled tier.
Two-letter + 6-digit codes from other schemes (ticket references, product SKUs, non-UK registries) can coincide with the prefixed form, e.g. SC/NI also appear in unrelated codes. Mitigation: The prefixed structural regex is restricted to the verified Companies House prefix allow-list and is only exported behind Companies House keyword evidence within the proximity window. In the exported SIT the prefix match is case-sensitive (prefixes are always upper case). NO and SO are additionally removed from the bare prefixed regex — NO is the ubiquitous "number" abbreviation (INVOICE NO123456, ACCOUNT NO123456) and SO is a common ERP sales-order document prefix, both of which co-occur with footer keywords such as "registered office" — so those two prefixes match only in the label-attached form.