Business surface intelligence for UK companies
Know the business behind the web.
Observes connects legal entities to the web estate they verifiably operate, and resolves a web presence back to the legal entity behind it. It records what can be observed about that estate, dated and never overwritten: an evidence-backed record of business identity, external state and history.
Verified company ↔ web identity, via API. Company in, verified web estate out. Domain in, the legal entity behind it out. Every relationship carries its evidence, its confidence class and its observation dates.
Which legal entity is behind this web presence?
A DNS record, a certificate, a mail policy or a page footer says nothing about a company until it can be attributed to one. Observes does not make a domain fact better by itself; it makes the fact attributable. "DMARC missing on a domain" is domain telemetry. "This verified UK company's web estate has no DMARC record" is company intelligence. The hard part is establishing which real legal entity operates a web presence, with evidence, and keeping that answer correct through time. That is the layer Observes builds, in both directions: from a company to the domains it can be evidenced to operate, and from a domain to the company behind it.
No accepted relationship rests on a name guess, a directory listing or a model's opinion. Finding a candidate and accepting a relationship are different stages: a name, a search result or an archived page can say where to look, and only first-party evidence read from the site itself, matched to the register, can make a relationship PROVEN or VERIFIED.
Verified business identity
The legal entity, connected to the web presence it can be evidenced to operate. Each accepted relationship carries its evidence class; the kind of evidence (the company number on the site's own pages, the registered name with the registered postcode, or the site's own statement of its operator); the source page; the first-party words found there; its observation dates; and its type where that applies: official site, group site, redirect or trading brand. Relationships on that evidence are PROVEN or VERIFIED. Weaker evidence is held at a lower class under its own name and reported apart, and a model's reading never rises above a candidate.
Company-level external intelligence
Once the identity is established, what Observes observes about that estate is attributable to the company rather than to a bare domain: DNS and name servers, mail records and mail provider, SPF and DMARC, HTTPS, TLS and certificate state, redirects, liveness and HTTP state, and the disclosure and conflict facts read from the site itself, such as a website presented for one company that names another as its operator. The value is the attribution, not the telemetry. Facts, never a score.
Evidence and history
Dated observations, not overwritten. Every fact carries two dates, when it was true and when Observes observed it, kept apart. A new observation never replaces an earlier one: corrections supersede, a false attachment is retracted, and the retraction stays on the record beside the original, so the record shows what was known and when. Keeping observations through time is the foundation for evidence-backed change intelligence, and change events built on those comparisons are pre-registered design rather than a live feed.
Where Observes records that no operating website was found, the finding carries the search receipt behind it. A blank or unresolved record is never an absence claim.
One identity layer. Several applications.
One graph, several views: the same verified identity, read for a different question.
- KYB and onboarding. Does this website actually belong to the legal entity being presented? Entity confirmed, website verified with its evidence and dates, a conflicting operator if the site names one, live or not.
- Cyber security and cyber insurance. Attribute observable cyber and email-security facts to the correct company and its verified estate: which domains genuinely belong to this business, and the mail, TLS and certificate state of each. Not a vulnerability scanner.
- Fraud and scams. Identity anomalies with evidence, never a score: the company a site claims to be against the operator its own legal pages disclose, or a domain with no evidenced relationship to the company it invokes.
- Credit and lending. Observable operating-presence intelligence alongside conventional credit data, never instead of a bureau: whether the company's evidenced web presence is there, is its own and is live, with the dates. Web behaviour does not determine creditworthiness.
- Supplier and third-party risk. You supply the company population, your own supplier list. Observes resolves each company to its verified domain and attaches the evidence and the external facts. It never discovers supplier relationships.
- Attack-surface attribution. Given a discovered domain, which legal entity can it actually be attributed to? The same identity layer, read from the domain back to the company, with the evidence.
How accuracy is measured
Precision is measured on seeded samples adjudicated blind by a human, with the sample drawn before anyone sees the answers, and reported only when a pre-registered measurement has been recorded. No figure appears on this site before its measurement exists. The method, in full.
Built on the register
- Watching the Companies House stream continuously since 30 August 2026 — seven audited days, zero unexplained gaps.
- 5,689,367 companies on the live register; 34.6 million timestamped observations.
Companies on the UK register are the identity universe: Observes is built to connect companies on the register to their verified web identity, on Companies House data under the Open Government Licence and on companies' own public websites, fetched by a crawler whose policy is published. Coverage is stated here only once its measurement is recorded.
Data that can survive being questioned.
Get early API access
Access opens by waitlist for now. Leave an address, and if you like, answer the one question we ask.