Identity resolution: definition and practice

Definition

Identity resolution establishes the subject of a research record. It connects a name, domain, location, registration, or other identifier to the business actually being assessed. In origination, the immediate question is often whether the supplied website belongs to the intended company rather than a namesake, unrelated affiliate, or acquired brand.

The term overlaps with entity resolution, but emphasises confirming the subject before collecting claims about it. A plausible name match is insufficient. Geography, service activity, legal identifiers, domain relationships, and ownership evidence can help distinguish candidates. Contradictory details should trigger investigation rather than being ignored because most fields look similar. A precise quotation about the wrong business is still unusable evidence for the intended target.

In origination work

Docket's Triage resolves identity before detailed research is paid for. The workflow should establish the correct business and website, flag shared domains, and separate unresolved records from confirmed ones. Avoid collecting a full dossier and trying to repair identity at the end. Preserve the identifying evidence so later reviewers can understand the match. If a company changes name or ownership, update relationships without erasing the earlier identity context that supported past research.

What identity resolution decides

Identity resolution answers the question: which real-world entity does this record describe? For a target company the candidates include the brand customers see, the trading name registered with a state, the legal entity that holds licences and contracts, the parent group, and the physical location. These are different things that share names. Deciding which one a record concerns is the first step in any research on a target.

The step comes before research for a practical reason: facts attach to entities, not to names. A lawsuit, a licence suspension or a government contract belongs to a specific legal entity, and applying it to the wrong one produces false findings. Resolving identity before collecting facts also avoids paying to research duplicates and wrong-site records, which is why it is the first stage of a disciplined research process.

Signals used to resolve identity

Useful signals include registry identifiers, addresses, telephone numbers, domain names, officer names, licence numbers, and tax or federal contractor identifiers where public. Identifiers that are issued by an authority and unique, such as a registration number within a jurisdiction, are stronger than names. Combinations of weaker signals can also be decisive: the same address, officers and formation date across two records is strong evidence.

Probabilistic record-linkage theory, set out in a classic 1969 paper, formalises the idea of weighing agreement and disagreement on several fields to decide whether two records match, and open-source tools implement it today. The theory also supports a third outcome beside match and non-match: a possible match that needs human review. Treating uncertain pairs as uncertain is more honest than forcing them into one of two bins.

Recording the decision

Each resolution should record the evidence used, the date and the confidence. If the evidence is a single shared phone number, say so. If the decision was made by a person, say who. When later evidence contradicts the decision, revise it and keep the earlier version, so that findings gathered under the earlier belief can be revisited.

Plan for change. Entities merge, convert, rename and dissolve, and a record that was correct last year may point at a dormant shell now. Re-check identity when a key attribute changes, and keep a continuity log that links old and new identifiers. This keeps a target's history connected across the changes that often accompany ownership transitions.

Entity resolution — Determining whether records describe the same real entity or different entities with a relationship between them.
Deduplication — Identifying and consolidating repeated records while preserving evidence, history, and genuinely distinct entities.
Evidence trail — A reviewable chain connecting a research conclusion to its supporting material, interpretation, and decisions.

Sources and further reading

A Theory for Record Linkage, Journal of the American Statistical Association (1969). The foundational probabilistic model for deciding whether two records describe the same entity.
Automatic Linkage of Vital Records, Science (1959). The earliest computerised record-linkage study, still the origin of the name-matching problem.
Splink: probabilistic record linkage, UK Ministry of Justice Analytical Services. Open-source documentation that makes the Fellegi-Sunter approach concrete.
OpenCorporates, OpenCorporates. Aggregated company registry data across jurisdictions, with provenance back to the filing source.
Legal Entity Identifier (LEI) search, Global Legal Entity Identifier Foundation. Open, free entity identifiers with parent relationships. The nearest thing to a global primary key for companies.

To apply these distinctions to your own acquisition criteria, discuss your research brief or inspect the demo platform.