Entity Resolution
Entity resolution decides whether different records refer to the same real person or organisation, so risk is assessed on the entity, not on fragments.
Definition
Entity resolution is the process of determining whether two or more records describe the same real-world person or organisation, and linking them when they do. It is sometimes called record linkage or identity resolution.
The records may sit in one system or many: a customer file, a payment message, a company registry extract, a sanctions list entry, a news article. Each carries a partial and sometimes inconsistent view. Entity resolution weighs the evidence across them (names, dates of birth, addresses, registration numbers, identity documents, relationships) and concludes "same", "different" or "not enough information".
It is related to, but broader than, fuzzy matching. Fuzzy matching tells you how similar two names are. Entity resolution uses that score as one input among several to decide whether two records are the same entity.
Why a regulated firm cares
Compliance decisions are made about entities, but data arrives as records. If one customer exists as three records with slightly different names, their activity is split three ways and no single view shows the full picture. If two different people are merged because they share a common name, one may be wrongly refused service.
The consequences show up in several places:
- Screening. A match against a list entry is only meaningful once it is resolved to the listed person or ruled out.
- Ownership. Aggregating ownership across a corporate structure requires knowing that the same shareholder appears in several layers, perhaps under different spellings.
- Monitoring. Transaction patterns, alerts and prior decisions should attach to the entity, so a new analyst sees the history.
- Evasion. Schemes that rely on slight identity variations are harder to run when records are consistently linked.
How it is evidenced in practice
Entity resolution in compliance is usually a mix of automated scoring and human judgement, and firms keep evidence of both:
- Defined matching rules. Which attributes count as strong identifiers (for example a national identity number or company registration number), which are supporting (address, nationality), and how conflicts are treated.
- Resolution outcomes with reasons. Each merge, split or "not the same" decision records the attributes relied on, so it can be reviewed later.
- Handling of uncertainty. A clear route for records that cannot be resolved with confidence, such as requesting further documents from the customer.
- Reversibility. The ability to undo a wrong merge without losing the underlying source records.
- Quality checks. Periodic sampling of resolved clusters to find over-merging and under-merging.
Regulatory basis
No regulation prescribes entity resolution as a technique. The requirement is indirect: several obligations cannot be met without it.
FATF Recommendation 10 requires customer due diligence, including identifying and verifying the customer and the beneficial owner. FATF Recommendations 24 and 25 concern the transparency of beneficial ownership of legal persons and arrangements. FATF Recommendations 6 and 7 require targeted financial sanctions to be applied to designated persons and those acting on their behalf, which depends on correctly identifying who a party is. OFAC's 50 Percent Rule treats entities owned 50 percent or more in aggregate by blocked persons as blocked, so a firm must recognise the same blocked owner across different holdings. In the UK, sanctions regulations under the Sanctions and Anti-Money Laundering Act 2018 similarly extend asset freezes to entities owned or controlled by designated persons.
This summary is not legal advice. Firms should confirm how these obligations are applied in their jurisdiction.
Worked example
A hypothetical corporate bank reviews a client, Quillmere Logistics. The ownership chart shows a 30 percent shareholder recorded as "M. A. Sorensen", and a further 25 percent held by a holding company wholly owned by "Mikkel Andreas Sørensen". A third record, from the bank's private banking arm, holds "Mikkel Sorensen" with a full date of birth and passport number.
Taken separately, each record looks like a minority stake held by a different person. The bank's resolution process compares them: the initials agree with the full name, the "ø" and "o" spellings are a common transliteration pair, and the registry filings for the shareholding and the holding company give the same residential address and date of birth as the private banking record. The analyst resolves all three to one individual and records the evidence.
The client's indirect ownership is now attributed to one person with a combined 55 percent stake, direct and indirect, which changes the beneficial ownership record, triggers screening of that individual in the corporate context, and informs the client's risk rating.
Related terms
Related Articles
Account Takeover (ATO)
A form of identity theft where a fraudster gains access to a user's bank accounts, email, or other services.
Adverse Media
Adverse media is published information linking a person or company to crime or serious misconduct, screened as part of customer due diligence.
Alert Rules
Configured parameters within AML software that automatically trigger an alert when suspicious activities match defined criteria.
Algorithmic Fraud
Automated and programmatic approaches to attempting financial fraud at scale utilizing bots and specialized software.