Entity-resolution and graph-analysis software

Entity-resolution and graph-analysis software links records that may refer to the same real-world actor and represents resolved actors and relationships as a network. It is dual-use analytic infrastructure for compliance, investigation, fraud detection, supply-chain analysis and, where lawful authority exists, economic-statecraft planning. A match, confidence score or graph relationship is an analytic lead, not a legal finding.

Function

Entity resolution confronts the deliberate messiness of financial identity. The same Russian procurement agent appears as different transliterations across registries; a sanctioned firm re-registers under a cognate name; a network of shelf companies shares only an address, a nominee director, or a phone number. Resolution engines match records probabilistically on names, identifiers, addresses, and relationships, scoring the likelihood that clusters refer to one actor, with machine learning increasingly supplementing rule-based matching. The resolved output feeds a graph database in which nodes are entities, accounts, vessels, and people, and edges are ownership, directorship, transaction, and logistics relationships. Graph queries then answer the operational questions: who ultimately owns this entity (UBO resolution); which broker connects otherwise separate procurement clusters (betweenness centrality); which single bank, agent, or facility would fragment the network if removed (chokepoint identification). The approach entered public consciousness through the ICIJ's offshore-leak investigations, which used graph tooling to navigate millions of Panama Papers records.

Strategic significance

Graph tooling can reveal intermediaries, ownership chains and dependencies that flat records obscure. Weaponised interdependence directs attention to network position, while the Economic Kill Chain treats mapping as an analytic phase. The same software supports ordinary compliance: applying OFAC's 50 per cent rule may require aggregating ownership interests, and screening can compare customers with relationship data as well as flat lists. Use for analysis or targeting remains subject to the quality of the records, the governing legal test and the authority of the decision-maker.

Limits

Resolution is probabilistic and adversarial. False merges can put the wrong actor under sanction; missed links let structured evasion pass; and every published designation teaches network designers which signatures to randomise, so the data environment degrades in response to its own exploitation. Ground truth is scarce where it matters most: secrecy jurisdictions, informal transfer systems, and state-run evasion networks generate few resolvable records, leaving graphs that are dense where the world is transparent and sparse where the targets are. Practitioners accordingly treat graph output as hypothesis generation for human analysis, not as targeting authority in itself. A resolution engine's confidence score is not a legal standard of proof. Analysts must validate the underlying records, resolve contradictory identifiers and distinguish legal ownership tests from graph inference before action.

See also

Financial intelligence (FININT) · Financial intelligence (FININT) analytical platforms · Beneficial ownership registries and databases · Chokepoint identification · Weaponised interdependence · Economic Kill Chain (EKC) · Panama-Paradise-Pandora leak exploitation · Sanctions list and watchlist screening technology · Economic statecraft

Sources

• OFAC 50 Percent Rule guidance.
• FATF beneficial-ownership guidance.
• NIST IR 7284.
• ICIJ Offshore Leaks Database.