Entity Resolution Explained
Entity resolution is the process of determining whether records that look different actually describe the same real-world person, organisation, property or other entity. It is a foundational capability because poor matching can create false relationships, while missed matches can hide important connections.
Why names are not enough
Names vary because of spelling, abbreviations, transliteration, trading names and formatting. Two identical names can also belong to different entities. Reliable resolution therefore combines multiple attributes rather than relying on a single string match.
Useful matching signals
Depending on the use case, signals may include registered names, company identifiers, addresses, domains, public contact details, dates, directors, locations and other stable attributes. Each signal should be weighted according to its reliability and the risk of false matching.
Deterministic and probabilistic approaches
Deterministic rules use known identifiers or exact relationships. Probabilistic approaches estimate whether records are likely to represent the same entity based on multiple similarities. Both can be useful, and both require thresholds and review processes appropriate to the consequences of an error.
Human review matters
High-impact matches should remain reviewable. A system should be able to show why records were linked and which evidence supports the link. This is particularly important where similar names, shared addresses or common directors could produce misleading associations.
Keeping identity current
Resolution is not a one-time task. Entities change names, addresses, domains and ownership structures. A useful system stores observations over time and can revisit earlier matches when new evidence changes the picture.