Request a demo
← Back to glossary

What is entity resolution?

AI and people intelligence

Entity resolution is the process of finding records that refer to the same real-world person, company, or thing and linking or merging them, often when no shared unique ID exists. It is also called record linkage, deduplication, or data matching, and it lets teams count, search, and analyze each entity once.

How entity resolution works

A review in Science Advances, vol. 8, no. 12 (Binette and Steorts, 2022), describes entity resolution as "the process of taking large noisy databases and removing duplicate entities (often in the absence of a unique identifier)." The authors treat it as the same task as record linkage, which "seeks to bring together all relevant information about a person, business, or entity." The task is also called deduplication, data matching, instance matching, and data linkage.

The review describes entity resolution as one stage of a data cleaning pipeline:

  • Schema alignment: Records are parsed so the datasets share a common set of attributes, such as name, employer, and location.
  • Blocking: Similar records are grouped into blocks, and only records in the same block are compared, which keeps the work manageable at scale.
  • Matching: Pairs of records are judged to refer to the same entity (a link) or to different entities (a nonlink).
  • Merging: Matched records are combined into a single representative record.

Matching can use deterministic rules, probabilistic methods that build on the Fellegi and Sunter framework, or supervised machine learning. The review traces the idea to 1946, when Halbert Dunn described record linkage as assembling the pages of a person's "Book of Life" into a volume. Uses include deduplicating the U.S. decennial census, counting casualties in armed conflicts, and telling apart authors with the same name.

Why entity resolution matters for talent teams

The same person often appears many times across talent systems: as an applicant from three years ago, a referral, a CRM contact, an alum, and a public profile with a newer title. Without entity resolution, a talent pool looks larger than it is, outreach reaches the same person twice, and a candidate's history is split across records.

Errors run both ways. A false match merges two different people into one profile, which can attach the wrong experience to a candidate. A missed match leaves one person split across records, which hides relevant history. Findem's guides point to the same issue: integrating people data across systems means solving identity resolution and contact record deduplication, and workforce analytics platforms should be tested for how they handle identity resolution.

How Findem approaches entity resolution

Findem's 3D People Graph resolves information from more than 100,000 sources, including GitHub, Stack Overflow, Crunchbase, and a customer's ATS and CRM, into a single career view. Profiles in the graph are verified across sources and refreshed over time.

In Findem's Labeling Engine, career and company records are organized into structured 3D data, and duplicates, noise, and outdated information are removed before expert labeling begins. Findem's Relationship Signals unify a company's talent channels into one view, so agents can surface anyone already connected to the company, including past applicants, silver medalists, referrals, alumni, employees, talent community members, and pipeline leads.

Further reading

Related terms

Sources

Frequently asked questions.

It is the work of identifying records that refer to the same real-world person, company, or thing and linking or merging them. It is needed whenever data comes from several sources or has been entered more than once without a shared unique identifier.

The terms are used for the same task. Researchers also call it data matching, instance matching, and data linkage. Some teams use deduplication for matching within one dataset and record linkage for matching across datasets, so usage varies by field.

A typical pipeline aligns the fields across datasets, groups similar records into blocks so only likely pairs are compared, decides which pairs refer to the same entity, and merges each match into one representative record.

A false match merges two different people or companies into one record. A missed match leaves one entity split across several records. Both distort counts and search results, so teams check samples of matches and review uncertain cases.

Findem's 3D People Graph resolves information from more than 100,000 sources, including a customer's ATS and CRM, into a single career view. Its Labeling Engine removes duplicates, noise, and outdated information before expert labeling, and profiles are verified across sources and refreshed over time.