Request a demo
← Back to glossary

What is a knowledge graph?

AI and people intelligence

A knowledge graph is a way of organizing data as a network of entities, such as people, companies, and places, connected by named relationships, such as "worked at" or "invested in." Because facts are stored as connections, software can follow paths between entities, combine data from many sources, and answer questions that span them.

How a knowledge graph works

Definitions vary. A widely cited survey in ACM Computing Surveys, vol. 54, no. 4 (Hogan et al., 2021), notes that the definition "remains contentious" and that definitions range from specific technical proposals to more inclusive general ones. The authors define a knowledge graph as "a graph of data intended to accumulate and convey knowledge of the real world, whose nodes represent entities of interest and whose edges represent potentially different relations between these entities."

In practice, each fact is a connection between two things. The W3C's RDF 1.1 Concepts standard expresses facts as triples of subject, predicate, and object, and a set of triples forms a graph. "Maria, worked at, Acme" is one triple. Add "Acme, acquired, Beta" and software can follow the path from Maria to Beta. Other graph models, such as property graphs, attach details like start and end dates directly to a connection.

The survey describes several reasons teams choose a graph:

  • Flexible structure: The schema can evolve as new kinds of data arrive.
  • Path queries: Query languages can find entities connected through paths of any length.
  • Shared meaning: Ontologies and rules define what the terms in the graph mean, so software can reason over them.
  • Integration: Data from diverse sources can be combined around the same entities.

The phrase has been used since at least 1972, and its modern use dates to Google's 2012 announcement of its Knowledge Graph, followed by knowledge graphs announced by Airbnb, Amazon, eBay, Facebook, IBM, LinkedIn, Microsoft, Uber, and others.

Why knowledge graphs matter for talent teams

People data is spread across an applicant tracking system, a CRM, an HRIS, referral lists, and public sources. Each system describes part of a person's career. A knowledge graph links those parts to the same person and connects them to companies, roles, and other people, so a question such as "who has led a sales team through a company's expansion and is connected to one of our employees" can be answered from connected records instead of separate searches.

A graph is only as good as its connections. If two records for the same person are not recognized as one, the graph splits that career in two. If time is missing, a past role can look like a current one. That is why identity matching and dated relationships matter as much as the volume of data.

How Findem approaches knowledge graphs

Findem's 3D People Graph connects trillions of data points across individuals, companies, and time. It is built from more than 100,000 data sources and combines a customer's ATS and CRM data with career histories, contributions, publications, patents, funding events, and company milestones.

The graph organizes a career the way it unfolds, across roles, teams, relationships, and company milestones over time. Findem calls the underlying model 3D data: person, company, and time. Profiles in the graph are verified across sources and refreshed over time. Findem's Labeling Engine then applies expert labeling to that structure, producing Success Signals about experience and fit and Relationship Signals about how people are connected. Read more about the People Intelligence Platform.

Further reading

Related terms

Sources

Frequently asked questions.

It is a way of storing data as things and the connections between them. People, companies, and places are the things, and relationships such as "worked at" or "acquired" are the connections. Software can follow those connections to answer questions that span many records.

A relational database stores data in tables with a fixed schema and links them through joins. A knowledge graph stores entities and relationships directly, lets the schema evolve as new data arrives, and supports queries that follow paths of any length between entities.

A triple is one fact written as subject, predicate, and object, such as "Maria, worked at, Acme." The W3C's RDF standard uses triples as its basic unit, and a set of triples forms a graph.

They are used to integrate, manage, and draw value from diverse sources of data at large scale. Common uses include search, recommendations, and analysis that depends on how entities are connected, such as how people, companies, and roles relate over time.

It is Findem's graph of people data. It connects trillions of data points across individuals, companies, and time, draws on more than 100,000 data sources, and is the structure Findem's Labeling Engine uses to produce Success Signals and Relationship Signals.