Knowledge Graph

A knowledge graph represents information as nodes (entities) and edges (relationships), with both treated as first-class. Its optimization is traversal: following connections several hops out — customer to account to transaction to counterparty — without knowing the path in advance, which is awkward and slow to express as SQL self-joins.

That makes it complementary to, not a replacement for, a data warehouse: the warehouse aggregates rows into measures, the graph walks relationships into context. Graphs earn their place where connection is the analysis — fraud rings, entity resolution, supply-chain dependency, lineage — and they’ve returned to prominence for AI grounding, because retrieval over explicit relationships holds up on multi-entity questions where pure vector search degrades. A knowledge graph is usually governed by an ontology that defines what its nodes and edges mean.

Go deeper: Knowledge Graph vs Data Warehouse: Aggregation vs Connection