Detailed definition
A BI knowledge graph connects business entities, measures, dimensions, hierarchies, synonyms, descriptions, and access rules. It describes not just where data lives, but how the business understands the data.
For AI-powered analytics, a knowledge graph gives the AI a structured map to reason over. Instead of guessing which table or column might answer a question, the system can match user language to governed concepts and relationships.
Why it matters
Most business questions rely on context. "Top customers", "active products", "net revenue", and "last quarter" may all depend on company-specific rules. A knowledge graph makes that context explicit and reusable.
It also creates a foundation for trustworthy self-service. Business users can ask questions in familiar language, while data teams still control definitions, joins, calculations, and permissions.
How it works
Data teams model the important concepts in the business and connect them to the underlying warehouse. The graph can include synonyms, descriptions, custom calculations, hierarchies, entity relationships, and access controls.
When a user asks a question, the analytics system resolves terms against the graph, builds a semantic query, and executes it against the connected data source using the approved mappings.
Practical examples
- "Account" connects to contracts, invoices, usage events, and customer success ownership.
- "Gross margin" connects to revenue, cost, exclusions, and reporting-period logic.
- "Region" connects to countries, sales territories, and row-level permissions.
- Synonyms such as "ARR", "annual recurring revenue", and "subscription revenue" can point to the correct concept when appropriate.
Common pitfalls
- A knowledge graph is more than documentation. The analytics system has to read it at query time to translate questions into the right SQL, not just show definitions to humans.
- It should not contain raw transactional data by default. In BI, it usually stores semantic metadata and mappings.
- A graph that is not maintained will drift. Ownership, tests, and version control help keep it reliable.