Why do tools report different numbers for the same metric?

Teams may calculate “revenue” with different refund policies, time zones, or filters. Independent SQL in each dashboard makes those differences hard to see. A semantic layer centralizes business definitions so reports, applications, and agents can reuse the same meaning.

How does it turn business meaning into a query?

The layer maps metrics, dimensions, entities, and relationships to source tables and fields. When a user asks for revenue by region, it resolves the calculation, join path, grain, time rule, and access policy before generating a query. Clear grain and allowed dimensions help prevent duplicate counting. Defining business semantics in the data engine shows this mapping in practice.

How does it differ from a metrics or context layer?

A metrics layer focuses on metric definitions. A semantic layer also models entities, dimensions, and relationships. An AI context layer organizes the knowledge and evidence an agent needs for a task. A semantic layer can supply governed data to it, while source quality and query execution still need separate controls.

FAQ

Does it replace a data warehouse?

No. It supplies meaning and query rules over stored data.

Can it prevent every metric discrepancy?

No. Source quality, mapping, and versioning still need control.

Who should own definitions?

Business owners and data teams should jointly approve and maintain them.