Semantic Layer

Concept

A semantic layer is a version-controlled definition of your business metrics and entities — what "revenue", "active customer" or "new order" mean in SQL — that every query and every tool resolves against.

Layman Explanation & Analogy

It is the company dictionary for numbers. Instead of every analyst, dashboard and AI tool inventing its own definition of "active customer", there is one written-down definition, stored with the code, that everything looks up.

Worked real-world example

A semantic layer defines `net_revenue` once as gross order value minus discounts, refunds and tax, filtered to `status = 'completed'`. Every dashboard tile, scheduled report and AI-generated query that asks for revenue gets exactly that expression.

What people get wrong & common traps

Putting business definitions in a prompt instead of a semantic layer makes them a suggestion rather than a rule — a model can follow them, paraphrase them, or quietly drop them under a long context. Definitions that matter financially belong somewhere deterministic, where a code review can catch a change to them.

Related terms