Home › Key concepts › Semantic layer and metrics
Key concept · Run (advanced) · the same on every platform
Semantic layer and metrics
Define each business number once, so dashboards and AI give the same answer.
What it is
A semantic layer stores metric definitions, such as 'net revenue = gross minus refunds, excluding test orders', in one place. Dashboards, spreadsheets and AI assistants ask the layer for the metric instead of writing their own SQL.
Why it matters
It ends the 'why do two dashboards disagree' meeting, and it is the safe way to let an AI answer questions about data.
An example
Finance and sales both ask for 'net revenue last month' and get the same number from the same definition.
The common mistake
Letting a model write raw SQL against tables, so the same question gets different answers.
Where you meet it on each platform
The idea stays the same; only the product name changes.
| Job | AWS | Azure | Google Cloud | Databricks | Open source |
|---|---|---|---|---|---|
| Semantic layerone agreed meaning for each metric | dbt semantic layer | Power BI semantic models | Looker (LookML) | Metric views | Cube Core |
Related ideas
Where it shows up
Part of the Run level on these paths: Data Engineering (shaping data).