Semantic Layer Overview

The Semantic Layer is the set of Data Management tabs that translate raw columns of your registered tables into the metrics, dimensions, and assignment sources your experiments read.

Where it lives in the console

Open Data Management. The body tabs, left to right, are:

• Metrics Catalog — every metric known to the project.
• Metric Group — hierarchical grouping of metrics by team or topic.
• Dimension — the cuts you can slice experiment results by in Explore.
• Feature List and Feature Combination — reusable feature sets (see Feature List).
• Tables — the registered tables that everything above is built on (see Define Your Data Model).
• Warehouse Native — the warehouse connection itself (see Connecting Your Warehouse).

The first three tabs — Metrics Catalog, Metric Group, and Dimension — are the Semantic Layer this section documents.

How the pieces fit together

Warehouse table  →  Registered table  →  Metric / Dimension  →  Used in an experiment

• A Fact Table is the source a metric aggregates over.
• A User Property Table is the source most dimensions read from.
• An Assignment Table is what an experiment reads exposures from.

Together these objects turn a question like "did checkout conversion improve for users in country = US" into SQL the warehouse runs in place.

What you can do here

• Define a metric and pick its field aggregation (Average, Sum, Ratio, …).
• Organize metrics into a Metric Group hierarchy.
• Register a dimension so results can be split by a user attribute or an event field.
• Point an experiment at the right Assignment Table.