Semantic Modelling

From star schema to a governed metric

Your dimensional model is the physical truth; the metric layer is the shared meaning on top of it. Here is how to get cleanly from one to the other.

A star schema tells you where revenue lives; a governed metric tells everyone what revenue means. The semantic layer maps a business term to a measure, a grain and the dimensions it can be sliced by — once, for every consumer.

Take a fct_sales fact with a net_amount measure. In a metrics layer you declare weekly_revenue as the sum of net_amount at a weekly grain, allowed to be grouped by product category and region. Every dashboard, notebook and AI agent that asks for weekly revenue now gets the same number.

That single definition is what lets you ground an LLM safely: it answers questions against governed metrics instead of guessing SQL.

Read the full Semantic Modelling track →