Semantic & Metrics Layers — Week 3 CheckWeek 3 of dbt & Analytics Engineering: metric definitions as infrastructure — the Airbnb/Minerva model applied to the transformation layer.

What problem does a semantic/metrics layer solve that well-modeled tables alone cannot?

Airbnb routes all metric definitions through Minerva so dashboards, experimentation, and discovery consume the SAME definition. The organizational tradeoff:

A metric definition changes (e.g., refunds now deducted at initiation). What must the change process include?

The metric layer sits ABOVE models. What belongs in the model layer vs the metric layer?

Self-serve analytics at Datadog scale (5,000 employees) works only when paired with:

An experiment readout and the finance dashboard disagree on the same conversion metric. With a working semantic layer, the FIRST question is: