Define metrics once and keep them correct with the dbt Semantic Layer on Amazon Redshift

This article demonstrates how to use the dbt Semantic Layer with MetricFlow on Amazon Redshift to define non-additive metrics like average order value once and compute them correctly across arbitrary dimension combinations at query time.

• Pre-aggregated gold tables fail for non-additive metrics like average order value when rolled up to different time grains
• MetricFlow generates correct SQL at query time by recomputing metrics at requested grain instead of averaging pre-aggregates
• Define metrics once in version-controlled YAML semantic models alongside dbt models
• Query metrics across any dimension combination without building combinatorial pre-aggregated tables
• Walkthrough includes seeding sample data, building gold models, defining semantic models and metrics, and querying at different grains
• Generated SQL shows MetricFlow computing SUM(amount) / COUNT(orders) at requested grain for accuracy

The dbt Semantic Layer on Amazon Redshift eliminates the choice between table sprawl and incorrect metrics by defining governed metrics once and computing them correctly for any dimension and time grain combination.