Devii · Data & analytics · 2026-05-29 · 7 min read
dbt: Version-Controlled SQL Transformations In The Warehouse
Models, tests, and documentation as code for analytics teams using cloud warehouses.
**dbt** (data build tool) lets analytics engineers define **models** as `SELECT` statements materialized as views or tables in Snowflake, BigQuery, Redshift, Databricks, and other adapters. Docs: `docs.getdbt.com`.
dbt handles **dependencies** via `ref()` functions, generates a DAG, and runs tests (`unique`, `not_null`, relationships) in CI. Documentation site pages ship from YAML descriptions.
dbt Core is open source; dbt Cloud adds scheduling, IDE, and governance features. Separate dev schemas or warehouses per developer to avoid clobbering production tables.
dbt does not extract or load raw data; pair with ingestion tools (Fivetran, Airbyte, custom pipelines). Treat breaking model changes like API semver for downstream dashboards.