This path is for you if dbt is a folder of SQL that nobody wants to review. Three parts: a project layout that can grow, first models and tests, and reviewing a model the way you would review a change. Start with the first part on this page. The conceptual what is dbt lives in the pipelines series, not here.
- 1 dbt lab: project layout that scales A messy dbt repo ages fast. Map staging, intermediate, and marts with clear naming and one grain per model so the project scales past your first three tables without tribal knowledge.
- 2 dbt lab: first models and tests Turn layout into working models. Write your first staging and mart SQL, wire refs, add unique and not-null tests, and run a small build you can actually explain in a review.
- 3 dbt lab: reviewing a model like a PR Review dbt model PRs for grain, joins, filters, tests, docs, and blast radius. Green CI means the files compile. It does not mean the metric is safe for the board.