Building a Real dbt Project

Where the work actually starts: defining sources, writing stg_ models, the conventions every team follows. FIND_VIDEO: search 'dbt staging layer sources tutorial' — recommended channel: dbt Labs / Kahan Data Solutions. Aim for 10 min or under.

Practice quiz: The staging layer is unglamorous. Get it right and everything downstream is easier; skip it and you'll pay forever.

The four built-in materializations and the cost model behind picking each. FIND_VIDEO: search 'dbt materializations table view incremental' — recommended channel: dbt Labs. Aim for 11 min or under.

Practice quiz: Each materialization trades build cost for query cost differently. Here's how to pick.

The most-used and most-misused materialization. Patterns that work, gotchas that bite. FIND_VIDEO: search 'dbt incremental model unique_key' — recommended channel: dbt Labs. Aim for 11 min or under.

Practice quiz: Three incremental strategies and the late-arriving-data problem that breaks all of them.

Generic tests in YAML, custom SQL tests, what to test for vs what to skip. FIND_VIDEO: search 'dbt tests unique not_null relationships' — recommended channel: dbt Labs. Aim for 10 min or under.

Practice quiz: Every model needs tests. Four generic tests cover most cases; custom tests handle the rest.

Auto-generated docs from your YAML. The highest-leverage 30 seconds you'll spend. FIND_VIDEO: search 'dbt documentation descriptions docs generate' — recommended channel: dbt Labs. Aim for 9 min or under.

Practice quiz: Documentation people actually read: focused descriptions, column meaning, lineage you can trust.

Construct production dbt staging models for a Fivetran Shopify connector: source definitions with freshness tests, deduplication via ROW_NUMBER(), and generic column assertions.

Convert a slow 45-minute full table scan into an incremental merge model using is_incremental(), unique_key, and lookback windows for late-arriving records.