Dry-running a model to check its logic and schema before a real build

The --empty flag lets you dry-run dbt build or dbt run by replacing all ref() and source() calls with a limit-0 query, so models execute against your warehouse and validate SQL syntax, dependency logic, and schema definitions without processing any actual row data. This makes it a fast, low-cost way to confirm that models compile and run correctly before wasting time or warehouse credits on full builds.

1 · Learn the must-know

• 1Running dbt run --empty or dbt build --empty replaces referenced tables/views with an empty (limit 0) subquery, so the query plan is validated but no real data is scanned.
• 2This is the fastest way to catch SQL syntax errors, broken refs/sources, and column-level schema mismatches before running a full, expensive build.
• 3Because zero rows are returned, --empty is not suitable for validating actual data quality, business logic correctness on real values, or data-dependent tests.
• 4Materializations that depend on data volume or actual row content (like incremental logic branching or certain snapshot behaviors) may not fully validate with --empty since underlying tables might be empty or the run may take specific empty-input code paths.
• 5The flag works with dbt run and dbt build, and combining it with dbt test is not meaningful since tests generally require real data to assert against.
• 6Using --empty is a recommended CI/CD or pre-merge practice to quickly verify model compilation and schema before running full or slice-based production-scale builds.

2 · Check your understanding