Build
Author Models And Tests
Turn sources into trusted transformations with reusable SQL and explicit quality gates.
Use this guide when you want to move from raw or staging data toward durable analytical outputs.
Inputs
- a trusted source table or view
- a project naming convention
- SQL logic for the intended transformation
- test expectations for the output
Flow
- create a model with clear materialization intent
- keep business logic in the model, not in downstream ad hoc queries
- add tests that prove uniqueness, not-null behavior, or relationship assumptions
- run the model and review the DAG and test results
- promote only after the output is stable enough for reuse
Visual Model
Verification
- the model runs successfully
- tests fail for real data quality issues, not for ambiguity in the contract
- the output grain and purpose are documented for downstream users