Refactor dbt models without changing consumer output
When one model is hundreds of lines long and nobody dares touch it.
- Transformation
Re-engineer something you own
10 RecipesRebuild something already in production and prove the new form behaves like the old one.
Convert a full-refresh dbt model to incremental and show the runtime delta
A dbt model is correct but a full refresh is too slow or too expensive to run routinely.
- Transformation
Backfill a bounded date range after a logic change without a full refresh
When a definition changed and months of history need recomputing, but a full refresh would take all weekend.
- Transformation
Convert a legacy SQL script or view into a dbt model and prove parity
When a scheduled SQL script is the real production logic and nobody can prove a rebuild would match it.
- Transformation
Convert a stored procedure into a dbt model and prove parity
When the nightly load is a stored procedure nobody fully understands anymore.
- Transformation
Replace a Fivetran or Airbyte connector with a dlt pipeline and reconcile the result
When the connector bill changed shape, or the connector has been broken with a pending fix for weeks.
- Ingestion
Convert a materialized view into a dbt model and preserve its required refresh semantics
When a materialized view refreshes on a schedule nobody wrote down and downstream reports assume that cadence.
- Transformation
Convert a notebook transformation into a dbt model and prove parity
When a transformation only exists as a notebook someone runs by hand before the numbers go out.
- Transformation
Convert a Spark ingestion notebook into a dlt pipeline and prove landed parity
When a Spark notebook is the only thing that lands a source into bronze and nobody can run it unattended.
- Ingestion
Convert a scalar SQL function into a dbt macro and prove equivalence
When a scalar SQL function is called from a dozen models and lives outside version control.
- Transformation