A perspective from Plainsight Migrating to Fabric or Databricks? Generative BI
Platforms

dbt on Databricks.

dbt does not move or own your data. It generates SQL that runs on Databricks SQL, so the same project, tests and docs build straight into the lakehouse you already pay for, governed by Unity Catalog.

Databricks logo
What running dbt on Databricks looks like

The same project, compiled to Databricks SQL.

The dbt-databricks adapter targets Databricks SQL. Your models compile to the SQL Databricks runs, so nothing about the way your team builds data changes with the platform. Incremental models use MERGE, slowly changing dimensions use dbt snapshots, and Unity Catalog governs permissions and lineage across the lakehouse. The modelling, tests, sources and documentation are identical to a dbt project on any other platform.

How it fits together

One project, a Databricks target.

Author once. dbt generates the SQL Databricks expects, and the platform-native patterns sit at the edges.

What you get

Databricks specifics, handled.

What stays the same is most of it. What maps to Databricks is the target and a few native SQL patterns.

Databricks SQL target

The dbt-databricks adapter compiles your models to native Databricks SQL and builds straight into the lakehouse.

MERGE and snapshots

Incremental models use MERGE so runs stay fast and cheap, and dbt snapshots track slowly changing dimensions over time.

Unity Catalog governance

Unity Catalog handles permissions and lineage at the platform level, with dbt tests, docs and lineage layered on top.

Questions

dbt on Databricks, in short.

How does dbt run on Databricks?

The dbt-databricks adapter targets Databricks SQL. dbt compiles your models into SQL and Databricks executes it against your tables, so the data never leaves the estate you govern. The project structure, tests, sources and documentation are the same ones you would use on any platform.

How do incremental models and history work on Databricks?

Incremental models compile to MERGE, so each run updates matched rows and inserts new ones instead of rebuilding the whole table. Slowly changing dimensions use dbt snapshots, which track history over time. Both compile to native Databricks SQL, so runs stay fast and cheap.

Where does governance fit?

Unity Catalog handles governance, permissions and lineage at the platform level, and dbt builds into the catalogs and schemas it manages. dbt adds its own tests, documentation and ref-based lineage on top, so you get both platform governance and project-level traceability.

Does targeting Databricks add a licence cost?

No. dbt Core is open-source and free to run, and so is the dbt-databricks adapter. You pay for the Databricks compute you already run; dbt adds no licence on top of it.

However you build on Databricks

Let's put dbt on your lakehouse.

Tell us how you build on Databricks today, and we will show you what a dbt project looks like on it.

Talk to us
Talk to us