Docs navigation

Docs / Connectors

Connect Databricks

Converise queries Databricks through a SQL warehouse using an OAuth service principal with read-only access. The credential is encrypted at rest and never returned to the browser.

1. Create a service principal

  1. 1

    Add the principal and an OAuth secret

    In workspace Settings → Identity and access → Service principals, add a service principal and generate an OAuth secret. Note the client ID and the secret.

  2. 2

    Grant warehouse and data access

    Grant the principal Can use on a SQL warehouse and read-only (SELECT) access to the catalog/schema with your experiment data.

  3. 3

    Find the SQL warehouse ID

    The warehouse ID is in the warehouse's Connection details. A serverless or auto-resuming warehouse avoids cold-start timeouts on scheduled refreshes.

On Azure Databricks the secret must still be a Databricks OAuth secret generated in the workspace as above - a Microsoft Entra ID (app registration) client secret will not work here.

2. Connect in Converise

  1. 1

    Open the connect form

    Go to Connectors and pick Databricks from the catalog.

  2. 2

    Fill in the principal

    Enter the Workspace URL (e.g. https://adb-…azuredatabricks.net), Client ID, Client secret, and SQL warehouse ID.

  3. 3

    Set up the data source and KPIs

    On the connector page point the Data source at the table (or view) that holds outcomes as catalog.schema.table - or paste your own SQL, or fill it from a preset (GA4 export on BigQuery). Pick the column roles from the live schema: subject, timestamp, variant, optionally the experiment key, event name and value columns; switch on a separate assignments table when your platform stores exposures apart from outcomes. Then list the KPIs - an event name or a column that counts, reduced per subject as a conversion (count), a quantity (sum) or revenue (sum). Set once; every experiment on this connector binds in three clicks.

  4. 4

    Schedule per experiment, not here

    There is nothing to schedule on the connector. When you bind an experiment, choose whether it refreshes automatically once a day (off by default - each run bills warehouse queries) and at which UTC hour, ideally after your warehouse ETL completes.

  5. 5

    Check the connection and save

    Click Check connection to run a test query, then save. When editing later, leave the credential fields blank to keep the stored secret.

Next step

Bind an experiment to the connector on its Data sources tab - see Binding evidence.

PreviousBigQueryNextSnowflake