Databricks Pushes Unity Catalog Governed Metrics Directly Into Power BI
Databricks launched Publish to Power BI for Unity Catalog metric views, currently in Private Preview. Metric views become Power BI tables, dimensions become columns, and measures auto-translate to DAX. Power BI queries the metric view directly, so joins and logic execute in Databricks.

Takeaways
- Databricks launched Publish to Power BI for Unity Catalog metric views, currently in Private Preview.
- Metric views become Power BI tables, dimensions become columns, and measures auto-translate to DAX.
- Power BI queries the metric view directly, so joins and logic execute in Databricks.
- DirectQuery preserves Unity Catalog row filters, column masks, and ABAC policies on every query.
- Power BI tasks in Lakeflow Jobs keep semantic models in sync as metric views change.
- Sync is one-way only; Power BI edits do not write back to Unity Catalog.
Databricks has introduced Publish to Power BI , a feature that converts Unity Catalog metric views into Power BI semantic models. A semantic model is Power BI’s dataset layer, where reports find tables, fields, relationships, and DAX measures.
Large analytics organizations often maintain conflicting versions of the same KPI across finance models, sales dashboards, marketing reports, and executive presentations. Publish to Power BI gives those teams a central place to define measures while preserving Power BI as the reporting interface.
A governed source behind familiar reports
Unity Catalog metric views define business dimensions and measures over data in the Databricks lakehouse. They belong to Unity Catalog’s semantic layer, which applies shared definitions, access controls, and lineage to metrics such as revenue, customer acquisition cost, and return on ad spend.
Publishing turns those definitions into Power BI objects without requiring report authors to recreate calculations or data relationships. The same metric views can also serve Databricks AI/BI dashboards and Genie, its natural-language analytics tool.
How metrics cross the boundary
| Unity Catalog object | Power BI result |
|---|---|
| Metric view | Table in the semantic model |
| Dimension | Column |
| Measure | DAX measure |
| Description and display format | Field metadata and formatting |
| Join defined in the metric view | Resolved by Databricks during queries |
The preview translates the calculations BI teams commonly use:
- Aggregations including
SUM,COUNT,AVG,MIN,MAX,MEDIAN,STDDEV, andVARIANCE. - Distinct counts, ratios, filtered measures, and measures composed from other measures.
- Current-period, offset, and trailing-window calculations.
- Currency, decimal, percentage, and other display formats.
- DirectQuery for live requests and Import mode for data loaded into Power BI.
In Databricks’ example, a marketing team defines a campaign_performance metric view with channel and region dimensions, plus measures for total spend, total revenue, and ROAS. The expression roas = MEASURE(total_revenue) / MEASURE(total_spend) publishes as DIVIDE([total_revenue], [total_spend]) , removing the need to rewrite that calculation by hand in DAX.
Policy checks stay in Databricks
Under DirectQuery, each report interaction sends a query to the metric view in Databricks. Unity Catalog then applies its access controls, row filters, column masks, and attribute-based policies before returning results. Joins also execute in Databricks, so Power BI authors do not need to duplicate the metric view’s relationship logic.
Import mode copies data into Power BI rather than querying Unity Catalog during each report interaction. Teams using that mode need Power BI-side controls for the imported data because report-time access no longer passes through Unity Catalog.
Publishing fits into data pipelines
Lakeflow Jobs can automate publication through a Power BI task . Teams can schedule the task or run it after upstream tables refresh, allowing new dimensions and updated measures to flow into the target semantic model.
Synchronization runs one way from Unity Catalog to Power BI. Changes made to published measures in Power BI remain local and do not update the original metric view. Teams therefore need clear ownership rules for generated measures and any Power BI-specific extensions.
Precomputation targets report latency
Metric-view materializations precompute eligible results while preserving the definitions in the metric view. When a DirectQuery request matches a materialization, Databricks can serve the stored result instead of recalculating it from the underlying tables. Databricks combines this mechanism with Lakehouse//RT to support interactive reporting over governed Delta data.
Actual latency and cost will depend on materialization coverage, SQL warehouse sizing, model shape, and concurrent report traffic. Preview evaluations should test representative dashboards and query patterns rather than relying on isolated measure benchmarks.
Databricks wants to own the definition
Dbt Semantic Layer, Cube, AtScale, and Looker’s LookML address the same organizational problem: deciding where shared business metrics should live. Databricks places those definitions beside governed lakehouse data and distributes them into the BI tools employees already use.
For organizations running both Databricks and Power BI, the practical effects are fewer duplicated DAX measures, less manual relationship modeling, and a defined path for propagating metric changes. The approach also lets teams retain existing Power BI reports and workflows while moving metric ownership into Unity Catalog.
Questions for the private preview
Private Preview access currently runs through Databricks account teams, and prospective users should validate several operational details against their existing Power BI estate:
- Coverage for custom DAX calculations and specialized metric-view expressions.
- Identity propagation and policy behavior under DirectQuery.
- Security requirements for datasets using Import mode.
- How republishing affects local Power BI measures and model customizations.
- Latency, concurrency, and warehouse cost under production workloads.
- Versioning, deployment, rollback, and ownership across development environments.
Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.
Read original report

