Create User-Defined Aggregations for Report Building using a centralized Semantic Model

[dgefcoe]

I have a centralized Semantic Model that is used for many reports. That model is mixed, with Fact tables as Direct Query and Dimensions in Dual DQ/Import mode. This model is used as a self-service data portal for many users.

In each of my reports, I connect to the Semantic Model (via DQ) as my primary data source. I would like to build User-Defined Aggregations (UDA) relevant to each report in that report's model (building a composite model for each report).

I cannot get this to work. Load times remain long, as if there is no recognition of the aggregations existing in the composite model. 

I know I am setting up the aggregations correctly because If I build the aggregations in the original Semantic model and connect to that via DQ as usual, the report will read from the aggregations and render quickly.

This is not ideal because I need to build out multiple kinds of aggregation, most of which are not applicable to any individual report and it balloons the size of the base semantic model considerably.

I would prefer to build out the aggregations in each report with only the relevant data. Is this possible?

[v-tsaipranay]

Hi @dgefcoe ,

Could you please confirm if you've submitted this as an idea in the Ideas Forum? If so, sharing the link here would be helpful for other community members who may have similar feedback.  Please feel free to contact us if you have any further questions.

Thank you.

[v-tsaipranay]

Hi @dgefcoe ,

Thanks for reaching out to the Microsoft fabric community forum.

Currently, Power BI does not allow user-defined (per-report) aggregations in composite models that use DirectQuery to connect to a shared Semantic Model. Aggregation awareness works only within the same semantic model, so local Import tables in a report are not seen as valid aggregations for the main dataset. This means queries still go to the DirectQuery source, leading to slower performance.

To achieve your goal, you can define aggregations directly in the central Semantic Model for shared use, or create specific semantic models for each reporting need with the necessary aggregation tables. Alternatively, you could pre-aggregate data in Fabric Warehouse or Dataflow Gen2 before importing it into your report model.

we suggest sharing this scenario and feedback through the Fabric Ideas - Microsoft Fabric Community so it can be reviewed for future consideration by Microsoft’s product team.

Hope this helps. Please feel free to rech out for any further questions.

Thank you .