Forum Discussion

jnjAnil's avatar
jnjAnil
New Member
9 months ago
Solved

Can we break the Power BI semantic model into different subsets

Can we split the Power BI semantic model as a main superset and smaller subsets that contain limited tables. The intent is to create a simplified semantic model for self‑service reporting, where only...
Need Help
  • anilgavhane's avatar
    anilgavhane
    9 months ago

    Steps to Create a Child Semantic Model from a Superset in Power BI

    1. Publish the Superset Model
      • Build and publish the main semantic model (superset) with all tables, relationships, and measures.
      • This acts as the master dataset.
    2. Create a New (Blank) Power BI File
      • Open Power BI Desktop and start with a blank report.
      • Add a dummy/blank table if needed to initialize the model.
    3. Connect to the Superset Semantic Model
      • Use Get Data → Power BI Datasets.
      • Select the published superset dataset.
      • This establishes a connection to the main model.
    4. Select Limited Tables (Subset)
      • Once connected, you’ll get the option to choose which tables to bring into the child model.
      • Import only the required tables/columns for the self‑service scenario.

    Note: AI helped me to format this result:

Mauro89's avatar
Mauro89
Icon for Super User rankSuper User
9 months ago

Hi jnjAnil 

 

Yes, you can achieve this using composite models and perspectives (in limited scenarios), but the approach depends on your specific setup:

 

Option 1: Composite Models (Recommended)

  • Create a main semantic model with all tables
  • Build a secondary model that connects live to the main model via "Connect to a Power BI dataset"
  • Add only the simplified tables/calculations needed for self-service users
  • The secondary model stays linked to the main model for data consistency
  • Users connect to the simplified model and only see what you've exposed

Option 2: Direct Lake with Shortcuts (Fabric)

  • Use Microsoft Fabric with Direct Lake mode
  • Create multiple semantic models pointing to the same OneLake data via shortcuts
  • Each model exposes only relevant tables for different user groups
  • All models share the same underlying data for consistency

Option 3: Perspectives (Limited)

  • In Analysis Services/Fabric, you can use perspectives to create filtered views of the same model
  • However, this requires Premium capacity and XMLA endpoint access
  • Perspectives are metadata-only - not true separate models

I hope any of these options might be a fit for you situation.

 

Best regards!

PS: If you find this post helpful consider leaving kudos or mark it as solution