Forum Discussion
multiple or single semantic model
in which scenario there is a need to define multiple semantic models? let say there are different departments in an organization such as HR, Finance etc , should i create semantic model for each department (make sure to avoid replication of semantic model) or create a single semantic model for all departments and apply RLS for each user based on the department they belong?
Hi powerbiexpert22,
Use Single Model + RLS when departments share the same fact tables and need a unified source of truth. RLS handles row-level restriction per user/department. Simpler to maintain, consistent KPIs.
Use Multiple Model when structures diverge, split when :
- Different grain (HR = employee level, Finance = GL line level)
- Sensitive data isolation required (payroll ≠ visible even as metadata to Finance)
- Different refresh cadences or development team ownership
- Model becomes a bloat of irrelevant measures per audience
The real-world best practice to avoid duplication is using Hybrid Model which means having one certified core model for shared dimensions (Date, CostCenter, OrgUnit) → department models connect via composite/live connection and add their own facts and measures. No dimension duplication, clean ownership per team.
How to choose the right pattern ? start with one model + RLS, split only when you hit a structural or compliance constraint.
5 Replies
Hey powerbiexpert22 ,
most likely there will be varying requirments regarding content of the semantic model, meaning the dimensional model for this reason, my recommendation: create separate semantic models for each department.
Regards,
TomHi powerbiexpert22,
Use Single Model + RLS when departments share the same fact tables and need a unified source of truth. RLS handles row-level restriction per user/department. Simpler to maintain, consistent KPIs.
Use Multiple Model when structures diverge, split when :
- Different grain (HR = employee level, Finance = GL line level)
- Sensitive data isolation required (payroll ≠ visible even as metadata to Finance)
- Different refresh cadences or development team ownership
- Model becomes a bloat of irrelevant measures per audience
The real-world best practice to avoid duplication is using Hybrid Model which means having one certified core model for shared dimensions (Date, CostCenter, OrgUnit) → department models connect via composite/live connection and add their own facts and measures. No dimension duplication, clean ownership per team.
How to choose the right pattern ? start with one model + RLS, split only when you hit a structural or compliance constraint.
Hi powerbiexpert22 ,
I would like to take a moment to thank oussamahaimoud for actively participating and sharing valuable insights on this thread, your contributions are truly appreciated.
Could you please confirm if you had a chance to review the information shared.
Let us know if it helped or if you need any further assistance.
We are happy to help.Hi powerbiexpert22 ,
I wanted to check if you had the opportunity to review the information provided. Please feel free to contact us if you have any further questions.
Hi powerbiexpert22,
If the requirements for all departments are the same, you can proceed with the RLS. Less effort in maintenance.
However, if each department has its own requirements, then you need to proceed with multiple models. It requires more effort in maintenance but you'll be sure that there will be no any data mix up.
Hope this helps.
