Learn from the best! Meet the four finalists headed to the FINALS of the Power BI Dataviz World Championships! Register now
can i create power bi report by using or combining two or more power bi semantic model as source?
Solved! Go to Solution.
You can use more than one Power BI semantic model in a single report, but there are two different scenarios:
1. Same report with multiple models,
You can connect to Model A and Model B and use visuals from both in one PBIX. However, the models stay separate: you generally can’t relate tables across them, and slicers/filters from one model usually won’t filter visuals from the other model.
2. Truly combining models,
You may be able to use a composite model (DirectQuery for semantic models) to add a second model and create relationships/combined measures. This depends on tenant settings and can introduce limitations such as performance, restricted modeling features, and more sensitivity to changes in the upstream models.
You will need permission to all underlying semantic models. If upstream models change (rename/remove fields), your report can break. Also, cross-model filtering and “single set of slicers” is not automatic unless you implement an integration approach.
If your end goal is a single, consistent reporting experience, the most reliable approach is usually to build a combined semantic model upstream and then build reports from that one model.
Power BI supports connecting to multiple semantic models using DirectQuery for Power BI datasets and combining them in one report.
Connect to Semantic Model A → Get data → Power BI semantic model
Then connect to Semantic Model B
Power BI creates a composite model
You can create relationships between tables from different models
This is possible. First, make sure the current PBIX file already contains data, then connect to the semantic models using DirectQuery.
To use measures from different semantic models together, there must be a shared (common) dimension between them. That common dimension should be used in the visual (e.g., in the table) so the measures can align correctly across models.
Also note that RLS and permissions are still enforced from the source semantic models. If a user doesn’t have access to a specific semantic model, the visuals that rely on it won’t return data (or may show an error), while the rest of the report can still load normally.
Hi @powerbiexpert22 ,
Sure we can .there are limitations with respect to RLS/OLS.
please go through link below.
https://learn.microsoft.com/en-us/power-bi/transform-model/desktop-composite-models
Please give kudos or mark it as solution once confirmed.
Thanks and Regards,
Praful
You can use more than one Power BI semantic model in a single report, but there are two different scenarios:
1. Same report with multiple models,
You can connect to Model A and Model B and use visuals from both in one PBIX. However, the models stay separate: you generally can’t relate tables across them, and slicers/filters from one model usually won’t filter visuals from the other model.
2. Truly combining models,
You may be able to use a composite model (DirectQuery for semantic models) to add a second model and create relationships/combined measures. This depends on tenant settings and can introduce limitations such as performance, restricted modeling features, and more sensitivity to changes in the upstream models.
You will need permission to all underlying semantic models. If upstream models change (rename/remove fields), your report can break. Also, cross-model filtering and “single set of slicers” is not automatic unless you implement an integration approach.
If your end goal is a single, consistent reporting experience, the most reliable approach is usually to build a combined semantic model upstream and then build reports from that one model.
Hi , @powerbiexpert22 , Yes you can bring in 2 or more semantic model as source and create a report on top of it .
When you connect to the first semantic model (A) it will be in Live mode . You can see it at the bottom of the desktop .
When you bring in the second model (B) by using the get data option pbi will provide a prompt asking it need to change the connection from live to Direct Query and asking for the user input to add the local model
once you select add a local model you can select the required tables you want to retain from model A
Choose the second model B and bring in the required tables
now you can create relationships between tables and use it as a single model .
But the keep in account of the constrains it brings RLS from both models and the users needs permission to access both the models to view the report .
When you publish it will create the semantic model A and B as the direct query source for your pbix file .
Thanks
If this response was helpful in any way, I’d gladly accept a kudo.
Please mark it as the correct solution. It helps other community members find their way faster
Hey @powerbiexpert22 ,
Yes, you can create a Power BI report that uses or combines two or more Power BI semantic models as a source, but not in the traditional way of simply merging them like tables in Power Query. Power BI does not allow you to use multiple live connections in a single report.
However, you can combine multiple semantic models by using a composite model with DirectQuery for Power BI datasets. In this approach, you connect to one semantic model first, then add another Power BI semantic model as an additional source. Power BI converts the report into a composite model, allowing you to create relationships between tables from different models and build measures across them.
That said, there are some limitations, such as performance dependencies on both models and restrictions on relationship types (typically many-to-one). For long-term scalability and cleaner architecture, many organizations prefer creating a new centralized semantic model that integrates all required data sources and then building reports on top of that single model.
In summary, while you cannot directly merge multiple published semantic models through multiple live connections, you can combine them using composite models, which is the most practical supported approach.
If this solved your issue, please mark it as the solution so others can find it easily.
If it helped, a quick Kudos is always appreciated it helps highlight useful answers for the community.
Thanks for being part of the discussion !!!
| User | Count |
|---|---|
| 57 | |
| 43 | |
| 32 | |
| 16 | |
| 13 |
| User | Count |
|---|---|
| 84 | |
| 70 | |
| 37 | |
| 27 | |
| 24 |