Forum Discussion
Using Import Models and Dataflows Together
Hi everyone,
I have a Power BI workspace where most of my reports use semantic models built in Import mode.
For one specific report, the dataset is much larger, and since I have a Power BI Pro license, I am planning to use a Dataflow to perform the data transformation in the Power BI Service instead of doing all the transformations in Power Query Desktop. The semantic model will then load data from the Dataflow.
My question is:
- Is it considered a good practice to have a mix of regular Import mode semantic models and Dataflow-based semantic models within the same workspace?
- Are there any limitations, performance considerations, or management issues I should be aware of when using this approach?
I'd appreciate any guidance or best practices from the community. Thanks!
Yes, this is a perfectly valid and commonly used architecture. There's no issue with having a mix of regular Import mode semantic models and Dataflow-based semantic models within the same workspace.
Using a Dataflow for larger or more complex ETL workloads can actually be considered a best practice because it separates data preparation from semantic modeling and allows the transformed data to be reused across multiple reports.
A few considerations to keep in mind:
Reuse: If multiple semantic models require the same transformed data, using a Dataflow avoids duplicating Power Query logic across PBIX files.
Refresh order: Ensure the Dataflow refresh completes before any dependent semantic models refresh. You can achieve this by configuring an appropriate refresh schedule.
Performance: Dataflows can reduce the amount of transformation work performed during semantic model refreshes, although the semantic model will still need to import the processed data.
Capacity and licensing: Since you're using Power BI Pro, be aware of refresh limits and shared capacity constraints. Large datasets may still require careful optimization of your model and queries.
Management: Keep naming conventions and documentation consistent so it's clear which semantic models depend on Dataflows.
If the large dataset is only used by a single report, you could continue using Power Query in Desktop. However, if the transformations are complex or the data may be reused in the future, moving the ETL logic into a Dataflow is a good design choice.
For more information, see:
Introduction to Dataflows: https://learn.microsoft.com/power-bi/transform-model/dataflows/dataflows-introduction-self-service
Best practices for Dataflows: https://learn.microsoft.com/power-query/dataflows/best-practices-developing-complex-dataflows
Semantic models in Power BI: https://learn.microsoft.com/power-bi/connect-data/service-datasets-understand
💡 Helpful? Give a Kudos 👍 — keep the community growing.
✅ Solved your issue? Mark this as the Accepted Solution ✔️
Best regards, Prince Singh | Data Science & Microsoft Fabric Enthusiast
11 Replies
Hii Learner_PowerBI
A semantic model connected to a dataflow is still an Import model; the dataflow simply provides a centralized transformation and staging layer, so both types can coexist in the same workspace. Ensure the dataflow refresh completes before the semantic model refresh. Also note that a Pro/shared-capacity semantic model is still limited to 1 GB compressed, eight refreshes per day, and a two-hour refresh duration, so a dataflow alone will not remove large-model limitations.
The link below clearly explains this:
https://learn.microsoft.com/en-us/power-bi/transform-model/dataflows/dataflows-configure-consume
Thank you rohit1991 for the time and support.
Yes, this is a perfectly valid and commonly used architecture. There's no issue with having a mix of regular Import mode semantic models and Dataflow-based semantic models within the same workspace.
Using a Dataflow for larger or more complex ETL workloads can actually be considered a best practice because it separates data preparation from semantic modeling and allows the transformed data to be reused across multiple reports.
A few considerations to keep in mind:
Reuse: If multiple semantic models require the same transformed data, using a Dataflow avoids duplicating Power Query logic across PBIX files.
Refresh order: Ensure the Dataflow refresh completes before any dependent semantic models refresh. You can achieve this by configuring an appropriate refresh schedule.
Performance: Dataflows can reduce the amount of transformation work performed during semantic model refreshes, although the semantic model will still need to import the processed data.
Capacity and licensing: Since you're using Power BI Pro, be aware of refresh limits and shared capacity constraints. Large datasets may still require careful optimization of your model and queries.
Management: Keep naming conventions and documentation consistent so it's clear which semantic models depend on Dataflows.
If the large dataset is only used by a single report, you could continue using Power Query in Desktop. However, if the transformations are complex or the data may be reused in the future, moving the ETL logic into a Dataflow is a good design choice.
For more information, see:
Introduction to Dataflows: https://learn.microsoft.com/power-bi/transform-model/dataflows/dataflows-introduction-self-service
Best practices for Dataflows: https://learn.microsoft.com/power-query/dataflows/best-practices-developing-complex-dataflows
Semantic models in Power BI: https://learn.microsoft.com/power-bi/connect-data/service-datasets-understand
💡 Helpful? Give a Kudos 👍 — keep the community growing.
✅ Solved your issue? Mark this as the Accepted Solution ✔️
Best regards, Prince Singh | Data Science & Microsoft Fabric Enthusiast
Thanks Prince0011 for the great explanation. It really helped and solved my doubt.
What I would recommend when working with your semantic models is always to make your semantic model size as small as possible. So remove any extra columns which are never gonna be used for any measures or any reporting to reduce the size of your semantic model which will help you keep under the pro size limit of 1GB.
Thanks GilbertQ . Sure will keep it mind the best practice in order to limit semantic model size.
Yes, this is a common practice that there is no conflict between Import models and Dataflow sourced models in the same workspace. Few things to keep in mind with your Pro license setup is that Dataflow will be in Import mode once taken by the semantic model as Pro will not support DirectQuery over Dataflows across workspace the way Premium does.
So you can gain reusability and offloading transformation logic to the service instead a different storage mode. Watch your refresh scheduling the Dataflow needs to finish refreshing before dependent semantic model refreshes. Also keep an eye on Dataflow storage limits and refresh duration on Pro as large datasets might push against timeout limits
Thank you krishnakanth240 for the explanation. It really cleared the doubts.
Yes, mixing regular Import models and Dataflow-backed models in the same workspace is 100% fine and very common. Power BI treats a Dataflow just like any other database source.
3 Things to Watch Out For (On a Pro License)
Refresh Staggering (No Auto-Trigger)
On Pro, a Dataflow refresh won't automatically trigger your semantic model refresh.
You must schedule them with a time buffer (e.g., Dataflow at 2:00 AM, Dataset at 2:30 AM) or use Power Automate to orchestrate them sequentially.
Pro Limits Still Apply
8 refreshes/day per artifact (Dataflow and Model count separately).
10 GB total limit across workspace items.
No Enhanced Compute Engine (so heavy joins inside the Dataflow won't be hardware-accelerated).
Workspace Management
Permissions apply to the entire workspace. If developers or builders only need access to the report, they'll also see the raw Dataflow.
Best Practice: As your setup grows, put Dataflows in an "ETL Workspace" and semantic models in a "Reporting Workspace."
Thanks MohitsinghMS for your support and time.
Hi Learner_PowerBI,
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.
Thank you.
