Power BI dataflows: prepare data once and reuse it
Power BI dataflows let a team clean and shape data once in the cloud, then reuse the result in many reports. They run Power Query in the service instead of inside each Desktop file. So a customer list or a calendar has one definition that everyone shares. This guide explains how dataflows work, when they are worth the effort and how the newer Gen2 version in Microsoft Fabric changes the picture.
- Power Query in the cloud
- One definition, many reports
- Gen2 in Fabric
Last updated
What are Power BI dataflows?
A dataflow is a set of Power Query steps saved in a workspace. It connects to sources, transforms the data and stores the output as tables. The editor looks and feels like Power Query in Desktop.
The difference is where the work runs. A dataflow refreshes in the Power BI service on its own schedule. After that, semantic models connect to its tables through the dataflows connector, as they would to a database.
Therefore the heavy shaping happens once. Ten reports that need the same product table all read one prepared copy. Also, the source system receives one set of queries instead of ten. That matters when the source is slow or sensitive.
Where shared data prep pays off
These six situations are where we most often suggest a dataflow.
Shared dimensions
Customers, products and dates get one definition, so every report groups data the same way.
Slow sources
An API or an old system is queried once, while reports read the stored result.
Analyst-owned prep
People who know Power Query can build pipelines without waiting for a database team.
Smaller Desktop files
Transformation logic leaves the report file, which then opens and refreshes faster.
Separated schedules
Data prep can refresh at night, because models pick it up on their own timing.
Gradual growth
A dataflow is a modest first step before a team commits to a full warehouse.
How does Dataflow Gen2 differ from the original?
Microsoft Fabric introduced Dataflow Gen2. It keeps the Power Query editor, but adds data destinations. So the output can land in a lakehouse, a warehouse or another store, not only in the dataflow's own storage.
That changes what a dataflow is for. The original kind feeds Power BI models. Gen2 is closer to a general loading tool, since other engines can read what it writes. However, Gen2 needs a Fabric capacity, while the original works in ordinary workspaces.
Which should you pick? If the team only needs reusable tables for reports, the original still does the job. If you are building a lakehouse or want to combine with pipelines, Gen2 fits better. We look at licenses and plans first, then advise. If you are deciding what belongs in a dataflow and what belongs in the model, see Power BI dataflow vs semantic model, and for a full warehouse design, see Power BI data warehouse consulting.
Pitfalls to avoid with Power BI dataflows
Six pitfalls account for most of the trouble we get asked to fix.
- One giant dataflow that does everything
- Refresh times that collide with model refresh
- Steps that stop query folding early
- No owner besides the person who built it
- Business logic copied into each report anyway
- No notice when a refresh fails
How to start with Power BI dataflows
Four steps get a first dataflow into steady use.
Pick a shared table
First choose a table that several reports already prepare separately, such as customers.
Build and test
Then move its Power Query steps into a dataflow and compare the output with the old result.
Schedule in order
Next set the dataflow to refresh before the models that depend on it.
Repoint the reports
Finally switch each report to the dataflow table and remove the duplicate steps.
Questions teams ask about shared data prep
Is a Power BI dataflow the same as a semantic model?
No. A dataflow prepares and stores tables. A semantic model adds relationships, measures and security on top, and reports read the model. So the two work together. The dataflow does the cleaning, and the model does the business logic. Putting measures into a dataflow is not possible, since it has no calculation layer.
Do dataflows in Power BI need a Premium or Fabric capacity?
Not for the basics. Standard dataflows run in a workspace with Pro licenses. Some features need a capacity or Premium Per User, for example linked and computed tables and the enhanced compute engine. Dataflow Gen2 needs a Fabric capacity. We confirm against Microsoft's current licensing pages, because the details shift.
Should we use a dataflow or a data warehouse?
It depends on scale and skills. A dataflow suits a team of analysts with moderate volumes and a few sources. A warehouse suits large volumes, many sources and engineers who write SQL. Teams often start with dataflows and move later. The Power Query logic they wrote still documents the business rules, so little is wasted.
Can a Power BI dataflow use incremental refresh?
Yes, with the right license. Incremental refresh on a dataflow reloads only recent periods, which shortens long runs. It needs a date column and a source that filters efficiently. Without query folding, the dataflow still reads everything before it filters. So we check folding before turning the feature on.
Can Power BI Agency build our dataflows?
Yes, or review the ones you have. Our consultants do the work inside your tenant and explain each step, so your analysts can maintain it later. We are a division of Sataware Technologies and independent of Microsoft. A written, fixed scope comes first, after a free assessment call.
Let's build your first dashboard together.
Tell us what you report on today. A consultant replies within 2 hours with next steps.
Trusted partners
Microsoft Solutions Partner · AWS Partner
Cloud & AI Platforms
Azure SQL, Microsoft Fabric and the data pipelines behind your reports.
AI Business Solutions
Power BI reporting on your Dynamics 365, Business Central and Power Platform data.
Security
Row-level security and Entra ID sign-in on every report and workspace.
Amazon Web Services
Power BI on your AWS data, with connectors for Redshift, Athena and S3.