How incremental refresh works in Power BI, and when to use it

Incremental refresh loads only new and recently changed data instead of reloading everything. Here is how it works and how to set it up well.

By the Seestem data team ·

What incremental refresh does

By default, a Power BI dataset reloads all of its data on every refresh. That is fine for small tables, but as history accumulates, refreshes get slower and the source database does more and more work.

Incremental refresh changes this. Power BI splits a table into time-based partitions, keeps the older ones as they are, and refreshes only the recent ones. The first refresh in the service loads the full history; after that, each refresh handles only the recent window.

When it is worth using

Incremental refresh is most valuable when a table is large, grows over time and has a reliable date or datetime column. It is a good fit for:

  • Fact tables such as transactions, orders, events or logs.
  • Sources where reloading all history puts noticeable strain on the database.
  • Datasets whose refreshes are approaching their time limits.

How it is set up

You create two datetime parameters named RangeStart and RangeEnd in Power Query, then filter your date column so it is greater than or equal to RangeStart and less than RangeEnd. In Power BI Desktop you then define the policy: how much history to store and how much recent data to refresh. Once the dataset is published, Power BI creates and manages the partitions itself.

Why query folding matters

The date filter should be passed back to the source system as part of the query, a behaviour called query folding. If folding breaks, Power BI downloads the whole table and then filters it, which removes most of the benefit. Keep the filter step early in the query, avoid transformations that stop folding before it, and check that the step still folds.

Common mistakes

  • Filtering on a column that is not a true date or datetime type.
  • Placing the date filter after steps that prevent query folding.
  • Storing far more history than reports actually need.
  • Forgetting that records changed far in the past will not be picked up unless the refresh window or change detection covers them.

Getting it right

Incremental refresh is straightforward in concept but easy to get subtly wrong. If your refreshes are slow or your source system is struggling, a short review of the model and queries usually shows whether incremental refresh, or a different fix, is the right answer.

How we can help

← All insights

Let's build what's next.

Not sure where to start? Ask for a free dashboard review. We'll get back to you as soon as we can.

Prefer to talk it through? WhatsApp us Contact us by email

Our working hours overlap with the US, Europe and Asia-Pacific.

Start a project