You Clicked One Filter. Why Did the Total Change?
When a Power BI total changes after a selection, check the selected scope and expected result. Use a small control example before changing the measure.

A client clicks one channel in the report. The total changes. They ask whether the calculation is broken, and the report builder is not yet sure how to explain it.
The first task is to establish what the selection was intended to include. A changing total can be entirely appropriate, or it can reveal a design that does not match the reader’s question.
Predict the result before clicking
In a fictional freelance project, the report shows a small service week split by channel. The builder writes down the expected total for all channels, then the expected result for one channel on one day.
Using a recognisable example makes the interaction easier to inspect. A selected channel should not be judged only by whether the chart looks plausible. Compare the output with the source values for that same scope.
Record which filters are active while checking. A date selection left over from an earlier demonstration can explain a surprising figure without any change to the formula.
Explain the selected scope in ordinary language
Power BI measures are evaluated in context, including relevant filters. Microsoft’s model-design guidance describes measures and the role of filtering in report queries. The practical question for this report is simpler: which records should contribute to this number now?
Put that explanation in the measure’s working definition. If a total is deliberately meant to ignore a particular selection, make that intention explicit and test it. Do not change the formula merely to keep a preferred headline number on screen.
The reader should be able to recognise whether they are seeing a whole-business figure or a selected part of the data.
Check interactions one at a time
Start from a known report state. Apply one selection, compare the expected result and note what changed. Then clear it before testing the next interaction.
Include a selection that produces no eligible records. Decide how the report should communicate that situation, rather than allowing a blank to be mistaken for a confirmed zero.
For a more complicated model, relationships and measure definitions also need review. Keep the first learning example small enough that you can reconcile the outcome without relying on appearance.
Make the client part of the check
Ask the client to explain what they expect when selecting a channel. If they want the overall benchmark to remain visible beside the selection, that is a presentation requirement to discuss, not proof that all changing totals are wrong.
Document the agreed behaviour and the control example. It gives both parties a reference when the report is changed later.
A useful interactive report does more than respond to clicks. It makes the changing scope understandable so the reader can tell which question the number is answering.
Continue with a related question
Explore Build a Power BI Report You Can Check and Hand Over, or read Your Chart Looks Fine. Did the Data Load Twice?.
For a written practice option, see Power BI with Matze: contents and requirements.
Review the provider’s course details. Affiliate link: Superstore may earn a commission if you buy.
Technical reference: Microsoft Learn: Power BI model design.




Discussion 0
What did this article make you think about?