Fix mismatched KPI totals

Power BI users often hit a totals problem: the numbers in each row look right, but the total card or waterfall does not match. Microsoft says the DAX language behind Power BI evaluates measures inside a filter context, and the `CALCULATE` function changes that context before returning a result. (learn.microsoft.com) Joseph Opene, whose GitHub profile lists Power BI work, posted a July 2026 example showing that wrapping variance logic in `CALCULATE` corrected mismatched totals in sales variance and year-over-year trend measures. The post centered on a dashboard where bridge totals were not tying out across slicers and time selections. (github.com) (x.com) The underlying issue is common in Power BI because a total row is not always the sum of visible rows. A measure is recalculated at the total level under a different set of filters, so a variance formula that works for one product, month, or region can return a different answer when the report asks for “all of them at once.” (learn.microsoft.com) (github.com) `CALCULATE` is the main tool for fixing that behavior because it lets analysts add, replace, or remove filters before a measure runs. Microsoft’s documentation lists filter modifiers such as `REMOVEFILTERS`, `ALL`, `ALLEXCEPT`, and `KEEPFILTERS`, which are commonly used to force a measure to evaluate under the exact time period or category scope the analyst intended. (learn.microsoft.com) That matters most in year-over-year reporting, where one side of the comparison usually depends on a shifted date context. In a sales variance bridge, the current-period measure and the prior-period measure have to be evaluated under consistent filters or the pieces in the bridge will not sum back to the headline variance. (learn.microsoft.com 1) (learn.microsoft.com 2) Microsoft’s Power BI training materials teach the same concept in formal labs: analysts create measures, test them in a matrix, and then modify filter context with `CALCULATE` so report totals behave as intended. That training treats filter control as a core reporting skill, not an edge case. (github.com) The practical result is less about syntax than trust. When a chief financial officer sees a variance card that disagrees with the bridge beneath it, the report stops being a management tool; when the same measures reconcile across product, region, and date filters, the dashboard can support decisions instead of arguments over arithmetic. (learn.microsoft.com)