Explain ratio totals and filter context

Understand the likely causes

• The total may divide combined successes by combined opportunities, while the apparent30% averages groups equally. CALCULATE can replace a same-column selection, and a denominator modifier can remove more context than intended.

Write the question before the fix

Define the numerator, denominator, grain and weighting. For groups with1/2 and9/90, the overall opportunity rate is10/92, whereas the equally weighted group average is30%. Both are valid arithmetic with different questions. Do not force the grand total to average visible cells merely because it seems familiar. Also define whether a no-opportunity group should return BLANK or a documented alternative. A display format cannot resolve a missing business definition. Preserve a small fixture so every candidate measure can be compared with an independent expected answer.

Trace the context stack

List report, page, visual and slicer filters, relevant interactions and active relationship paths. Then trace modifiers inside the measure. A direct ProductKey=A filter can replace B on that same column; KEEPFILTERS instead intersects and can leave no rows. Filters on other columns do not disappear automatically. A share denominator that removes product filters may still retain date, which can be correct. The record below separates the population choices from formatting. No DAX or report is opened, and expected BLANK behavior is reasoned under the lesson’s documented measure assumptions.

group1:1 success /2 opportunities
group2:9 successes /90 opportunities
overall contract:(1+9)/(2+90)=10/92
group-average contract:(50%+10%)/2=30%
filter review: replace, intersect or remove only the documented columns

Repair the smallest semantic mismatch

If the intended measure is an overall rate, divide aggregated numerator by aggregated denominator in the total context. If the business explicitly requires an equal group average, iterate the intended group population deliberately, including its blank policy. If a fixed-product result should respect existing selection, use the appropriate intersection semantics. Narrow an overbroad denominator removal to the documented dimensions. Do not weaken RLS or change relationships solely to make a percentage look right. Validate numerator and denominator separately, because duplicated facts can inflate both and leave an apparently correct ratio.

Verify totals and reader meaning

Test no selection, one group, conflicting product filters, a retained month and a zero denominator. Compare both row and total contexts with the expected population. Explain the weighting in a meaningful measure name or nearby report context so consumers do not sum overlapping or nonadditive results. This guide executes no DAX, UI or service operation. Its arithmetic record supports a review plan only. Actual engine evaluation and report interactions require separate authorized evidence, and a corrected local definition must be checked in dependent reports before a shared-model change is considered complete.

Keep it from recurring

• Keep unequal-denominator, blank and conflicting-filter fixtures with ratio measures. Document every denominator modifier and preserve independent amount reconciliations so percentage formatting cannot conceal a modeling defect.