Why do the rows of a breakdown not add up to the total?

There are three usual causes: an other row that absorbs the long tail and is sometimes hidden, rows where the property is unset and are dropped, and above all the fact that unique counts are not additive. A user whose property value changed during the window appears in two rows but is counted once in the total.

Breakdown by Property

Beginner

Updated Jul 11, 2026 Shahin editorial 2 min read

PronunciationEN/ˈbreɪkdaʊn baɪ ˈprɑːpərti/

Breakdown by property splits a single metric into one series per value of a chosen property, turning one aggregate number into a comparable set of numbers.

For the split to be meaningful the property must attach to the same rows being counted. Breaking a user-level metric down by an event-level property forces the tool to pick a rule, such as first value, last value or any value, and different tools pick differently, which is why the same breakdown can produce different totals in two products. High-cardinality properties create a long tail that the report collapses into an other row, so the visible top rows are not the whole population. The common misreading is treating a breakdown of absolute counts as a breakdown of rates: the largest segment usually leads on volume simply because it is the largest. The subtler risk is Simpson's paradox, where a metric improves inside every segment while the overall figure worsens because the mix of segments changed.

FAQ