Segment Comparison
Intermediate PronunciationEN/ˈsɛɡmənt kəmˈpærɪsən/
Segment comparison places two or more defined groups of users side by side in one report so their behaviour can be read against a shared metric.
Each segment is a filter definition evaluated over the same date range and the same metric, and the report draws one series per segment. The value is in providing a reference point: a single group's number means little until it sits next to another group's. The common error is reading the gap between two segments as cause and effect; users selected themselves into these groups rather than being randomly assigned, so a difference is a correlation. Segments can also overlap, which means one person may appear in two series and the shares need not add up to one hundred.
FAQ
No. People were not randomly split between the segments, so any difference may come from who those people already were rather than from the attribute used to define the group. Segment comparison generates a hypothesis; establishing cause requires random assignment.