Results by variant

For a survey tied to an experiment, the results screen opens a breakdown by variant straight away: how many visitors each variant showed the survey to, how many answered, the response rate, and each score question’s net score or average per variant.

Choice questions get a table of their own, every option against every variant, so you can see whether the visitors on the new page came for different reasons or picked different answers, not only whether they scored differently.

Any other survey can be compared the same way, by any experiment its visitors happened to be in. The variant is recorded on every answer when it is given, so the split is there afterwards for experiments nobody thought to connect at the time.

It is descriptive on purpose. There is no significance test and no winner, and a variant with fewer than thirty respondents is flagged, because a handful of answers is not a verdict. The experiment’s own results decide the outcome; the survey explains it.