Response Rates

The share of responders who finish every question is about 57% for two questions and about 45% for three or more. The number of questions does not change how many people start.

A model that holds the other choices constant keeps these effects:

• An open-text first question cuts the odds of a response to about a fifth.
• An event trigger about doubles them.
• A centered modal (middle_center or center) raises them about 1.8 times.
• A choice question first and a targeted audience each raise them about 1.4 to 1.5 times.
• A popup delay over 15 seconds cuts them by about a quarter.

The mobile difference, the 0–10 rating scale, and the number of questions do not hold up once the other choices are held constant.

Limits

• These are correlations. Teams that set up event triggers may also write better surveys in other ways.
• With partial responses off, a person counts only after finishing every question, so surveys with partial responses on look better partly because of how they count. Turn it on to keep first answers, not to raise the rate.
• A centered modal interrupts the user more than a corner popover. Its higher response rate does not measure what the interruption costs. Suggest it only when the user wants the most responses from a short, important question.
• API surveys depend on how each app captures survey shown, so their rates are not comparable. Hosted forms count only people who opened the link.