How many questions a survey should have before people abandon it
There's no fixed number of questions that works for every survey, because abandonment depends on the number of questions, how long each one takes, the context it's sent in, and how invested the respondent is in the topic. What is consistent across most studies on this is that abandonment doesn't rise linearly: it stays relatively low up to a point, then accelerates.
Where that point tends to be
As a general reference (not a strict rule), surveys with 5 to 10 questions tend to keep completion rates high, between 10 and 15 questions abandonment starts becoming noticeable, and past 20 questions the drop-off is usually steep unless the audience is highly invested (employees answering a mandatory HR survey, for example). Estimated time matters more than the raw count: 10 single-choice questions take less effort than 5 open long-text questions.
Why abandonment accelerates instead of rising evenly
Early on, the person has invested little time, so the cost of abandoning is low and they keep answering out of inertia. Past a certain point, they've invested enough time to feel like they're "almost done", which keeps them going a bit longer — but if the survey keeps stretching past what they expected, the frustration of having invested time without seeing the end becomes the reason to close it instead of finishing it.
How to cut length without losing what matters
Before cutting questions at random, it helps to split them into two groups: the ones that will change a real decision, and the ones that are "interesting to know" but not actionable. Only the first group justifies the abandonment cost it generates. Conditional logic (showing or hiding questions based on a previous answer) also helps a lot here: instead of asking everyone about something that doesn't apply to them, that section is skipped entirely for the people it doesn't apply to.
The progress bar matters more than it seems
When someone doesn't know how much is left, they tend to overestimate what remains and abandon earlier than necessary. Showing a progress bar or the current question number out of the total reduces that uncertainty-driven abandonment, without changing a single question in the survey.
When a long survey is actually fine
There are contexts where a 20+ question survey works fine anyway: when answering is mandatory or organizational (performance reviews, employer-driven workplace climate surveys), when there's a real incentive for completing it, or when the topic genuinely matters to the respondent. Outside of those cases, it's safer to assume every additional question past the tenth has a real cost in completion rate, and decide whether it's worth paying.