Question
"Keep it under 10 minutes" is folk wisdom in survey research, repeated everywhere and sourced nowhere. We wanted the actual shape of the drop-off curve: is there a cliff, where is it, and what does each additional minute cost in completes and in quality?
Data
890,000 survey starts across 3,900 studies fielded between June 2025 and May 2026. For each start we recorded the exact abandonment point (question index and elapsed time), the survey's estimated length, device class, and the final quality score of completed responses. Studies with screener-based terminations were handled separately so quota screen-outs do not masquerade as drop-off.
The shape of abandonment
| Estimated length | Completion rate | Median abandonment point |
| Under 5 min | 91% | first question |
| 5-8 min | 89% | first question |
| 8-12 min | 83% | ~60% progress |
| 12-18 min | 71% | ~50% progress |
| 18-25 min | 58% | ~40% progress |
Three findings researchers should act on
- The first question is the biggest cliff. Across all lengths, 31% of total abandonment happens on question one. Respondents are not quitting your survey - they are bouncing off its opening. A hard or personal first question costs more completes than four extra minutes of length.
- Length is free until ~8 minutes. Completion is statistically flat from 2 to 8 minutes. Cutting a 7-minute survey to 5 buys nothing; use the room for the questions you actually need.
- After 8 minutes, each additional minute costs ~2.3% completion - and the losses are not random. Drop-off skews toward busier, younger, and mobile respondents, so long surveys do not just shrink your sample, they quietly reshape it.
Length degrades quality before it kills completion
Among respondents who did complete, quality decayed with position: in surveys over 15 minutes, straight-lining in the final third was 2.1× the rate of the first third, and median open-text length fell 23% from the first open question to the last. A 20-minute survey does not just lose 40% of starters - the survivors give you worse data at the end than the beginning.
The estimate must be honest
Surveys whose real median completion time exceeded the advertised estimate by more than 50% showed double the mid-survey abandonment. Respondents forgive length; they do not forgive being misled about it.
Recommendations
- •Open with an easy, engaging, obviously relevant question - never a grid, never income
- •Design to 8 minutes; treat every minute past it as a purchase costing ~2.3% of your sample
- •Put must-have questions in the first two-thirds; treat the final third as at-risk territory
- •If the instrument cannot fit 8 minutes, split it into two studies - sample math usually favors two short precise studies over one long degraded one