The usability sample size calculator.
How many participants a usability test needs, and what share of problems each extra user finds. Live curves, honest numbers, and nothing you type leaves your browser.
How many users does a usability test need?
The chance one participant trips over an average problem. Research puts the average near 31%, so leave it there unless you have a reason not to.
How many people you sit with, one at a time. Drag the slider or click the curve below.
Expected share of problems found
84%
With 5 participants, expect to surface about 84% of the problems in the tasks you test. The rest stay hidden until later rounds.
Jump to a coverage target
| Participants | Problems you can expect to find | Still hiding |
|---|---|---|
| 1 | 31% | 69% |
| 2 | 52% | 48% |
| 3 | 67% | 33% |
| 4 | 77% | 23% |
| 5 | 84% | 16% |
| 6 | 89% | 11% |
| 7 | 93% | 7% |
| 8 | 95% | 5% |
| 9 | 96% | 4% |
| 10 | 98% | 2% |
| 11 | 98% | 2% |
| 12 | 99% | 1% |
Everything runs in your browser. Nothing you type leaves your device.
Rule of thumb: five users find about 85%.
At average problem visibility, five participants surface roughly 85% of the problems in the tasks you test. It falls straight out of the discovery formula, and three decades of follow-up studies back the shape. Run several small rounds with fixes between them rather than one big round.
Why do extra users find less and less?
The maths behind the curve is the discovery formula from Nielsen and Landauer (1993): with average problem visibility of 0.31, the share found after n users is 1 minus (1 minus 0.31) raised to n. Plug in 5 and out comes 85%. Nielsen’s own write-up walks through the intuition, and Lewis (1994) later confirmed the model holds even when problems vary in visibility.
The words, in plain English.
- Problem visibility
- The chance that one participant trips over a given problem, written as p. Research averages it near 0.31, meaning a typical problem catches about one participant in three. Obvious problems score higher, subtle ones lower.
- Discovery share
- The slice of all problems your study can expect to surface: 1 minus (1 minus p), raised to the number of participants. Five users at average visibility lands near 85%.
- Diminishing returns
- Each extra participant finds fewer new problems than the last, because the easy problems are already caught. The sixth user adds far less than the second, which is why the curve flattens.
- Qualitative vs quantitative
- Qualitative studies find problems, and five users do that job. Quantitative studies measure how often things happen, and that counting needs twenty participants or more. This calculator serves the first job.
- Iteration
- Testing, fixing, then testing again. Three rounds of five with fixes between rounds find far more than fifteen users in one go, because each round hunts fresh ground.
Should I test 15 users at once, or 5 users three times?
This is the point the slogan skips: five users is a round size, not a project total. Nielsen (2000) recommends exactly this rhythm, several small tests with redesigns between them, because each round hunts fresh ground. It also matches how research engagements run here: fieldwork, fixes, then back into the field.
When does the five-user rule not apply?
Distinct audiences barely overlap in the problems they hit, so budget three to four users per group (Nielsen, 2000). And when the question turns quantitative, how many users complete checkout, what the satisfaction score is, reach for the survey sample size calculator instead. Finding problems and counting them are different jobs with different sample sizes. The quick guide to UX research shows where each method fits.
Questions the calculator gets asked.
Five, for a standard qualitative round. Research going back to the early nineties finds five participants surface around 80 to 85% of the problems in the tasks tested. Add three to four more per extra audience group, and switch methods entirely when you need numbers rather than findings.
A 1993 paper by Nielsen and Landauer modelled problem discovery mathematically and found the average problem catches about 31% of participants. Feed that into the discovery formula and five users lands at roughly 85%. Later studies by Virzi, Lewis, and Faulkner confirmed the shape, with Faulkner finding five users caught 85% on average across real projects.
Less risky than it feels, riskier than the slogan admits. Faulkner found five users caught as little as 55% in the worst case, and an often-quoted study by Spool and Schroeder found the first five catching only 35% on a complex task (reported in Faulkner, 2003). Serious teams run several small rounds with fixes between them rather than one round of five and done. The calculator shows the average; iteration covers the variance.
Because users eleven through fifteen mostly rewatch problems the first five already found. Fix what round one surfaces, then test again: the next five hunt through the fixes and find what remains. Three rounds of five beats one round of fifteen on total problems found.
No. All the maths runs in your browser on your device. Nothing you enter is transmitted, stored, or logged, which keeps the tool comfortable to use under POPIA.
Calculator reviewed October 2026.
Where do these numbers come from?
Every figure on this page traces to one of the studies below, cited Harvard style. Nothing here is agency folklore.
- Faulkner, L. (2003) ‘Beyond the five-user assumption: benefits of increased sample sizes in usability testing’, Behavior Research Methods, Instruments, & Computers, 35(3), pp. 379–383.
- Lewis, J.R. (1994) ‘Sample sizes for usability studies: additional considerations’, Human Factors, 36(2), pp. 368–378.
- Nielsen, J. (2000) ‘Why you only need to test with 5 users’, Nielsen Norman Group, 19 March. Available at: https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/ (Accessed: 12 October 2026).
- Nielsen, J. (2012) ‘How many test users in a usability study?’, Nielsen Norman Group, 3 June. Available at: https://www.nngroup.com/articles/how-many-test-users/ (Accessed: 12 October 2026).
- Nielsen, J. and Landauer, T.K. (1993) ‘A mathematical model of the finding of usability problems’, Proceedings of the ACM INTERCHI’93 Conference on Human Factors in Computing Systems, Amsterdam, pp. 206–213.
- Virzi, R.A. (1992) ‘Refining the test phase of usability evaluation: how many subjects is enough?’, Human Factors, 34(4), pp. 457–468.
Let's talk
Ready to stop guessing?
Tell us what you're trying to figure out. We'll design the study that answers it: UX, market, or both.