Why you only need to test with 5 users
Five users uncover ~85% of the usability problems a test can find; additional users mostly re-observe known issues.
Research question
How many participants does a usability test need before additional users stop paying for themselves?
Method
Nielsen & Landauer fit a Poisson model to problem-discovery data across projects: the share of problems found follows 1−(1−L)ⁿ, with average single-user detection rate L ≈ 31%.
Findings
With L≈31%, five users surface about 85% of findable problems. The curve flattens hard — the fifteenth user mostly watches the same failures as the first five.
Limitations
The 31% is an average across studies — complex products and diverse audiences lower it. The claim applies to qualitative, iterative testing only: quantitative benchmarks need 20+ users, card sorting needs ~15, eyetracking needs ~39. Distinct user groups each need their own ~5.
What it means in practice
Budget for three rounds of five rather than one round of fifteen: find, fix, re-test. The redesign after each round is where the value is created.