Guide · Running a test

How many testers do you actually need?

Fewer than you think, but not for the reason most people quote. Here's the research behind "test with five users", what it gets right, and when you need more.

6 minute readUpdated 18 November 2026

In short

  • Jakob Nielsen's classic finding: five testers find about 85% of usability problems, if each one finds about 31%.
  • His real advice is to test in small rounds: five people, fix, then test again.
  • Use more when you have different kinds of users, or you're hunting rare problems.

01Where "five users" comes from

In 2000, Jakob Nielsen published "Why you only need to test with 5 users". It built on a model he and Tom Landauer published in 1993: if a typical tester runs into about 31% of the problems in a design, each new tester finds fewer new ones, because they hit the same problems as earlier testers.

Play with the number of testers and watch how quickly the curve flattens:

85%of problems found, on average

Each dot is one usability problem. Based on Nielsen and Landauer's model, where one tester finds about 31% of problems.

02The part people forget: test again

Nielsen didn't say five testers is enough, full stop. He said that if you could afford 15 testers, you should run three rounds of five instead: test, fix what you found, then test the new version.

Spend this budget on 3 studies with 5 users each.Jakob Nielsen, Nielsen Norman Group

The second round checks your fixes worked and finds the problems the first round's big ones were hiding. That's why our Launch pack is ten testers, then five more after you ship your fixes.

03Where the rule breaks

The 85% is an average, and the average hides a lot. When Laura Faulkner tested the rule in 2003, groups of five found anywhere from about 55% to nearly all of the problems, depending on which five people you happened to get.

  • Different kinds of users find different problems. If buyers and admins both use your app, you need a few of each.
  • Rare problems need many more people. Jeff Sauro at MeasuringU works out that to be 85% sure of seeing a problem that affects only 1% of users, you'd need around 189 testers.
  • Measuring things, like how long a task takes or what share of people succeed, needs bigger samples than finding problems does.

04So how many should you use?

Your situationTesters
A quick check of one page or flow3 to 5
One kind of user, a whole onboarding5 to 10
Two kinds of user5 or more of each
Checking your fixes worked5 new people

We use ten per standard test because it covers some variety in your users and shows patterns clearly. When eight of ten testers miss the same button, there's nothing to argue about.

Start small, for free.

Your first test is three matched testers and a short fix list. No card.

Keep reading

Related guides.

Get started

See your app through a stranger's eyes.

Your first test is free: three matched testers and a short fix list. No card.

Rather be on the other side? Become a tester