How to user test your app with just 5 people
- What the model claims: With an assumed 31% per-user discovery rate, Nielsen’s five-user model estimates that five similar users reveal about 85% of the usability issues that setup can expose – not 85% of bugs, accessibility needs, or problems across distinct audiences.
- One tester beats zero: A single relevant participant already surfaces about a third of what there is to learn, because “the difference between zero and even a little bit of data is astounding.”
- Small rounds, not big studies: Later participants increasingly repeat issues already observed, so test, improve, and test again – and give distinct groups like novices and VoiceOver users their own coverage.
- Apple’s three questions: Ask “Do you know how to XYZ?”, “Is it easy to XYZ?”, and “How can we make this better?” rather than “do you like it?”, and while watching, don’t argue, defend, or dismiss.
- Give them a task, not a tour: Set one specific task (“sign up for the app, then log your first workout”) and find your app’s equivalent of Mercury Weather’s Two-Second Glance Test.
- Treat findings as hypotheses: Fix blockers and high-severity patterns first, weigh severity and recurrence before changing deliberate workflows, and give more time to anything users uniformly praise – that’s how Halide’s reduced-processing option grew into Process Zero.