A research method where representative users attempt to complete tasks with a product while researchers observe and record what happens. Unlike surveys or analytics, usability testing reveals why problems occur, not just that they exist.
Common contexts
- Testing a new onboarding flow with five participants before it ships to validate the task completion rate
- Running sessions on a competitor's product to identify industry patterns that users will bring as expectations to your design
- Conducting guerrilla testing in a café with printed paper prototypes to get same-day directional feedback
Use when
Run usability testing whenever you have a design ready to be challenged — even a rough prototype with five participants will surface the most significant interaction failures. Earlier is almost always better: testing a wireframe costs a day; testing a shipped feature costs a sprint.
Avoid when
Don't use usability testing to validate strategic or business decisions — whether to build a feature, which market to target, or how to price a product are research questions that require different methods. Using usability tests for strategic questions wastes sessions and produces the wrong kind of evidence.
Five participants is enough to find the majority of usability problems — the value of adding more participants comes from seeing edge cases, not from the first five sessions, which is why that budget is almost never worth arguing about.
Real-world examples
- Nielsen Norman Group's research established that 5 users in a moderated usability test reveal approximately 85% of the most critical usability problems — a finding that democratised usability testing by making it affordable for small teams.
- Google's UX research team runs weekly 'Friday Research Reviews' where any product team can sign up to observe 3 live usability sessions, maintaining a culture where design decisions are grounded in observed user behaviour.
- ASOS increased mobile checkout conversion by 50% after a single round of usability testing identified that their guest checkout flow required users to create an account mid-purchase — a blocking issue invisible in analytics data alone.