Definition
Usability testing is a UX research method in which representative users attempt specific tasks on a product, prototype or website while a researcher observes — surfacing friction points, confusion, and breakdowns that the design team would never see on their own.
Usability testing is the highest-ROI single research practice available to product teams. Five users surface roughly 80% of usability problems in any flow — at a cost in the range of $1,500-$5,000 — and the value of catching those problems before they hit production is typically 10-50x that.
The discipline of usability testing is in the observation, not the task design. Watch what users do, not what they say. Note where they hesitate, click the wrong thing, ignore the right thing. Stated preferences (what users say they want) are notoriously unreliable; observed behaviour (what they actually do) is the data.
Origin
Usability testing as a formal practice was popularised by Jakob Nielsen in the 1990s, building on human-factors research traditions. The '5-user' heuristic comes from Nielsen's empirical research showing the curve of problem discovery flattens after the fifth participant.
How it works
- Define 3-5 specific tasks that represent core user goals.
- Recruit 5-8 representative users.
- Set up the prototype, screen recording, and observation environment.
- Brief the user — explain you're testing the design, not them.
- Read each task; observe; don't lead; capture verbatim quotes and hesitation points.
- Synthesise patterns across users; map issues to severity; recommend fixes.
When to use it
Use when
- Before launching any new flow.
- When analytics show unexpected drop-off and you need to understand why.
- When the team disagrees about whether something is intuitive.
Skip when
- For changes within well-established patterns the user already knows.
- When you don't have time to act on the findings.
Key metrics
- Task completion rate (target 80%+)
- Time to complete each task
- Error count per task
- Severity-rated issues found
Examples
- Five users in five hours caught seven issues we'd missed in three months of design.
- The task completion rate was 40% — meaning more than half of real users would have failed at the most basic task on this screen. Easy fix once we saw it.
- Users kept saying 'oh wait, this is what I should click?' — that's the sound of a UI failure.
In practice at Makreate
Makreate UX engagements run usability tests at the prototype stage and again pre-launch. We use moderated remote testing for most work (Zoom + screen share + recording), which lets us pull users from any geography in the client's target market within 48 hours. Each round typically catches 5-12 issues; most are quick fixes in Figma; all of them would have cost more downstream.
UX Design →Common mistakes
- Asking users to evaluate the design instead of complete tasks.
- Recruiting users who don't match the actual target audience.
- Leading questions or hints during the session.
- Only running once, instead of testing at multiple iterations.
- Not prioritising findings by severity — fixing the wrong thing first.
Frequently asked
How many users do I need for usability testing?
5-8 users surface 80% of issues in a single round. More users add diminishing returns per study.
Should I test moderated or unmoderated?
Moderated catches more nuance and follow-up; unmoderated is cheaper and scales. For high-stakes flows, moderated; for quick checks, unmoderated tools (Maze, Lookback) work well.
What if I can't recruit my exact target users?
Close-enough proxies still surface most issues — the bigger risk is testing with no one. Generic usability flaws show up regardless of who's testing.