Skip to main content
You know that moment when you watch someone use your product for the first time and they click completely the wrong thing? And then do it again? And you’re sitting there, biting your tongue, desperately wanting to say “no, the button is right there” 😅 That’s usability testing. And that discomfort is the whole point.

​
What it is

Usability testing is the practice of observing real users attempting real tasks with your product - without help, guidance, or explanation. You’re testing whether the design communicates clearly enough for someone unfamiliar with it to figure it out on their own. It’s distinct from value testing: usability asks can they use it?, value asks do they want it?. You need both, but they’re different questions that require different sessions. Steve Krug literally wrote the book on this - Don’t Make Me Think is the classic. His central argument: good usability is invisible. Users shouldn’t have to think about how to use your product - they should just be able to use it. Every moment of confusion is a design failure, not a user failure.

​
How to run a usability test

Recruit the right people - Users who match your target persona. Five users will surface most of the significant usability issues (Krug’s rule, and it holds up). Give them a task, not a tour - “Find a product and add it to your cart” not “let me show you how the checkout works.” Realistic tasks reveal real problems. Tours hide them. Shut up and watch - This is the hard part. Don’t hint, don’t explain, don’t rescue them when they get stuck. The stuck moments are the data. If you can manage it, ask them to think out loud as they go - “what are you looking for? what do you expect to happen here?” Note, don’t judge - You’re collecting observations, not verdicts. “User clicked the wrong button three times” is an observation. “The navigation is broken” is a conclusion. Start with observations; conclusions come later. Debrief immediately - Talk through what you saw with your designer and engineer right after. Fresh observations shared together beat a written summary read separately every time.

​
Moderated vs. unmoderated

Moderated - You’re in the session (in person or remote), watching live and able to ask follow-up questions. Richer insight, more time-intensive. Unmoderated - Users complete tasks asynchronously via a tool like Maze or UserTesting. Faster and cheaper, scales better, but you lose the ability to dig deeper when something interesting happens. For early-stage testing, moderated is almost always worth the extra effort. For validating specific flows on a more mature product, unmoderated works well.

​
How many sessions do you need?

Krug’s answer: five. Nielsen’s research backs this up - five users will uncover around 85% of usability problems. More sessions reveal diminishing returns. The practical implication: don’t wait until you have twenty participants lined up. Run five sessions, fix what you find, then run five more if needed. Iterating is faster and more useful than one large study. Lesson learned: the most revealing usability sessions are always the ones where someone does something you were absolutely certain nobody would ever do. They always do it 🤦