What Is Usability Testing and Why It Matters

The short answer

Usability testing is a method where you give real users specific tasks in your product and watch how they complete them. For example: "Order delivery for tomorrow" or "Find out how much the team plan costs."

The key word is observation, not opinion. People are bad at predicting their own behavior and will happily say "it's all clear" right after clicking the wrong thing five times. A test shows what people actually do.

You can test almost anything: a clickable Figma prototype, a live website, a mobile app, even paper sketches of screens.

Moderated vs unmoderated

Moderated sessions work best early, when you don't yet know where the problems are. Unmoderated sessions are useful when you want to check a specific flow with more people.

Remote vs in-person

• In-person: the participant and facilitator share a room. You see facial expressions and body language, and it's easy to help with equipment. Good for physical devices and for audiences who struggle with video calls.
• Remote: over a video call with screen sharing. Cheaper and faster, lets you reach people in other cities, and the participant uses their own device in a familiar setting.

These two dimensions are independent: a remote test can be moderated (a call with a facilitator) or unmoderated (a recorded session through a tool).

What problems it uncovers

• Confusing navigation: people can't find a section or look for it somewhere other than where you put it.
• Unclear wording: a "Continue" button leads somewhere unexpected; jargon only the team understands.
• Invisible elements: a key button blends into the background or sits below the fold.
• Form friction: unclear password rules, data wiped after an error, awkward phone number input.
• False expectations: the user believes something was saved when it wasn't.

Usability testing does not tell you whether people want the product, and it does not replace analytics. It answers a narrower question: can people use it?

Why five users are often enough

The idea was popularized by Jakob Nielsen of Nielsen Norman Group. The logic is simple: the first participant reveals many problems, the second repeats some and adds a few new ones, the third adds even fewer. With each additional person you mostly see repeats.

That's why several small rounds beat one big test with 20 people: test with five, fix what you found, test again. Each round surfaces problems that were previously hidden behind bigger ones.

Two important caveats:

• Five people refers to one homogeneous group. If your product serves different audiences (buyers and sellers), each needs its own sample.
• The rule applies to qualitative problem-finding. Quantitative metrics such as task time or success rate need more participants.

How to run your first test

1. Define the goal. What are you checking: checkout, sign-up, search?
2. Write 3 to 5 tasks as realistic scenarios without hints. Not "click Cart," but "you've picked a product, now buy it."
3. Recruit participants from your target audience, not colleagues from the next desk.
4. Ask them to think aloud so you hear what they expect to see.
5. Don't help. If someone is stuck, ask "what did you expect?" instead of pointing to the button.
6. Record observations, not interpretations: "looked for payment in the profile menu," not "the menu is bad."
7. List the problems and prioritize them by how badly they block the task.

Common mistakes

• Testing on colleagues and friends who know the context and avoid criticism.
• Asking leading questions like "That was easy, right?"
• Testing only at the very end, when changes are expensive.
• Drawing conclusions from a single participant.
• Running the test and never fixing what it found.