How to Run a Small Website Usability Test
Ask someone to find your returns policy without telling them which menu to open. Watch where they look first and where they hesitate. A few sessions won’t tell you what percentage of visitors struggle, but they can show you a specific obstacle worth fixing and testing again.
A practical example
Give a likely customer this task: “You bought a shirt yesterday and it is too small. Find out whether you can return it and who pays postage.” Do not say “Click Returns in the footer.” Finding that link is part of what you want to observe.
A short session you can run
Invite three people who resemble the intended audience for a first round; three is a practical starting point here, not a representative sample. Explain that you are testing the site, not them. Use fictional order details, offer any access support they need, and ask before recording. Let them stop whenever they wish.
Give each person the same task and let them try it on a device they would normally use. If someone pauses, ask “What are you looking for?” rather than pointing at the answer. If you help them finish, mark where you helped. The GOV.UK guide to moderated usability testing describes how to design realistic tasks and observe without giving the route away.
A useful note reads “Opened Delivery twice, then searched for ‘postage’,” not “bad at navigation.” If two people find the policy but cannot find who pays, changing the menu may not help. Clarify that sentence and try the task with fresh participants. Report what you observed, not a claim that two-thirds of all customers have the problem.
Choose one task
Pick one task that matters to your business. Examples include finding the price of a service, booking a call, locating the returns policy, or buying a product. Write the task as a short instruction, such as “Find out whether you can get a refund after thirty days.” Keep the wording neutral and avoid naming the menu item or button that the visitor should use. If you name the path, you are testing your navigation labels rather than the visitor’s own instincts, which is the point of the test.
Recruit three people who fit your audience
Look for people who resemble your real visitors, not people who know your business. Friends, family, and colleagues often know too much, so they may find a route that a stranger would never take. Some people will be happy to help for a short session, perhaps in exchange for a coffee card or a thank-you note. Keep the session short, usually twenty to thirty minutes, and make clear that you are testing the website, not the person. Ask whether they are comfortable with notes being taken, and explain how you will use the information.
Set up the session
Use the same device and browser for each session, and make sure the site is in the state a real visitor would see. Clear saved logins, so that the person does not land on a page they would not normally reach. Start with a brief explanation that you are testing the site and not the person, and that there are no wrong answers. Ask them to think aloud as they go. If a session runs over, let the person decide whether to continue, and do not push past the time you agreed on.
Observe without coaching
Ask the person to share their screen, or sit beside them if you meet in person. Give the task instruction once, then stay quiet. It is tempting to point at the right menu or explain a label, but that removes the most useful information. Ask open questions only after the attempt, such as “What were you expecting to see there?” Write down what the person does and says, including small pauses, scrolling back and forth, and comments like “I think it is here.” The sentences people say out loud are often the clearest description of a problem.
Note where people get stuck
After the three sessions, review your notes and mark each point of friction. A blocker is a step where the person could not complete the task without help, or where they went down the wrong path and did not recover. A minor issue is a moment of confusion that they resolved quickly. Count how many people hit each blocker. Two or three people hitting the same problem is a strong signal, even with a small sample. A single person getting lost might reflect their own habits, so note it but do not rush to redesign the page for it.
Fix one blocker and retest
Resist the urge to fix everything at once. Choose the most serious blocker, change one thing that addresses it, such as a clearer button label, a shorter form, or a link placed where people expect it. Then run the same task with three new people. The retest shows whether the change helped, and it keeps you from making several changes at once and not knowing which one worked. If the blocker remains, try a different fix and test again. Improvement comes from repeated small rounds, not from one large redesign.