Guide · UX and design

Usability testing: how to test your website or app with 5 users

In short: a usability test is when you watch real users complete concrete tasks in your product while thinking aloud. Five users are enough to find most serious problems, and a test can be planned, run and analysed in a few days. This guide gives you the whole method: recruiting, test script, good and bad tasks, what to measure, remote or in person, and what it costs.

12 min read · Updated 1 October 2026

A webshop owner sits behind a screen and watches the first test participant try to find out what shipping costs. She looks in the menu, scrolls to the bottom, opens the terms and gives up after two minutes. The owner has had the site for four years and never saw the problem, because he knows the shipping price appears in the basket. In 30 minutes he learns more about his shop than from a year of Google Analytics data. That experience is why usability testing is the method we recommend most often.

Usability testing needs neither a lab nor a big budget. It takes five people who resemble your users, some good tasks and the discipline to stay quiet and observe. In this guide you get a complete method you can use yourselves, and a picture of when it makes sense to get help.

What is a usability test, and what can it reveal?

A usability test is a session where one participant completes realistic tasks in your website, app or prototype while a facilitator observes and asks the participant to think aloud. The test reveals where people hesitate, misunderstand, click the wrong thing or give up. It also shows which words they use, what they expect to find and what gives them confidence. What the test measures poorly is opinions and preferences. Focus on what people do, and treat what they say they like with care.

Why are five users enough for a usability test?

The well-known rule of thumb from usability researcher Jakob Nielsen is that five users find most of the problems in a design. The explanation is simple: serious problems hit almost everyone, so they show up with the first two or three participants. Participants four and five confirm them and add a few new ones. After that you learn less with each extra person. It is therefore better to test with five, fix and test again with five new people than to test fifteen at once. If you have several very different user groups, such as customers and staff, test with three to five from each group.

How do you run a usability test step by step?

1. Set a goal

What do you want to learn? For example whether new customers can book without help.

2. Choose participants

Five people who resemble your real users.

3. Write tasks

Three to five realistic tasks written as scenarios.

4. Pilot

Try the script on a colleague and fix unclear tasks.

5. Run the sessions

30–45 minutes per participant with recording and a note-taker.

6. Analyse

Collect findings, count how many hit each problem and rate severity.

7. Fix and retest

Fix the most serious problems first, and test with five new people.

Seven steps from goal to fixes. A small project can be done in a week.

How do you recruit test participants?

Existing customers through the newsletter, a receipt or customer service.
A short survey on the website asking whether people would like to help.
Your network and social media, but avoid friends and family who will be polite.
For B2B: your sales team can ask customers they have a good relationship with.
Recruitment agencies or testing platforms when the audience is hard to reach.
A few screening questions to make sure participants match the target audience.

How do you write good test tasks?

Weak tasks

“Click Booking and choose a time.”
“Find the About us page.”
“Do you think the home page looks nice?”
“Use the filter to find a sofa.”

Strong tasks

“Your car is due for inspection next week. Book a time that suits you.”
“You want to know whether the company works in your area. Find out.”
“Based on what you see, what do you think the company offers?”
“Find a grey sofa under DKK 8,000 that can be delivered before Christmas.”

Tasks should describe a situation and a goal and never reveal the route.

A test script you can copy

Welcome (2 min): “It is the product we are testing. You cannot do anything wrong.”
Consent (1 min): explain the recording, what it is used for and when it is deleted.
Background (5 min): two or three questions about how the person handles the task today.
Think aloud (1 min): “Please say what you are looking for and what you expect.”
Tasks (20–30 min): read one task at a time, and hold back from helping.
After each task: “How easy or difficult was that on a scale from 1 to 7?”
Wrap-up (5 min): “What was most confusing?” and “What did you miss?”
Thanks and gift, and ask whether the participant may be contacted again.

The hardest part for most people is staying quiet. When the participant gets stuck, ask “what are you thinking now?” or “what would you do if I weren’t here?” Answer questions with a question. It is in the silence that you learn the most. Many of the same techniques are used in user interviews, which are good preparation before you design anything.

What should you measure in a usability test?

Five measures that give a clear picture without advanced statistics. With five participants the numbers are indicative and best used alongside the observations.

Measure	What it shows	How to record it
Completion	Whether the task can be done	Completed, completed with difficulty or failed
Time on task	How efficient the flow is	Stopwatch from reading the task to completion
Errors and detours	Where the design misleads people	Note wrong clicks and back-button use
Perceived difficulty	How much effort it felt like	Scale from 1 to 7 after each task
Quotes	Why the problem occurs, in the user’s words	Write verbatim, with a timestamp in the recording

Remote or in-person usability testing: which should you choose?

Three ways to test. Moderated remote testing is the most practical starting point for most teams.

Format	Advantages	Drawbacks	Best for
In person, moderated	Body language, physical products, older users	Travel, a room, fewer sessions per day	Field apps, kiosks, complex professional systems
Remote, moderated	Quick, cheap, participants from anywhere	Technical issues with screen sharing	Websites, web apps and prototypes
Remote, unmoderated	Many responses, no facilitator	No follow-up questions	Measuring simple tasks in a finished product

Testing mobile apps takes a little extra planning. In person you can film the screen from above or use screen mirroring. Remotely, the participant can share their phone screen in a video call. Always test on the participant’s own phone when possible, because screen size and habits affect the result.

How do you analyse and prioritise the results?

Write each problem on a sticky note or a spreadsheet row, and note how many of the five participants hit it. Then rate the severity of each problem. A problem that makes three out of five give up is critical. A problem that delays one person a few seconds is cosmetic. Fix the critical ones first, even if the cosmetic ones are easier. Present the results with short video clips: management is convinced faster by two minutes of video than by ten pages of report.

Critical: the user cannot complete the task. Fix before launch.
Serious: the user completes it with great difficulty or errors. Fix in the next iteration.
Minor: the user hesitates or takes a detour. Fix when there is time.
Cosmetic: irritation with no effect on the outcome. Collect and review later.

What does a usability test cost?

If you do it yourselves, the biggest cost is your time: typically two to three days spread across planning, five sessions and analysis, plus a small thank-you for the participants. With help from a freelance UX designer at typically DKK 600–750 an hour, a test with five participants, recruiting and a short report typically takes 15–30 hours, roughly DKK 9,000–23,000. With agencies at DKK 1,100–1,300 an hour the same scope is typically DKK 17,000–39,000. Testing a prototype before any code is written is the cheapest version, because fixes cost minutes in Figma. Read more about wireframes, mockups and prototypes.

A worked example: a plumbing company with online booking gets many calls from customers who cannot find an available slot. A test with five customers shows that four out of five think grey dates are booked, when they are simply outside the selected week. The fix is a change of colour and one line of copy. If it saves ten calls a week at five minutes each, that is over 40 hours a year for the office, and more customers finish the booking themselves. If you have an existing product with many problems like this, a UX audit can be a good place to start.

Which mistakes should you avoid in a usability test?

Helping the participant when they get stuck. The finding disappears.
Leading tasks that use the same words as the buttons.
Asking whether people like the design instead of watching what they do.
Testing with colleagues, friends or family.
Testing too late, when the code is written and the budget spent.
Writing a long report nobody reads. Three findings and three video clips work better.
Testing once and never again. Test after every major change.

Usability testing is a fixed part of how we design, and you can read more about our approach under user research. To see how testing fits into a UX designer’s overall work, read what a UX designer does. And if you have a product or prototype that needs testing, you can get a fixed price within 24 hours.

Questions about usability testing

What is the difference between a usability test and a user interview?

A user interview is about understanding needs, habits and problems. You talk with the user about how they work and what they are trying to achieve, often before any design exists. A usability test is about seeing whether a concrete design works. The user gets tasks and completes them in the product or prototype while you observe. The interview tells you what to build, and the test tells you whether you built it so people can use it. Most projects benefit from both.

Can we test with our own employees?

Only if the employees are the real users, for example in an internal system. For a customer-facing product your staff know too much: they know the internal names, where features live and what the product is supposed to do. That hides exactly the problems a new customer will meet. Colleagues from other departments can be used for a quick pilot of the script itself, but the real tests should be with people who resemble your customers.

Should test participants be paid?

It is good practice to give something for their time, typically a gift card or a product, and it makes recruiting much easier. For B2B tests where participants are customers or professionals, the thanks can also be early access to new features or a discount. Make sure the reward is not so large that people sign up only for the gift. Also explain how recordings and personal data will be used and deleted, and obtain consent before the test.

What is an unmoderated usability test?

In an unmoderated test participants complete the tasks alone, typically through an online tool that records screen and voice. There is no facilitator, so you can collect many responses quickly and cheaply. The downside is that you cannot ask follow-up questions when something goes wrong, and the tasks must be extremely clear. Unmoderated tests suit measuring specific things in a finished product. Moderated tests, with a facilitator present, are best for understanding why something does not work.

Can a usability test replace A/B testing?

The two methods answer different questions. A five-person usability test shows why people get stuck and which problems exist. An A/B test with thousands of visitors shows which of two versions statistically drives more purchases or enquiries. Use usability testing to find the problems and shape improvements, and use A/B testing to measure the effect once you have enough traffic. Small sites rarely have the traffic for A/B tests, and there usability testing is the most effective method.

Can a usability test check accessibility?

Partly. A test with users who rely on a screen reader, magnification or keyboard only reveals problems that automated tools miss. But accessibility also needs a systematic review against WCAG 2.1 AA, covering contrast, alt text, focus order and error messages. The best combination is a technical review first and then a test with a few users with different needs. If you are covered by the accessibility act, document both and check the requirements with Sikkerhedsstyrelsen.

Want us to build it for you?

You get a fixed-price proposal within 24 hours.

Get a fixed price for a usability test