Short answer: Usability testing means watching real people try tasks on your website, one person at a time, while they say out loud what they think. Five people per round find most of the common problems. Fix them and test again.
What usability testing is
The U.S. Department of Homeland Security (DHS) describes usability testing as a way to learn how easy or difficult it is for people to use something, by watching them use it.
In a moderated test, as GOV.UK calls it, you watch participants try to complete specific tasks. Asking them to think aloud helps you understand what they are doing, thinking and feeling.
The W3C notes that a test can follow established protocols or be very simple and informal.
Why watch people instead of guessing
As Nielsen Norman Group (NN/g) says, “zero users give zero insights.”
The W3C adds that informal tests throughout a project are more effective than only one formal test at the end.
In our view, analytics and tests answer different questions. Tools show where people click, and a test lets you hear why they got stuck. Our guide to website heatmaps covers what clicks, scrolls and recordings show. To find out which of two versions of a page gets more visitors to act, our guide to A/B testing shows how to test a page on your website.
How many people to test
For a typical test of one group of users, NN/g recommends five participants to uncover most of the common problems.
In its research, the first test with five people found 85% of the usability problems. After you fix them, a second round with five more finds most of the rest.
NN/g recommends three rounds of five people over one test with 15, and as many small tests as you can afford.
Test one person at a time. According to Digital.gov, a useful test has one participant per session and enough participants to see patterns.
Who to test with
GOV.UK says participants need to be actual or likely users of your service. Digital.gov warns against testing with experts, friends, family or close colleagues of the team.
The DHS kit uses colleagues only for quick feedback in the early stages, as a fast second opinion. When a draft is close to complete, it tests with real users.
Include people with disabilities. The W3C notes that testing with them usually reveals general usability problems that affect all users.
People who use assistive technology often have set-ups that are hard to copy, so GOV.UK says it is usually best if they use their own devices.
A few test sessions don’t prove a site is accessible. The W3C says to combine user testing with a check against WCAG, and our guide to website accessibility checkers covers what free tools miss.
How to write test tasks
Good test tasks, according to GOV.UK, set a clear goal and feel relevant and believable to the participant. They are hard enough to show problems and don’t give away the answer.
NN/g adds that tasks should be realistic and ask for an action. If you tell someone to click a menu label, you won’t learn whether that label makes sense to them.
In the Digital.gov examples, “How would you view the status of your application?” leaves the path to the participant. “Which of the links in the header would you click to login?” guides them.
NN/g also warns that small errors in how you word a task can change how people do it. It often asks participants to read each task out loud, so they read all of it and the notes show which task they are on.
For a small business site, we suggest tasks such as “You want to book an appointment for next Saturday. Show me how you would do that.” Our guides to website navigation and to calls to action cover the menus and buttons those tasks put to work.
How to run a session
GOV.UK says sessions usually take between 30 and 60 minutes, with at least 15 minutes between them. Participants are often nervous, so tell them you are testing the site, not them.
The sample script from Digital.gov runs through an introduction, a warm-up, the set-up, the tasks, a follow-up and a wrap-up. It tells participants there are no wrong answers and asks them to think aloud. It also asks them not to worry about hurting anyone’s feelings, because the team needs their honest reactions.
It starts with first impressions: “What are your first impressions?” and “What is this site for?” Our guide to the website homepage covers what people see first.
During a task, GOV.UK advises you to stay quiet and mostly watch and listen. Step in to understand something interesting, or if the participant gets completely stuck.
When a participant gives an opinion, ask open-ended questions, such as “What makes you say that?”
The Digital.gov script ends with this question: “If you had a magic wand and could instantly change one thing about this website, what would it be?”
Anyone else watching stays quiet. Digital.gov says observers don’t moderate and keep their open-ended questions for after the test.
In person or remote
NN/g says remote tests are popular because they often take less time and money than in-person studies.
In a moderated remote test, you and the participant are online at the same time. In an unmoderated one, the participant works alone, sometimes with follow-up questions set in advance.
NN/g still recommends in-person testing whenever possible, because body language is easier to read in the room.
The DHS kit adds that a video call is especially useful for websites and forms, because the screen of the participant can be shared and recorded.
In our view, a video call with screen sharing is the easiest way for a small business to start.
Permission and recordings
Digital.gov publishes sample participant agreements, with and without payment. They say taking part is voluntary, and that people can skip a question or stop at any time.
The session is recorded only if the participant approves, and they can ask you to stop recording at any time. The agreements also promise to remove details that could identify the person, such as their face or name.
GOV.UK adds that any personal data you collect, on paper or in recordings, has to be stored securely.
Turn what you saw into fixes
Following Digital.gov, debrief with your team right after each session, and look for patterns across participants.
Report the problems in order of priority, not as an unsorted list, and turn them into changes to the site.
Then test again. NN/g advises running many small tests rather than one big one.
If the problems run through the whole site, our guide to website redesign covers when to do one and how to plan it.
How to run your first usability test
- Write down what you want to learn, and turn it into a few short tasks with a clear goal and no hints.
- Recruit five people who use or could use your site. Leave out your staff, friends and family.
- Book one person per session, for 30 to 60 minutes with a break of 15 minutes or more between sessions, and ask for permission to record.
- Ask them to think aloud as they work. Watch, listen and stay quiet unless they get completely stuck.
- Debrief after each session, fix the problems several people hit, and test again.
These steps follow GOV.UK, Digital.gov and NN/g.
Before you run your test
- Each task has a clear goal and no hint of where to click.
- Your testers use or could use your site, and none of them works with you.
- Each session has one participant, and anyone watching stays quiet.
- You have permission to record, and the participant knows they can stop.
- Someone takes notes, or the session is recorded.
- The next round is planned to check your fixes.
RememberWatch real people try real tasks on your site, one at a time, without help. Five people per round find most of the common problems. Fix them and test again.
Our web design service builds your site in WordPress with Bricks and checks its speed in Google PageSpeed Insights and on real phones before launch. Everything lands in your name.
Questions about usability testing
What is usability testing?
Usability testing means watching people try real tasks on your website to see how easy it is to use. The DHS describes it as observing people while they use something. Participants usually think aloud, so you hear what they expect at each step.
How many users do you need for a usability test?
For a typical test of one group of users, NN/g recommends five participants, enough to uncover most of the common problems. It suggests several small rounds instead of one big test: fix what the first five found, then test with five more.
What questions should you ask in a usability test?
Start with first impressions, such as what the participant thinks the site is for. Then give tasks without hints, such as booking an appointment, and ask them to show you how they would do it. Follow up with open-ended questions, like “What makes you say that?”, and end by asking what one thing they would change.
Can you do usability testing remotely?
Yes. In a moderated remote test, you meet the participant online at the same time and watch them work. In an unmoderated one, they complete the tasks alone. NN/g still recommends in-person tests whenever possible, because body language is easier to read.
Can I test my website with friends or employees?
For a quick second opinion early on, the DHS kit tests with colleagues. For results you can act on, test with people who use or could use your site. Digital.gov warns against testing with experts, friends, family or close colleagues of the team.
Sources
Read on November 7, 2026. From Digital.gov (GSA): How to conduct a usability test. How to develop user testing participant agreements. How to write a research plan. From the U.S. Department of Homeland Security: DHS Usability Testing Kit. From GOV.UK: Using moderated usability testing. From the W3C Web Accessibility Initiative: Involving Users in Evaluating Web Accessibility. Involving Users in Web Projects for Better, Easier Accessibility. From Nielsen Norman Group: Usability (User) Testing 101. Why You Only Need to Test with 5 Users. Thinking Aloud: The #1 Usability Tool. Remote Usability Tests: Moderated and Unmoderated. Task Scenarios for Usability Testing. Also from our site: the web design service page and our guides to the website homepage, navigation, calls to action, heatmaps, website redesign and accessibility checkers.