Website usability testing means watching people from the intended audience try realistic tasks on a website, prototype, or service. Choose moderated sessions when you need to ask follow-up questions and understand unexpected behavior; choose unmoderated sessions when tasks are clear enough to complete alone and you need consistent results across participants. Start with a decision the study will inform, then use neutral tasks, observe what people do, and turn evidence into design changes.
What website usability testing can—and cannot—tell you
Usability testing evaluates how people complete tasks with a website, prototype, or service. It can reveal where participants hesitate, make errors, misunderstand content, or fail to reach a goal. It is different from functional quality assurance, which checks whether software behaves as specified, and from expert inspection, where specialists assess an interface without observing representative users.
ISO 9241-11:2018 offers a framework for understanding usability, but it does not prescribe a particular evaluation method. The method should follow the question you need answered: for example, whether people can find a return policy, understand a form, or complete a checkout task. ISO 9241-11:2018
A qualitative formative study helps you discover and understand problems. A quantitative study or benchmark measures defined outcomes—such as completion, time, or errors—across a larger sample. A small qualitative study can expose useful issues, but it should not be used to claim precise population-wide rates.
#1 Best Overall
Choose a testing method that fits the question
Moderated or unmoderated
| Method | Best suited to | Main trade-off |
|---|---|---|
| Moderated | Exploratory work, complex tasks, and diagnosing why someone gets stuck. A researcher can clarify context and ask tailored follow-ups. | Requires a live facilitator and scheduling each session. |
| Unmoderated | Narrow, clearly worded tasks that participants can complete alone; useful when you want consistent instructions across more sessions. | No live intervention or opportunity to clarify what happened. Ambiguous instructions can undermine the session. |
In a moderated remote test, participant and researcher interact live. In an unmoderated remote test, software presents instructions and records the participant without a researcher present. Neither mode is automatically better; match it to your need for probing, the task complexity, and logistics. Nielsen Norman Group: remote usability testing
Remote or in person
Remote sessions make it possible to include participants in different locations. In-person testing can make direct observation and physical context easier when those matter to the question. Choose based on the setting, access needs, and whether the researcher must see or understand something that is difficult to convey remotely.
Qualitative or quantitative
Use qualitative work to find and understand friction; use a suitably designed quantitative study to measure defined outcomes. Before collecting numbers, specify what counts as success, partial success, an error, or a request for help. A completion rate without a clear definition can be misleading.
How to conduct a website usability test
- Decide what the study will inform. Write one or more research questions, identify the target user group, and select the pages, journey, or prototype in scope. Keep the scope narrow enough to observe closely.
- Recruit relevant participants. Screen for characteristics connected to the service and its intended users. Include people with disabilities when accessibility or inclusive usability is part of the question, and prepare appropriate materials, technology, and facilitation.
- Select the format. Decide whether you need a moderator, whether the session should be remote or in person, and what access and logistics require. Use unmoderated testing only when participants can understand and complete the instructions without live help.
- Write realistic, neutral tasks. Describe an outcome that matters to the participant, not the interface action you expect. For a store, “Find out whether you can return this item if it does not fit” is more neutral than “Click Returns in the footer.” Tasks should be believable and challenging enough to expose friction without giving away the route.
- Prepare a guide and pilot it. Include an introduction, consent and recording explanation, task wording, neutral prompts, and an observation checklist. Try the materials and technology before sessions; check unmoderated instructions especially carefully because no live moderator can repair unclear wording.
- Run sessions consistently. Explain that you are evaluating the service, not the participant. Invite people to think aloud, use neutral prompts, and avoid steering them toward success. GOV.UK says moderated sessions commonly take 30 to 60 minutes, depending on task count and complexity. GOV.UK Service Manual: user research
- Record observable evidence. Note whether each task was completed, partial, or unsuccessful; errors, hesitation, detours, misunderstood content, participant comments, and relevant context. Collect only measures that help answer the research question.
- Analyze and act. Group observations by task and user impact. Separate what you saw or heard from your interpretation of why it happened. Prioritize changes, make them, and run another study when a new design decision warrants it.
Use think-aloud without leading participants
Think-aloud asks people to describe what they are doing or thinking as they work. It can expose a mismatch between what someone expects and what the interface communicates. GOV.UK puts the purpose simply: “Asking them to ‘think aloud’ as they move through the service helps you understand what they are doing, thinking and feeling.” GOV.UK Service Manual
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Rank #3
In moderated sessions, invite comments with neutral prompts such as “What are you looking for?” or “What would you expect to happen?” Avoid questions that suggest the answer or evaluate the participant. In an unmoderated session, participants may stop verbalizing and cannot be prompted, so recordings may provide less explanation. In either mode, treat observed actions and task outcomes as evidence alongside verbal comments—not as a substitute for them.
How many participants do you need?
For a typical qualitative usability study focused on one user group, Nielsen Norman Group recommends five participants as a practical starting point for uncovering common problems. It is not a universal statistical sample size, nor a guarantee that a fixed share of issues will be found. Nielsen Norman Group: why you only need to test with five users
Rank #4
Use a different design when the study spans distinct user groups, involves high-risk tasks, compares alternatives statistically, or needs precise performance estimates. Quantitative benchmarking generally requires a larger sample and clearly defined performance measures. Digital.gov’s plain-language guidance recommends testing a website or document with three to five people; that is practical guidance, not a statistical guarantee. Digital.gov: test your content
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Measures and accessibility considerations
Choose measures before the session
Potential measures include task success, partial success, critical errors, time, navigation path, and post-task satisfaction. Define each in advance. For example, decide whether a participant who needs a hint has completed the task or only partially succeeded. Pair measures with behavior and participant explanations: metrics show what happened, while observation can help explain why.
Recommended Free Tools
Best Value
Make accessibility part of the study design
A generic usability protocol can miss accessibility barriers. W3C WAI advises involving users with disabilities and tailoring participants, methods, evaluation parameters, and assistive technology to the accessibility question. Ensure the venue, prototype, materials, and facilitation suit the participants and the barrier being investigated. Focus data collection on those barriers rather than assuming time-on-task or satisfaction alone captures them. W3C WAI: involving users in evaluating web accessibility
Capture screenshots to document interface states
When a usability finding depends on a particular page state, a screenshot can help the team document what the participant saw. For a local do-it-yourself capture, open the relevant page in a browser, reproduce the state, and use the browser’s screenshot or print-to-PDF feature. For repeatable captures in a study pipeline, a browser automation script can navigate to the URL and save a screenshot; it may need additional setup for authentication, dynamic content, consent overlays, or the exact viewport. A screenshot records appearance, not participant behavior, and should not replace consent, observation notes, or recordings.
Or skip the browser setup
ScreenshotNeo is a website screenshot API and MCP server. One GET request can return a PNG, JPEG, WebP, or PDF. Its clean-shot steps accept cookie or consent banners and remove more than 60 known consent platforms, newsletter popups, and chat widgets; each step can be turned off. Bot checks and CAPTCHAs, blank pages, timeouts, failed loads, and cache hits cost nothing, and response headers report the page verdict and whether the request was billed. An MCP server provides screenshot tools for AI agents using Claude, Cursor, or other MCP clients.
For a WebP screenshot of a target page, replace the example URL and supply your API key:
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpSee the ScreenshotNeo API documentation for request options. The service offers 1,000 screenshots a month free with no card; paid plans start at $5 for 3,000. Sign up for free screenshots.
Quick Recap
Common usability testing mistakes
- Giving away the route in the task. Name the participant’s goal, not a menu label, button, or exact sequence of clicks.
- Rescuing participants too quickly. In moderated sessions, allow a reasonable pause and use neutral prompts; help only according to a consistent protocol.
- Treating opinions as proof of behavior. Pair what participants say with task outcomes, observed actions, and context.
- Overgeneralizing from a small qualitative study. Report themes and observed problems, not precise population rates.
- Ignoring accessibility needs. Match participants, assistive technology, and study setup to the accessibility question.
- Using unclear unmoderated instructions. Pilot the full task flow as a participant would encounter it, because no moderator will be present to clarify it.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
