The practical answer
Choose a small set of important tasks, invite people who resemble the intended audience and observe them attempting those tasks without coaching. Record where they hesitate, fail or misunderstand, then fix the most consequential issues and test again.
Your team knows where everything is on the website. Customers do not. A page that feels obvious in a staff meeting may leave a first-time visitor searching for basic information.
Testing a website with real users helps you see that gap before paying for a redesign based mainly on personal preference. You are looking for evidence about tasks, language and decisions.
This is useful work within UI/UX and product design. The goal is to understand what prevents people from doing something important, then decide which improvements deserve attention.
Choose the decision the test should inform
Start with a question such as “Can a suitable customer find our service area and request a quotation?” A test with a clear question is easier to plan and interpret than a general request for opinions.
Choose a few priority tasks rather than asking participants to inspect every page. A service business might test finding a relevant service, understanding the process and starting an enquiry. A school might test locating admission requirements and identifying the correct application route.
Keep the distinction between usability and market demand. A successful task test can show that someone can use a journey. It does not prove that enough people want the offer or will become customers.
Invite people who resemble the intended audience
Colleagues can help identify obvious errors, but they often know the organisation’s language and structure already. Include people with the relevant goals, experience and device habits.
Think about important differences between groups. A parent researching a school and an administrator managing admissions have different tasks. A person using assistive technology may reveal barriers that other participants do not encounter.
A small qualitative round can identify useful problems. Nielsen Norman Group distinguishes this kind of work from studies designed to support statistical claims. Do not present a few observed sessions as a representative survey of every customer.
Write tasks without giving away the answer
Use a realistic scenario. “You need a website for a new training centre. Find out what information to provide when requesting a quotation” is more useful than “Click Services, then click Contact”.
The second instruction tests whether the person follows your directions. The first lets you observe how they understand the website.
Avoid leading language copied directly from a navigation label unless a real customer would naturally use it. You want to learn whether the labels make sense, not coach the participant into repeating them.
Prepare the session responsibly
Explain the purpose and how notes or recordings will be used. Obtain the necessary permission and let people decline recording. Make clear that you are testing the website, not assessing their intelligence.
Use test data and an agreed environment for forms, accounts and transactions. Prevent the session from creating a real order, payment or application unintentionally. Tell the operating team how to recognise any test enquiry.
Check the equipment and the time needed. A remote session should have a workable way to observe the task without asking the participant to expose unrelated personal information.
Observe before explaining
Ask the participant to attempt the task and describe what they are thinking where comfortable. Notice what they read, where they hesitate and what they expect a button to do.
Do not correct every misunderstanding immediately. If you rescue the participant at each difficult point, the final result will hide the difficulty. Record when help was required.
Keep observations separate from your interpretation. “The participant opened three pages before finding the service area” is an observation. “The menu is badly designed” is a conclusion that needs examination.
Record a useful result for each task
| Record | Example of what to note |
|---|---|
| Task outcome | Completed independently, completed with help or not completed |
| Difficulty | Missing information, unclear label, form error or navigation problem |
| Evidence | What the participant did or said at the relevant point |
| Consequence | Delay, abandonment, wrong action or minor inconvenience |
| Next question | What the team needs to investigate or test |
Use timing cautiously. One person’s slow attempt can reveal a difficulty, but it is not automatically a benchmark for all customers. The context matters.
Prioritise by consequence and recurrence
A broken enquiry form deserves attention before a minor visual preference. A label that repeatedly leads people into the wrong process is more consequential than an isolated suggestion to move a decorative image.
Consider how often the issue appeared, which users it affects and how serious the result could be. Also investigate a severe issue observed once; frequency alone should not decide everything.
Assign an owner and describe the intended improvement. Keep the evidence with the issue so the designer or developer understands the problem rather than receiving only a proposed solution.
Test the revised journey
After changes, check whether people can complete the task more clearly. Include relevant new participants where practical so memory of the earlier session does not become the explanation for success.
Keep a short record of what changed and what remains uncertain. User testing should inform the design process repeatedly, not become a one-time certificate that the website is “user-friendly”.
Frequently asked questions
Can I run a useful test without a research laboratory?
Yes. A carefully planned session with suitable participants and realistic tasks can reveal valuable problems. The method and interpretation matter more than a formal-looking room.
Should I ask users what redesign they want?
You can listen to suggestions, but focus on their needs and observed difficulties. Participants do not need to design the solution for the test to be useful.
Does finding problems mean we need a full redesign?
No. The evidence may support targeted content, navigation or form improvements. Use the findings to decide the scale of change.
Further reading
Reference addresses are provided as plain text for copying.
https://www.nngroup.com/articles/how-many-test-users/https://www.gov.uk/service-manual/user-research/plan-user-research-for-your-service
Discuss your next step
Considering a redesign? Share the tasks your visitors need to complete and the difficulties you suspect. I can help plan a focused review before the business commits to wider design changes.





