You can learn most of what is wrong with a website by watching about five people try to use it. Give each person realistic tasks, stay quiet, note where they hesitate or fail, fix the biggest problems, then test again. It costs hours, not a research budget.
Why five is enough (and where it is not)
The Nielsen Norman Group’s well-known argument is that small tests run in sequence find more problems than one large study. Their model assumes a single user reveals about 31 per cent of usability problems on average, that returns diminish after that, and that the first five-user study finds roughly 85 per cent of problems. They recommend three rounds of five over one round of fifteen, with fixes between rounds (Nielsen Norman Group).
Two caveats from the same source matter:
- The 31 per cent figure is an average across the author’s projects, not a constant.
- If you have clearly different audiences, test more people: at least three per group.
- Five users is for finding problems. Measuring things like completion rates needs a larger quantitative study, which the article suggests is closer to 20 users.
Step 1: Pick three to five tasks
Choose the actions that matter most to your business: finding a service, checking prices or availability, contacting you, buying, booking. Limit each session to about 30 to 45 minutes so people stay fresh.
Step 2: Write realistic task scenarios
Poorly written tasks ruin tests. The Nielsen Norman Group’s guidance on task scenarios says tasks should be realistic, actionable and free of clues (Nielsen Norman Group):
| Weak task | Better task | Why |
|---|---|---|
| “Click Services, then find the web design page” | “You need a new website for your business. Find out whether we can help.” | Does not name menu labels |
| “Where would you click to contact us?” | “You have a question about timing. Get in touch.” | Asks for action, not explanation |
| “Find the page about pricing” | “Work out roughly what this would involve before you get in touch.” | Describes the outcome |
Add enough detail to make the task completable, such as a hypothetical date or budget. Participants generally accept hypothetical details.
Step 3: Recruit five people who resemble your customers
Friends and colleagues who know your business will not behave like first-time visitors. Better options are people from your customer list who agree to a short session, contacts in the right industry who have not seen the site, or a local community group. Offer a thank-you of your own choosing and tell people the site is being tested, not them. Where you record screens or voices, say so and get agreement first, and handle recordings with the care the Privacy Act expects of personal information.
Step 4: Run the session
- Explain you are testing the website, not the participant.
- Ask them to think aloud as they work.
- Give one task at a time, read from a printed card.
- Do not help, hint or defend the design. If they ask where to click, say “What would you do if I were not here?”
- Note where they hesitate, backtrack, misread a label or give up.
- Finish with two questions: what was confusing, and what would make them more likely to contact you.
A phone or laptop with screen sharing and a notes sheet is enough. Test on the device most customers use, which is often a phone.
Step 5: Decide what to fix first
After five sessions, list each problem and tally how many participants hit it. Then prioritise:
- Blocks a key task (nobody could find the contact route): fix first.
- Hit by three or more of five: fix next.
- Slows people but they recover: batch for the next release.
- Single-person preference: note it, do not act without more evidence.
Fix the cause, not only the symptom. If three people missed a button, the issue may be page hierarchy, not the button colour.
Step 6: Retest
The Nielsen Norman Group stresses retesting because fixes may not work and can introduce new problems. After the obvious issues are gone, later rounds can probe deeper questions such as how the site is organised. If you are reviewing structure, our guides on website navigation design and homepage essentials are good starting points, and what makes a business website convert gives context for the tasks you choose.
Limits
Five-person testing finds usability problems, not market demand. It will not tell you how many visitors share a problem or whether a design will lift enquiries. Combine it with analytics, and treat it as a way to find friction quickly and cheaply.
Where Radiance Impact fits
Radiance Impact’s Technology division builds and redesigns business websites and landing pages. See the Technology page for the range of work, or contact us if you would like help planning a test or acting on its results.
Sources
- Why You Only Need to Test with 5 Users, Nielsen Norman Group
- Turn User Goals into Task Scenarios for Usability Testing, Nielsen Norman Group
