Many projects skip usability testing because they picture it as expensive: labs with one-way mirrors, huge samples, weeks of analysis. In reality a good qualitative test needs a few people, a quiet room or a video call and an afternoon of work. What really matters is preparation.
Why a few people are enough
The idea, popularized by Nielsen, is that the most serious usability problems surface early: after the first participants you start seeing the same obstacles repeat. For a qualitative test, whose goal is to find problems rather than measure percentages, five or six people with a similar profile are often enough for one round. Three small tests a few weeks apart, with fixes in between, are more useful than a single large test at the end.
This applies to qualitative tests. If the goal is to compare two versions with reliable numbers, or to estimate a success rate, you need much larger samples and a statistical approach.
Define the goal
Before looking for participants, write down in one sentence what you want to learn. Examples:
- «Can people find and change their saved shipping address?»
- «Is the new sign-up process clearer than the old one?»
- «Is the difference between the three subscription plans understood?»
A precise goal drives the choice of tasks and stops the session from turning into a generic chat.
Write the tasks
Tasks are the heart of the test. They must be realistic, have a clear end point and not hint at the solution.
| Badly written task | Well written task |
|---|---|
| «Click Settings and change the address» | «You've moved house: make sure your next order arrives at the new home» |
| «Use the price filter» | «Look for a pair of running shoes under 90 euros» |
The first kind of task contains the interface's own words and turns the test into a word hunt. The second describes a situation and leaves the path to the user. Prepare three to six tasks, ordered from simplest to most complex, and try the sequence on a colleague before the real sessions.
Recruit participants
Participants should resemble the people who will actually use the product. Management software for dental practices should be tested with practice reception staff, not with the team's developers. Some practical tips:
- prepare a few screening questions to rule out people outside the target;
- avoid colleagues, close friends and anyone who already knows the project;
- offer a small fee or voucher: respect participants' time;
- recruit one more person than you need, because someone always cancels at the last minute.
Run the session
A session usually lasts between forty-five and sixty minutes. A proven structure is:
- Welcome (5 minutes): explain that you are testing the product, not the person, and that there are no wrong answers.
- Opening questions (5 minutes): habits and context, for example how they do the same activity today.
- Tasks (30-40 minutes): ask them to think aloud while working.
- Wrap-up (5-10 minutes): overall impressions and open questions.
The moderator must resist the urge to help. When a participant gets stuck and asks «Should I click here?», the best answer is to turn the question around: «What would you expect to happen?». Only if they are completely blocked and cannot continue should you give a hint, noting that the task was not completed independently.
If possible, a second person takes notes while the moderator focuses on the conversation. Recording screen and voice, with the participant's consent, lets you review the critical moments.
What to observe
What participants say is useful, but what they do is more so. People tend to be polite and say that «all in all it was easy» even after struggling for two minutes. Watch for:
- where they hesitate or go back;
- what they look for without finding it;
- where they click expecting something that does not happen;
- signs of surprise or frustration;
- whether they complete the task, and roughly how long it takes.
Analyze the results
Right after each session, note the main problems while your memory is fresh. At the end, gather all observations, group them by theme and count how many participants hit each problem. A problem seen by four people out of five deserves priority; one seen by just one person needs careful judgment, because it may be an isolated case or a weak signal of something important.
The final report should be short: the five to ten most relevant problems, with a brief description, a video clip or quote, and a proposed fix. Nobody reads a forty-page document; an ordered list with concrete examples, on the other hand, pushes the team to act.
Remote testing
Video calls with screen sharing work very well and widen the pool of participants. It is worth doing a technical check a few minutes before, having a plan B if sharing fails, and remembering that seeing the person's real environment, with their own device and habits, is often an advantage.
In short
- For a qualitative test, five or six participants per round are enough; several short rounds work better.
- Tasks should describe real situations without using the interface's own words.
- During the session you observe, you don't help: behavior counts more than opinions.
- Analysis ends with a short, ordered list of problems and proposed fixes.
