Moderated prototype testing: how to observe without leading
Plan a moderated usability test with neutral tasks, record what participants do and separate observations from assumptions.
Article language: English

In this article
Imagine a participant looking at a shop prototype. She is searching for collection information when, after three seconds of silence, the designer says, “Please click that little label under the price.” She clicks, the task is completed and everyone breathes a sigh of relief. The team has just learned that a person can click where they are told. It knows little more about whether she can find the information herself.
A moderated usability test involves watching someone attempt a task with a facilitator present. The role calls for attention and a little restraint. You need to make room for a conversation, understand moments of uncertainty and still avoid doing the task for the participant. It is a particular kind of hospitality: look after the person without pointing out every door.
You can examine a problem in a prototype without an elaborate laboratory. You do need a question, a suitable task, the right person and a way to record what happens. Without them, the recording may contain plenty of cursor movement and little to support a decision.
Decide what you want to learn
“Let’s check the website” can open a conversation with no end. “Let’s see whether a buyer can work out if a specific variant is available for collection at their chosen location” sets a boundary. You can tell which information the prototype needs and what would count as completing the task.
Separate the team’s research question from the instruction to the participant. The team might be investigating whether stock information is visible. The participant needs a buying situation in which that information matters. “Find the Availability tab” supplies a name and suggests a route. It does not test the same thing as trying to arrange a collection.
Here is a sample task: “You want to collect a green organiser on Friday from the pickup point on Example Street. Find out whether that is possible and what you need to do next.” The details are fictional, but the situation has a recognisable goal. It does not name a button or direct the participant through a sequence of screens.
GOV.UK’s guide to moderated testing recommends tasks with a clear, credible goal that do not give away the answer or suggest how to complete them. It is a useful principle for preparation. The situation still needs to fit the actual audience for the service.
Set out the conditions within the situation too: is Friday a requirement or a preference, and may the participant choose another pickup point? Without those details, a facilitator may later mark a reasonable solution as an error because they had a particular route in mind. Give the participant the constraints needed to decide, while leaving them free to choose how to proceed. A good task leaves room for a solution the team did not anticipate. It should not require anyone to guess a missing part of the scenario.
Agree on the expected result before the session. In this example, the participant should identify the variant, pickup point, availability condition and next step. Reaching the right screen alone does not prove understanding. Someone can find the information, read it as meaning the opposite and confidently travel to collect a product that is not there.
Prepare a prototype state that allows you to check that result. If the colour or pickup point cannot be changed, the test has to be narrower. Participants should not have to guess whether they have found a design problem or a limitation of the prototyping tool. Explain what kind of material they are seeing at the start, without revealing the route under investigation.
Choose participants according to the question. Someone who has used the company catalogue for years may understand abbreviations that mean nothing to a new buyer. Recruitment for UX research should therefore come before filling calendar slots. A friend being free on Wednesday is not, by itself, a criterion for a good match. For more detail, see the guide to recruiting UX research participants (in Polish).
Agree on how to respond when a prototype state is missing. If a participant chooses a route the mock-up does not support, record the choice and explain the limitation. Do not attribute the problem to the participant. You can then open an agreed screen to examine their understanding of the information that follows, but mark that portion as having begun after help. An interesting alternative route should not disappear from the evidence merely because the person who made the prototype did not anticipate it.
Before the first proper session, walk through the task with someone helping you check the scenario. This is a rehearsal of the tools and instructions. You may find a broken link, an unclear fictional name or a missing state. Do not present that rehearsal as a finding about customers.
A short fictional scene from a session
The facilitator introduces the situation. The participant opens a product, chooses a colour and pauses at the word “Available”. She says, “So I suppose I can collect it on Friday.” She has not changed the pickup point. The team wants to ask about the small link beneath the message straight away.
Instead, the facilitator might ask, “What do you understand ‘available’ to mean here?” That invites an explanation of what she has read without pointing to the missing element. She replies, “That the green one is in the shop I’ve selected.” In the prototype, however, the message refers to the dispatch warehouse. There is now a specific mismatch to record.
The participant then tries to continue. The facilitator lets her proceed, watching to see whether a later step clears up the misunderstanding. If the prototype blocks the route, the facilitator can end that part and move on to the next task. There is no need to pretend the mock-up can complete an entire purchase.
When help is needed, record it precisely. “The participant was shown how to change the pickup point after independently interpreting the message” describes what happened. Behaviour after that prompt may still be interesting, but it is no longer evidence that she found the pickup point on her own. That distinction does not invalidate the session. It lets you use its different parts honestly.

GOV.UK research practitioners emphasise neutral instructions and reactions, without signalling that a participant is following a good or surprising route. People may react to tone of voice and facial expressions too. Watch the reflexive “great”: it sounds friendly but can act like a green arrow.
You do not have to turn yourself into a motionless piece of furniture. You can acknowledge that you are listening, ask someone to elaborate and respond to discomfort. The point is to separate support for the person from hints about the interface. If a participant wants to stop, stop. No question about a button requires sacrificing someone’s comfort for the sake of a complete table.
Silence needs interpretation too. Someone may be reading, thinking or struggling to understand the task. Rather than assuming they are stuck after a few seconds, watch what they do. If you ask a question, start neutrally: “What are you considering now?” Do not build your own diagnosis into it: “Can’t you see the pickup information?”
Leave evaluative questions until the task has been completed or stopped. Then you can ask what was difficult and how the participant understood the result. During the task, “Is this convenient?” pulls attention away from the attempt and invites a polite comment about the design.
Record behaviour, interpretation and your own idea separately
Three kinds of note help in a session record. The first describes an observation: the person read “Available” and did not change the pickup point. The second preserves what she said: she believed the message referred to the local shop. The third is the team’s hypothesis: the label does not explain the scope of availability. Those statements do not have the same status.
“Let’s add a bigger button” is another step again. It may be a good idea, but it does not follow automatically from the observation. The problem could be the wording, the order of choices or a mistaken assumption about the selected location. Before choosing a fix, compare several moments and establish what information needs to be conveyed.
A useful record includes the prototype version, task, observed state, participant’s words and any intervention by the facilitator. If you make a recording, agree its purpose, access and storage arrangements before the session. A project report usually needs only enough of an excerpt to reconstruct the problem, without unnecessary participant data.
For a remote session, check the technical setup as well: opening the prototype, sharing the correct window and being able to read the text. If a participant uses a phone, do not judge their behaviour from a feed that cuts off part of the screen. A pause to set up the tool is not a difficulty with buying. Record it separately so setup time is not counted as task time. If the tool gets in the way of how that person normally operates, agree suitable conditions before interpreting the results.
Observers need to know their role. Ask them to save questions for the discussion afterwards. When three people start explaining the layout at once, the session becomes a product presentation. The participant is not obliged to guess which question to answer first or defend their way of using the site before a panel.
After the session, take a moment to compare notes together. Mark which conclusions come from what the participant said and which remain assumptions. If two people interpret the same moment differently, return to the record. An observer’s confidence is no substitute for what the participant actually did and said.
Keep the cases where nothing went wrong too. If a participant correctly distinguished the warehouse from the pickup point, record which information helped them. Comparing that route with a misunderstanding may reveal a useful cue already present in the design. Do not transfer it everywhere automatically: it may have worked because of that person’s earlier experience. Show both a successful route and a barrier in the report, so a fix does not remove something that helped some participants.
Connect findings to design decisions. “The participant confused warehouse availability with the pickup point” can lead to a clearer message and another test after the change. “People don’t read” leads almost nowhere, apart from another meeting where the team repeats the sentence together.
Record the findings as a hypothesis card and next test. A possible hypothesis is that naming the warehouse clearly next to the message will reduce misinterpretation. Only another test can show whether the selected change actually resolves the observed problem. This is discussed in the article on hypothesis cards and business experiments (in Polish).
The Aeneia Design offer is a starting point for discussing a website project and how to check it. Bring a specific task and a prototype that shows the states needed for the proposed test.
Say which questions the session cannot answer
A few observations can uncover obstacles and help prepare improvements, but they do not automatically provide a percentage for the entire customer group. Describe who took part, which tasks they attempted and which conditions were artificial. A reader of the report should understand the limits of the evidence.
A short five-second test addresses questions about initial orientation. A moderated task lets you examine later decisions and obstacles. There is no need to choose one method as a universal winner. The choice follows the uncertainty you want to reduce. If you want to test the task without a facilitator, prepare a separate unmoderated study with a clear ending and a way to check for faults. The task must then make sense without explanations during the session. For related guidance, see the guide to the five-second test (in Polish). You can also consult the article on unmoderated research (in Polish).
If you are interested in a buying journey spread over several days, a single session will show only part of it. A diary study can collect accounts at different moments. By contrast, checking whether a form works on the published site calls for a separate acceptance check of the implementation (in Polish). A good prototype does not prove that a message reaches an inbox. The topic is explored in the guide to diary studies (in Polish).
Finally, choose a small number of improvements supported by the evidence and prepare a new version of the test. Keep the earlier record so the reason for changing a message or route remains visible. Research has done its job when the team can make a better-supported decision and the visitor has an easier route. Not when the facilitator gets ever faster at pointing to the right button.








