Marcus Andrey Study
Ler em português Library

UX Research · Step 4 · 7 min read

Moderating and thinking aloud

Thinking aloud is the most useful usability method there is, and the easiest to spoil by talking. How to brief participants, keep them talking without steering them, answer their questions without answering them, ask follow-ups after a task, and know when you've biased a session.

Every time you help a participant, you erase a finding. It feels kind to point at the button they can't find. It feels rude to answer a question with a question. But a usability test exists to see what happens when you're not there, and you won't be there when real people use the product. Good moderating looks like doing very little. It's actually a lot of quiet decisions about when to speak, what to say, and what never to say.

What you'll be able to do after reading:

  • Explain what thinking aloud shows, and its three main risks
  • Brief participants so they think aloud from the first task
  • Keep them talking and answer their questions without leading
  • Ask follow-up questions after a task, and spot a session you've biased

1. What thinking aloud shows

In a think-aloud test, participants use the design while continuously saying what they think: what they're looking for, what they expect, what puzzles them. Nielsen called it "the single most valuable usability engineering method" in 1993 and still stood by it in 2012 (NN/g, Nielsen, Thinking Aloud: The #1 Usability Tool, 2012).

What makes it so valuable is that it shows why. You hear people's misconceptions, and misconceptions usually turn straight into redesigns: when people misread an element, you change it. Nielsen adds that the method is robust: unless you blatantly put words in people's mouths, even a poorly run study gives reasonably good findings.

Usability testing with five people introduced the method. This lesson is about the part that makes or breaks it: you.


2. Three risks, and what to do about each

Nielsen lists the downsides plainly (2012):

  • It's unnatural. Most people don't talk to themselves all day. Luckily, they quickly get absorbed in the tasks. Show a short example before you start (a few seconds of you thinking aloud on an unrelated site, or a short video) so they know what's expected.
  • Filtered statements. People want to seem smart, so they wait until they've worked it out and then give you a tidy commentary. You want the raw stream of thought. Prompt them to keep talking.
  • Biasing behaviour. Prompts are necessary, but an untrained facilitator's interruptions easily change what people do. When you know you've biased a moment, note it and discard that part. The worse case is not knowing.

One more limit: thinking aloud slows people down, so the times you measure in a think-aloud test aren't real task times. NN/g notes the extra time spent talking contaminates task-time measures (NN/g, Nielsen and Budiu, Success Rate, 2001). If you need to measure time, that's a different study (see Measuring usability).


3. The briefing

Two minutes before the first task do most of the work:

  1. "We're testing the design, not you. If something is hard, that's exactly what we need to find."
  2. "Please think aloud: say what you're looking at, what you're trying to do, what you expect to happen." Then show the short example.
  3. "I may not answer your questions. That's not me being rude; I want to see what you'd do if I weren't here. Ask anyway: questions are useful to us."
  4. "Stop whenever you want." And ask for consent to record (Research ethics and consent).

The third point is the one most people forget. Say it up front, and boomeranging a question later feels like part of the deal, not a snub.

A usability session also feels different from an interview. In an interview you're warm and make lots of eye contact to build rapport. In a test, NN/g describes a more rigid interaction: sporadic eye contact, and a facilitator who avoids being too friendly. The attention belongs on the screen (NN/g, User Interviews 101, 2023).


4. During a task

  1. POK, Ana was away… I'm looking for something like "edit". Hmm.Thinking aloud: what they look for, in their words
  2. MIt's under the three dots, top right.Helping. The finding just disappeared
  3. M(waits five seconds)Silence. Let them keep looking
  4. PIs "Split" the thing I want? Should I press it?A direct question to you
  5. MYes, try that one.Answering. Now it's your path, not theirs
  6. MWhat would you do if I weren't here?Boomerang. The question goes back to them
  7. PI'd press it, I guess. Oh, it split it in three. That's not what I meant. Weird.The real problem shows up
  8. MWeird?Echo. Their word, as a question
  9. PI thought "Split" would ask me how. It just did it. I'd want to choose first.The expectation behind the problem: a finding
Two minutes of a session (P is the participant, M the moderator), with the moderator’s moves named. The struck-out lines are what most of us say by instinct; each one would have hidden the finding at the end. Silence, boomerang and echo keep the participant in charge.Example transcript: Marcus Andrey, CC BY 4.0. Echo and boomerang from NN/g (Pernice, 2014).

When they go quiet, wait. Then a neutral prompt: "What are you thinking?" or "What are you looking for?" Never "Are you looking for the menu?": that names the answer.

When they ask you something, don't answer it. Kara Pernice at NN/g gives three safe moves (NN/g, Pernice, Talking with Participants During a Usability Test, 2014):

  • Echo: repeat their last word or phrase as a question. "Weird?"
  • Boomerang: send the question back. "What do you think?", "What would you normally do?"
  • Columbo: ask half a question and trail off, acting less like the expert. "So you were… [pause]".

And before you say anything, decide whether you need to: was it a real question, or just thinking aloud? Did you already understand enough? Will a probe really add something?

When they're truly stuck, and the session can't go on, help in the smallest step you can ("Is there anywhere else you might look?"), note that the task failed at that point, and move on. Getting someone unstuck isn't a crime; forgetting to record it is.

What not to do: explain the design, defend it, say "good" or "perfect" (praise tells them they're on the right path), or show where something is. If you designed it, this is the hardest part. Sit on your hands.


5. After a task

Once the task is done, or abandoned, there's room for a few follow-up questions, again as a funnel: broad first ("What did you think of doing that?"), then narrower ("You paused on this screen. What was going through your mind?") (NN/g, Rosala and Moran, The Funnel Technique, 2022). Point to the screen, not to the answer: "Tell me about this part" rather than "Did you see the filters?".

If you're collecting a rating after each task, ask it now, before the follow-ups, so your questions don't change their answer (Measuring usability).


6. Common mistakes

  • Helping, hinting or showing the way during a task.
  • Saying "good", "great" or "perfect" when they do something right.
  • Prompts that name the answer: "Are you looking for the search?"
  • Skipping the briefing, so participants think aloud only when reminded.
  • Measuring task time in a think-aloud session.
  • Not marking the moments you biased, so they end up in the findings.

7. Vocabulary

Term Meaning
Think-aloud Participants say what they're thinking while they use the design
Moderator / facilitator The person who briefs, gives tasks and decides when to speak
Prompt A neutral nudge to keep the participant talking
Echo, boomerang, Columbo Three ways to respond to a participant without leading
Assist Help given during a task; the task counts as failed or assisted
Bias When the moderator's words change what the participant does

8. Try it (no tools needed)

  1. Ask a friend to set an alarm on their phone for a time you choose, thinking aloud as they go.
  2. Your only allowed words are "What are you thinking?" and the echo of their last word.
  3. Afterwards, write down every moment you wanted to help.

Reflection questions: What did they say that surprised you? What would you have lost if you'd answered their first question?


9. If you already practise

  • Record one of your sessions and mark every time you spoke during a task. Classify each: prompt, echo, boomerang, Columbo, help, praise. Aim to remove the last two.
  • Give observers a list of moments to watch for (hesitations, wrong paths, misreadings), so they aren't tempted to judge the participant.
  • Train new moderators by having them take notes first, then moderate with you as note-taker, then alone.
  • Steve Krug shares the usability-test scripts he uses, free, on his site. Adapt one instead of starting from a blank page.

Checklist

Learn more

Books, courses and videos to go deeper

Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.

Usability testing with five people · Running the session and listening · Tasks and scenarios · How people think · Test with three people · Measuring usability

Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close