Study first
Review the ideas behind the questions
Review the session before participants arrive. Strong moderated research uses believable tasks, neutral facilitation, useful notes, and privacy controls that protect the participant while preserving evidence.
Design Tasks That Reveal Behavior
A moderated task should help the participant act naturally enough for the team to learn. Tasks are weaker when they name the exact feature or turn the session into a preference interview.
- Moderated usability testing watches participants try to complete specific tasks using a service.
- Good test tasks should set a clear goal, feel relevant and believable, be challenging enough, and avoid giving away the answer.
- Task scenarios should be realistic, actionable, and avoid clues that describe the steps.
In Practice
Start From The User Goal
If the task names the feature the team wants tested, rewrite it around the customer's goal so the interface has to prove itself.
Use Several Smaller Tasks When Needed
A complex buying or signup journey can often be tested as smaller tasks in a logical order, with setup time between them.
Common mistakes
Writing a task that tells the participant which menu, feature, or button to use.
Ask for the customer goal and remove step clues so the team can see whether the path makes sense.
Q&A
What makes a usability task believable?
It gives the participant a realistic customer goal and enough context to act without being told the path.
Why avoid asking what the participant would click?
Asking them to act gives stronger behavioral evidence than asking them to describe imagined steps.
Facilitate Without Steering
The moderator should make the participant comfortable, then watch, listen, and ask neutral follow-ups when needed. The goal is to understand behavior, not help the participant produce a polished answer.
- Before tasks start, participants should know the service is being tested, not them.
- A moderator can ask participants to think aloud, then mostly stay quiet and observe.
- Think-aloud testing needs representative users, representative tasks, and space for users to do the talking.
In Practice
Prompt For Thoughts, Not Answers
A short neutral prompt can keep the participant talking, but explaining the expected path turns the session into coaching.
Intervene Only For Learning
If a participant is completely stuck, a neutral recovery prompt can keep the session useful; it should not show the answer.
Common mistakes
Explaining the correct path whenever a participant hesitates.
Let the participant work, listen for confusion, and use neutral prompts only when needed to keep learning.
Q&A
What should the moderator do when a participant becomes nervous?
Explain the session, give them time to settle, and make clear that the service is being tested, not the participant.
When is a follow-up question useful?
Use it when something interesting or unclear happens and the team needs to understand what is behind it.
Capture Evidence Carefully
Session evidence should be useful in analysis and safe to handle. Notes, recordings, observers, personal data, and assistive-technology setup all affect what the team can responsibly learn.
- Good notes should record observations rather than personal interpretations.
- Before notes or recordings begin, informed consent is required from all participants at the session.
- For assistive-technology sessions, participants often need their own setup or device because their configuration can be hard to replicate.
In Practice
Use An Observer When Possible
A note-taking observer lets the moderator focus on the participant and improves the chance of having a complete record.
Protect Notes After The Session
Personal or confidential details should be stored securely and removed from materials when they are not needed for analysis.
Common mistakes
Letting observers write interpretations as if they were facts.
Ask observers to record clear single observations, then separate findings and actions during analysis.
Q&A
What should notes capture first?
They should capture what the participant did, said, seemed to think, and seemed to feel, while keeping interpretation separate.
When should dummy data replace real personal data?
Use dummy data when the service cannot securely process real data or there is not enough time to set up safe test conditions.