Plan, moderate and report a small usability test of your prototype using realistic tasks and the think-aloud method.
Why it matters
A usability test means watching real people try to complete real tasks with your design. It is the single most reliable way to find out whether a design works. Teams are often surprised: the button everyone in the meeting thought was obvious goes unnoticed by four out of five participants. Five sessions of 30 minutes, run with a clickable prototype, usually reveal the biggest problems.
The core idea: tasks, think-aloud and silence
In a typical moderated test, a facilitator sits with one participant (in person or on a video call), gives them realistic tasks and asks them to think aloud: say what they are looking at, trying and expecting. The facilitator mostly stays quiet and watches. A note-taker, if you have one, records what happens.
You are testing the design, not the person. Say this at the start, and mean it.
Planning the test
1. Goals. What do you need to learn? "Can people split a bill unequally and send requests without help?"
2. Participants. Five people who match your persona's criteria and did not help design the product. For practice, the same recruiting approach as your interviews works.
3. Tasks. Write tasks as realistic scenarios with a goal, not instructions:
| Weak task | Better task |
|---|---|
| Click "Split this bill", then choose "No". | The internet bill just arrived. Leo agreed to pay $20 because he works from home. Ask everyone for their share. |
| Find the settings page. | You keep getting reminders at 7 a.m. Make them come later. |
Avoid words that appear on the button you want them to find; otherwise you test word-matching, not understanding.
4. Script. Write an intro (thanks, purpose, "we're testing the design, not you", think aloud, consent to record), the tasks, a follow-up question after each task, and closing questions.
5. Success criteria. Decide in advance what counts as success for each task (for example "reaches Confirmation with Leo at $20").
6. Pilot. Run the whole test once with a friend and fix confusing tasks.
Moderating
- Start each task by reading it aloud and handing over a written copy.
- When people go quiet, prompt neutrally: "What are you thinking?"
- If they ask "Should I tap this?", reflect it back: "What would you do if I weren't here?"
- Do not help unless they are completely stuck. If you must, note it: the task becomes "completed with assistance".
- Watch behaviour more than opinions. "I like it" means little if they took three minutes and tapped the wrong button twice.
- After each task, ask a short question, such as "How easy or difficult was that?" (you will measure this formally in the next lesson).
Remote unmoderated tools exist with free trials or limited free tiers, but limits change often; for this course, moderated sessions on a free video call or in person are the simplest option.
Analysing and reporting
Right after each session, write down what happened on each task: success, success with difficulty, failure, and the key observations with timestamps if recorded.
After all sessions, merge observations into findings. Each finding should have:
- What happened (the observation), and how many participants it affected.
- Why it probably happened (your interpretation).
- Severity (for example: critical blocks the task; serious causes delay or errors; minor is an annoyance).
- A recommendation.
Example finding:
F1 (critical, 4 of 5): Participants did not notice the "Split equally: No" option and
edited the bill total instead. They read "Split equally" as a heading, not a control.
Recommendation: replace the radio pair with a clear "Change amounts" button under the
list of shares.
A short report has a one-paragraph summary, the task success table, the top findings ordered by severity, and highlights such as short quotes or clips (with consent).
Common mistakes
- Leading tasks that name the control to use.
- Helping too soon. Struggle is the data.
- Defending the design during the session.
- Testing with teammates. They know too much.
- Long reports nobody reads. Lead with the top three findings and recommendations.
- Testing once. Fix, then test again: two small rounds beat one big one.
Quick recap
- Usability tests watch people attempt realistic tasks while thinking aloud.
- Plan goals, five matching participants, scenario tasks, a script and success criteria; pilot it.
- Moderate neutrally, let people struggle, and record behaviour.
- Report findings with counts, severity and recommendations, then iterate.
Try it
On the Do tab you will plan, run and report a five-person usability test of your prototype.