Write a moderated usability test script whose tasks give nothing away

Write the script for a moderated usability test session.

What we are testing: {{prototype, staging build or live product, and the flow or screens in scope}}
Participants: {{the user group or groups and how they were recruited; say if there is more than one distinct group}}
What we need to decide afterwards: {{the design question this round has to answer}}
Goals real users come here to accomplish: {{3 to 6 goals in plain words, e.g. "change the delivery address on an order already placed"}}
Session length and format: {{e.g. 45 minutes, remote over video, think-aloud}}
Interface wording to keep out of the tasks: {{menu items, button text and feature names on the screens in scope}}

Write it in four parts.

Facilitator introduction, about two minutes, read aloud. Make clear that we are testing the product and not the person, and that nothing they do can be wrong. Ask them to think aloud as they go. Explain that I will mostly not answer questions during the tasks, because I want to see what they would do without me in the room, and that I will answer everything at the end. Ask for consent to record. Say nothing about what we changed or what we hope to learn.

Tasks. Turn each goal into a scenario the participant acts out:
- Give them a reason for doing it and every specific they need (names, dates, amounts), so they never have to ask.
- Use none of the interface wording I listed. If a goal cannot be described without one of those words, flag it, because that is already a finding about the interface.
- Ask them to do it. Never "how would you" or "where would you click".
- Order the tasks so an early one does not teach the path to a later one, and explain the order you chose.
- Include one open task where they apply their own criteria alongside specific tasks with a clear end state.
For each task, write the scenario exactly as I read it, what counts as success, the point at which I step in, and two neutral prompts for when they go quiet, in the style of "What are you looking for right now?" and never a hint.

Observer sheet. One row per task, with columns for completed (yes, with help, no), where they hesitated or went back, anything they said that shows how they think the product works, and severity. Add a line telling observers to stay silent and record what happened rather than what they think it means.

Debrief, about five minutes. Their overall impression, the moment they felt least sure, and then the questions they held back during the tasks.

Close with the plan for the round: five participants for a single user group; three or four from each if there are two distinct groups; at least three from each if there are more. Then say which kinds of finding should be fixed before a second round of five, instead of running everyone in one batch.

Tip: The task rules follow Nielsen Norman Group guidance on turning user goals into usability test scenarios: realistic, acted out rather than described, and free of interface wording or steps. The round size follows the Jakob Nielsen model in which one user surfaces about a third of the usability problems and five surface about 85 percent, which is why three rounds of five with fixes between them find more than one round of fifteen. The model assumes comparable users, so split the count across distinct groups. Paste the interface wording list in full: a task that borrows a button label tells you only that people can read.