UX Design Fundamentals · 20 min · 140 XP
Running a usability test
Prepare the session properly, then turn five people's confusion into a ranked list somebody can act on.
Writing the tasks and keeping quiet is covered in Figma for Product Teams → Test it with five people. This lesson is what happens either side of the session: getting ready, and turning what you saw into decisions.
Before anyone sits down, write a script. It doesn't need to be clever: a sentence about what this is and that you're testing the design rather than them, permission to record, one easy warm-up task, the real tasks in order, and a wrap-up question. Reading from a script is what makes five sessions comparable. Without one you will unconsciously run the last session better than the first.
Pilot it with a colleague first. Every pilot finds something: a dead prototype link, a task that takes twenty minutes, a question you can't say out loud without laughing. Finding it on a colleague costs nothing. Finding it on your first real participant costs you one of five.
Say what you're recording and who sees it, and mean it. People behave differently when they're unsure whether their screen is being shared with their manager. Tell them they can stop at any point, and make stopping genuinely easy.
During the session, capture observations, not conclusions. One row per thing that happened: participant, task, what they did, and their words if they said something. “P3 clicked Settings twice looking for billing” is an observation. “Navigation is confusing” is a conclusion, and writing it down early means you stop collecting the evidence that would have told you which part.
P# Task What happened Quote
── ──────── ─────────────────────────────────── ─────────────────────
P1 Refund Opened Orders, scrolled past the "I thought refunds
row twice, used search instead lived under Payments"
P2 Refund Clicked the row, missed "Refund" —
in the ⋯ menu for 40s
P3 Refund Completed in 15s —
P3 Invite Stopped at the seat limit, no idea "Does this cost me
what to do next money?"Afterwards, group the rows by problem, not by participant. Five per-person summaries are five stories; one list of problems with the people who hit each is a decision-making document. Then rate each problem's severity (Nielsen Norman Group's familiar 0–4 scale works fine) by combining how many people hit it, how badly it blocked them, and whether they got past it on a second try.
Frequency is not severity. One person who could not complete a core task outranks five people who thought a label was slightly odd. A single blocker is a finding; a single preference is an anecdote. Say which you have.
Do it this week: run one session from a written script, take notes as observations only, and write up the findings as a ranked list with a severity and the participant numbers beside each. If the write-up is longer than a page, you have written stories rather than findings.
Loading your workspace…