Figma and design tools · Beginner · 17 min
Test a Figma prototype with users and log what breaks
By Padding · Updated 18 Nov 2026
Figma (web) and the Figma mobile app · November 2026
You will run short test sessions with the Larkfield Clinic prototype on a phone and a laptop, sort every problem into a prototype glitch or a real design problem, and end with an issue log where each row links to the frame it happened on.
On this page
A prototype test produces two kinds of trouble, and they look the same in the moment. Someone taps the date and lands on the "not ready yet" screen. Is that because the date really should be tappable and the design hides the way to change it, or because the prototype was only ever wired for one path? If you log both kinds as "problems", part 10 fixes things that were never broken. If you dismiss both as "just the prototype", the real problems ship.
This is part 9 of the series that takes the Larkfield Clinic booking app from paper to a tested design. In part 8 you made the prototype safe to test: the task paths work, dead ends land on a Not part of this test screen, the data looks real and the share link opens on Home. Here you put it in front of people. You run sessions of about 20 minutes, some on a phone and some on a laptop, and the output is not a report. It is an issue log: one row per problem, sorted by kind, each one linked to the frame where it happened, so the next part can open the exact screen and change it.
If you have never run a usability session, read Run a five-person usability test first. That tutorial covers moderating and note-taking in general. This one covers what is different when the thing under test is a Figma prototype.
Step 1 · Build the issue log before the first session
Create a spreadsheet called Larkfield prototype test – issue log. Add one row of headings:
ID: L01, L02 and so on. Short IDs are what you write in Figma comments later.Frame: the frame name exactly as it appears in the file, such asPick a timeorYour details / For someone else.Frame link: a link that opens that frame in Figma. Filled in after the sessions.Task: which test task the person was doing.What happened: what you saw and heard, in plain description. "TappedTue 9 Decheading twice, then said 'can I not change the day?'" Not "the date picker is confusing".Kind: one ofDesign,PrototypeorUnsure.Participants: who hit it, as P1, P2 and so on, never names.Device: Phone, Laptop or Both.
Add a second tab called Sessions with one row per participant: P number, device, date and time, and a column for tasks completed. Keep names and contact details out of this file entirely; they stay wherever your consent records live.
Then write the tasks at the top of the Sessions tab so you read them out the same way every time. Part 8 prepared four tasks: travel jabs, a blood test, Tomi's check-up and the reminder. For this round, rewrite them as three tasks that follow one booking from start to finish, including changing a time before confirming, which none of part 8's tasks asks for. Part 10 and the second round use these three:
- "You need a nurse appointment for a dressing change sometime next Tuesday morning. Book one."
- "You have just picked 10:30 but realise you can only do after 11. Change it before you confirm."
- "Once it is booked, find out when you will be reminded."
The new tasks need a few additions to part 8's build. Make them on the prototype page the way part 8 did, before any session:
- On
Choose a service, linkNurse appointment(part 8 sent it toNot part of this test) to a copy of Pick a time namedPick a time / Nurse Tuesday. It shows Tuesday 9 December, with 10:30 and 11:15 free among taken slots. - Give each free slot its own path, so the summary shows the time the participant chose. Make Confirm and Booked copies named
Confirm / Nurse 10:30,Confirm / Nurse 11:15and a Booked copy for each. Link10:30toYour details, with Continue on its filled-in version going toConfirm / Nurse 10:30. Then duplicateYour detailsand its filled-in version for11:15, link11:15to the copy, and point the copy's Continue atConfirm / Nurse 11:15. - Leave the back arrows on part 6's Back action, so two taps from Confirm return to
Pick a time / Nurse Tuesday. Leave theChangelinks on both Confirm copies going toNot part of this test, as part 8 wired them: this round tests whether people change the time by going back, and part 10 decides whatChangeshould do. - Change the message on
Booked / Reminder textso it names the nurse appointment on Tuesday rather than Tomi's check-up. - Run part 8's tap-every-element check on the screens you changed, play tasks 1 to 3 once, and save a named version with Save to version history. Keep the name
Test build, round 1and list the three tasks in the description, so the newest round 1 version is the build you actually tested.
Check it worked
Your spreadsheet has two tabs. Issue log has the eight headings and no rows yet. Sessions has this round's three tasks written word for word at the top and an empty row for each booked participant, labelled P1 to P6 (or however many you have). In the file, tapping Nurse appointment opens Pick a time / Nurse Tuesday, tapping 11:15 there and continuing ends on Confirm / Nurse 11:15, Change beside the time on either Confirm copy opens Not part of this test, and version history shows the newest Test build, round 1 with the three tasks in its description.
Step 2 · Set up the phone and the laptop the same way each time
Test on the device people would book on. Most Larkfield patients would use a phone, so put most sessions on a phone and one or two on a laptop to see whether the layout holds up when people use a mouse and a wide screen.
On the phone, use the Figma mobile app, signed in to your own account. Hand participants your phone rather than sending them the link, because the app needs a Figma account and most patients will not have one. Open the file, start the prototype from the presentation view, and in the prototype options turn hotspot hints off. With hints on, a tap in the wrong place flashes a blue box over everything clickable, which tells the participant where to tap and wrecks the finding you are trying to observe.
Opening the share link in a phone's web browser also works, but it is a weaker test: a mobile browser only plays hotspot interactions and adds its own address bar and toolbar, which push your screen down and are not part of the Larkfield design. Use the browser only as a fallback. Part 8 checked the app before this round's new frames existed, so repeat its checks once in the app: restart the prototype and confirm it returns to Home, and scroll the list on Pick a time / Nurse Tuesday to confirm it scrolls.
On the laptop, open the share link in a browser, signed in as you. In the presentation view's options menu, turn off "Show hints on click", then go full screen so the file name and Figma toolbar are out of sight. Learn the Restart control (or press R): it takes the prototype back to the flow's starting point, which is how you reset between tasks and between participants without showing anyone the canvas.
Before each session, play task 1 through to Booked on that device, then restart. This catches a link you broke since yesterday and leaves the prototype on Home.
Check it worked
On both devices, tapping or clicking an empty part of Home does nothing visible: no blue boxes. On the laptop you see only the Home frame, full screen. Pressing R (or Restart) after reaching Booked returns you to Home in one step.
Step 3 · Tell people what the prototype cannot do
A participant who does not know they are using a mock-up will spend their time fighting it. Before the first task, say something like this, in your own words:
"This is an early version of a booking app, made of pictures that link together. Some things will not work the way a real app would. Where you would normally type, tap the first box once and the details fill in for you, so you do not need to type anything. If you tap something that is not ready yet, you will see a screen that says so. That is fine, it tells us where you expected to go. Tell me out loud what you are looking for and what you expect to happen."
Then keep a short limits card next to you, written from part 8's build and this round's additions. It lists every place where the prototype behaves differently from the app you intend to build, so you recognise them instantly instead of puzzling mid-session. For Larkfield, the card says:
Your details: tapping the first field jumps to a filled-in copy of the screen. Any other field, tapped before or after that, opensNot part of this test.Pick a time / Nurse Tuesday: only Tuesday 9 December can be booked. Other days, in the week strip, the date-picker overlay or from the day heading, openNot part of this test.Choose a service:Travel health,Blood testsand the child check-up still run part 8's task paths through to their own Confirm and Booked. A participant who picks one by mistake can finish a booking for the wrong service, so note where they went rather than stopping them.- Every screen: the back arrow works; the phone's own back gesture may not.
Booked: nothing on it sends or shows a real reminder. For task 3, once the participant has answered, showBooked / Reminder textby opening that frame on your laptop (it is not linked in the prototype) and ask whether it matches what they expected.
Check it worked
Your limits card exists on paper or in the Sessions tab, it has one line for every dead end and faked control in the prototype, and each line names the frame it applies to. Read your opening lines aloud once: they mention the one-tap fill on the form and the "not ready yet" screen.
Step 4 · Run each session and note the frame, not just the moment
Give the tasks one at a time, in order, and say as little as you can while people work. Restart the prototype between tasks only if they did not finish where the next task begins.
The note-taking habit that makes this tutorial work: every observation starts with the frame name. You know the frames because you built them, and the screen headings match the frame names (Pick a time, Your details). Write Pick a time: and then what happened. If someone lands on Not part of this test, write which frame they came from and what they tapped, because that is the useful part: it shows where they expected the app to go.
Notes taken this way look like this:
Choose a service: P2 scrolled past Nurse appointment twice, said "I want a dressing, not a nurse".Pick a time / Nurse Tuesday: P2 tapped the Tue 9 Dec heading, landed on Not part of this test, said "how do I see Wednesday?"Your details: P2 filled the form with one tap, then tapped the phone field, landed on Not part of this test, said "I wanted to check the number", went back.Confirm: P2 read the summary aloud, tapped Confirm without pause.
Twenty minutes is enough for three tasks plus a few minutes of questions at the end. Ask two closing questions every time, the same way: "What, if anything, was hard to find?" and "Is there anything you expected to be able to do and could not?" The second one tends to surface prototype limits and missing features in the participant's own words.
Fill in the Sessions row straight after each session: device, time, and which tasks were completed without help.
Check it worked
After each session, every line in your notes starts with a frame name that exists in the file. Each visit to Not part of this test has the frame it came from and the thing tapped. The participant's row on the Sessions tab is filled in before the next person arrives.
Step 5 · Sort every observation into design, prototype or unsure
Do this the same day, while you still remember what people did with their hands. Go through your notes line by line and ask one question of each: would this still happen in the finished app, built as designed?
- Prototype: no. The trouble comes from something the prototype fakes or leaves out, and the person's intent was reasonable. P2 tapping the filled-in phone field and reaching the dead end is a prototype limit; the built app will open a keyboard. Log it as
Prototypeso nobody redesigns the field. - Design: yes. The design itself misled, hid or confused. P2 scrolling past
Nurse appointmentbecause they think of it as "a dressing" is a naming problem that the built app would share. Log it asDesign. - Unsure: you cannot tell yet. P2 tapping the day heading on
Pick a timemight mean "I want to choose another day", which the design does not offer anywhere, or it might mean the prototype's single-day wiring confused them. Log it asUnsureand say inWhat happenedwhat would settle it.
Three patterns are worth knowing because they come up in almost every prototype test:
- A visit to the dead-end screen, or a wrong turn into another task's path, is a prototype event but often a design signal. If two people opened
Blood testsorTravel healthwhile doing the nurse task, the service names are not doing their job, wherever the tap happened to lead. - A missed hotspot (the person tapped near a button but just outside its link) is a prototype glitch if the hotspot is smaller than the drawn button. If the drawn button is itself small, it is a design problem: check the frame before you decide.
- Hesitation on a fake control is a design finding if the control looked fine but the person did not trust it. "I'm not sure this has saved" on
Your detailsis about feedback in the design, not about the fields being faked.
Copy each observation into the Issue log as its own row. When two people hit the same problem, add them to one row's Participants cell rather than making two rows.
Check it worked
Every line from your session notes is in the Issue log, and every row has a Kind. Each Unsure row says what would settle it, such as "ask P4 in the next round" or "wire a second day and retest". No row's What happened is a solution ("make the date a picker"); solutions wait for part 10.
Step 6 · Link each issue to its frame in Figma
The log is only useful if the person fixing an issue can open the exact screen from it. Open the Larkfield file on your laptop and go to the page with the tested frames. For each row in the log:
- Select the frame named in the
Framecolumn on the canvas or in the layers panel. - Right-click it and choose Copy/Paste as, then Copy link to selection. This copies a link to the selected frame, not the whole page.
- Paste the link into the row's
Frame linkcell, then open it in a new tab to make sure it lands on that frame and not just the page.
Then add the issue IDs to the file itself. In comment mode, drop a comment on the part of the frame the issue is about, and write only the ID and a one-line summary: L04 – Design: nobody found how to see Wednesday. One comment per issue, placed on the element, not the frame corner. Anyone with access to the file can read these comments, so keep P numbers and quotes in the spreadsheet and out of Figma.
Leave the tested frames exactly as they are, including Not part of this test and every link between them. Part 10 compares its revised screens against these, and a second round of testing reuses the same tasks and dead ends.
Check it worked
Every row in Issue log has a Frame link that opens the named frame. Every Design and Unsure row has a matching comment pin in Figma with its ID; Prototype rows can have one too if you like. Clicking through the comments on Pick a time shows the same IDs as filtering the log by that frame.
Step 7 · Tally the log and hand it on
When the last session is done, tidy the log so part 10 can start from it without asking you anything.
Sort by Kind, then by how many participants each issue affected. Add a Count column if it helps. Look at the Device column: an issue that only happened on the laptop may be about the wide layout, and one that only happened on the phone may be about thumb reach or the screen size. Note that in the row.
Then write three short lines at the top of the Issue log tab:
- How many sessions you ran, on which devices, and how many people completed each task without help (copy the totals from
Sessions). - The two or three
Designissues that more than one person hit. - Which
Prototypeissues should be fixed in the prototype before a second round, because they got in the way of the test itself. A dead end people kept hitting, or a hotspot nobody could find, falls here.
Do not change any screens yet. Deciding which problems matter most and how to fix them is the next part's job, and it needs this test's version of the file intact to compare against.
Check it worked
The Issue log tab opens with a three-line summary, then rows sorted by kind and count. Every row has a frame, a link, a kind, participants and a device. The Larkfield file is unchanged apart from the new comments: the tested frames, connections and the Not part of this test screen are all still there.
Next steps
Part 10, Revise the design after a prototype test, takes this log, ranks the issues and changes the screens and prototype for the top problems. To give each issue a severity rating before you start fixing, use Rate usability problems by severity.
If your sessions kept stalling on the prototype itself, go back to Get a prototype ready for a usability test and run its checklist again before a second round. For the moderating side of testing in more depth, see Run a five-person usability test.
Part of these paths
Related tutorials
- Get a prototype ready for a usability testYou will have a Larkfield Clinic prototype that a participant can open from one link, finish every test task in, and never get stuck in, with a saved version to fall back on.
- Run a five-person usability testYou will run five in-person think-aloud sessions on a live product in one day, hold back from helping when people struggle, and finish with a list of observed problems showing how many of the five hit each one.
- Revise the design after a prototype testYou will have a revised Larkfield Clinic prototype that fixes the worst problems from the first test, a page that shows each change before and after beside its evidence, and a change log that tells the second round of testing what to watch.
Need this done for you?
Padding does this work for growing companies, from the first diagnosis to a system that keeps working.
Get new tutorials