Skip to main content

UX and research · Beginner · 14 min

Plan a usability test with tasks and success criteria

By Padding · Updated 18 Nov 2026
Any document editor and a spreadsheet · November 2026

You will have a written usability test plan for a booking flow, with scenario tasks that do not give away the answer, a success rule for each task and a session script another person could run without you.

On this page

A usability test is only as good as the tasks you hand people. Most weak tests fail before the first participant arrives: the task says "Use the Book button to book a flu jab", the participant finds the button labelled Book, and the team learns that people can read. The plan is where you stop that from happening.

This tutorial plans a test for Larkfield Clinic, a fictional community clinic, and its appointment booking app. The flow has six screens: Home, Choose a service, Pick a time, Your details, Confirm and Booked, after which the patient gets a reminder. You will write down what the test must find out, turn that into scenario tasks, rewrite the tasks so they stop pointing at the interface, decide what counts as success, choose what to record, and write the script the facilitator reads aloud. Swap in your own flow as you go; the steps are the same.

By the end you will have one document and one spreadsheet. The test is ready when a colleague who has never seen the flow could pick them up and run a session.

Step 1 · Write down what the test has to find out

Open a new document called "Usability test plan: Larkfield booking". Before any tasks, write a short section headed What we need to learn. Each line is a worry or a question the team has about the flow, written as something a session could confirm or rule out.

For Larkfield, the team's list might read:

  • Do people find the right service when the clinic's name for it differs from theirs (the clinic says "Travel health", patients say "jabs for my holiday")?
  • Do people notice that the first slot shown is not the earliest one available?
  • Do people understand they are booking for someone else when they book for a child?
  • Do people know, on Booked, whether a reminder is coming and how?

Keep it to three to six lines. If you already have a research plan, copy its questions here instead of writing new ones; the tutorial on writing a one-page research plan covers how to get them agreed.

Under the list, add one line saying what is being tested: "Clickable prototype v3, mobile size, all six screens linked. Reminder shown as a text message mock-up after Booked."

Check it worked

Your document has a "What we need to learn" section with three to six lines, and each line names a specific part of the flow (a screen, a label, a decision) rather than "is it easy to use". There is one line naming the exact prototype or build version.

Step 2 · Turn each question into a goal, not an instruction

For each line in your list, write the goal a real patient would have that leads them through that part of the flow. A goal is what the person wants done in their own life. An instruction is what you want them to do in your interface. Write goals.

Put them in a numbered list headed Tasks (draft):

  1. Book travel vaccinations for a trip next month.
  2. Book the earliest available appointment for a blood test.
  3. Book a check-up for your seven-year-old son.
  4. Find out whether you will be reminded about the appointment you just made.

Each task should end at a point you can see on screen, so you know when it is over. Tasks 1 to 3 end on Booked. Task 4 ends when the participant tells you what they expect to happen.

Order them so earlier tasks do not teach the answers to later ones. Here, task 4 comes last because it builds on a booking that already exists. If two tasks pass through the same screen, put the one you care most about first, while the participant is still fresh.

Check it worked

Read each draft task and ask "would a patient say this sentence about their own day?". Every task should pass. Each task has a clear end point you have written next to it (for example "ends on Booked").

Step 3 · Rewrite every task so it does not give away the answer

This step is where most of the value of the plan comes from. Draft tasks almost always leak the interface: they use the clinic's labels, the order of the screens, or a hint about where to look. Participants then match words instead of thinking, and the test finds nothing.

Go through each draft task and look for three kinds of leak.

Interface words. Any word that appears as a label, button or heading in the flow. "Book travel vaccinations" is weak if the menu item says "Travel vaccinations"; the participant scans for the matching phrase. Replace it with the words a patient uses.

Steps in the right order. "Choose a service, then pick a time" describes the flow back to the participant. Give them the goal and let them find the order.

Hidden shortcuts. "Book the earliest appointment" tells them earliest matters. If the question is whether they notice the first slot is not the earliest, they must not be told to look for it.

Then turn each task into a short scenario: who they are, why they need this, and one or two concrete details they will need on the way. Details let them make the same choices a real patient would, and give you something to check at the end.

Rewrites for Larkfield:

  • Before: "Book travel vaccinations for a trip next month." After: "You are going to Ghana on 14 December and your friend said you need some jabs before you go. Get that sorted with Larkfield Clinic."
  • Before: "Book the earliest available appointment for a blood test." After: "Your GP wants you to have a blood test and said the sooner the better. Arrange it at Larkfield. You can do any weekday morning."
  • Before: "Book a check-up for your seven-year-old son." After: "Your son Tomi is seven and is due a check-up. Arrange one for him. You can take him any afternoon after school."
  • Before: "Find out whether you will be reminded about the appointment." After: "You tend to forget appointments. Looking at what you have now, what do you expect to happen between now and the day?"

Notice what changed. "Travel vaccinations" became "jabs". "Earliest" became "the sooner the better", which is how a doctor says it, and leaves the participant to decide what to pick. "Book for your son" became a scenario where the son has a name and a constraint, so you will see whether the details end up under the right person.

Keep a column next to each scenario called Leaks removed and write what you took out, for example "removed 'travel vaccinations' (menu label)". This helps the next person who edits the plan avoid putting it back.

A last check on wording: read each scenario aloud. If it takes more than about twenty seconds, it is too long for a participant to hold in their head. Cut background they will not use.

Check it worked

Open the prototype next to your document. For each scenario, search the screens for every noun and verb in it. None of the scenarios should share a distinctive word with a button, menu item or heading in the flow. Your "Leaks removed" column has at least one entry for each task.

Step 4 · Decide what success looks like for each task

Without a written rule, two observers will disagree on whether a session "went fine". Write the rule before testing so nobody adjusts it afterwards to fit what they hoped to see.

For each task, write three things in a table:

  • Success: the end state that counts as done, stated so you can check it on screen or in what the participant says.
  • Partial: done, but with something wrong or with help.
  • Fail: not done, wrong outcome, or the participant gave up.

Add a Stop rule row for the whole test: when the facilitator ends a task and moves on. A common rule is "the participant says they would give up, or eight minutes pass, or they ask for help twice". Pick numbers that suit your flow and write them down.

For Larkfield:

| Task | Success | Partial | Fail |
| --- | --- | --- | --- |
| 1 Travel jabs | Reaches Booked with the travel health service selected and a date before 14 December | Reaches Booked with the right service after one hint, or picks a date after 14 December without noticing | Books a different service, or gives up |
| 2 Blood test | Reaches Booked with the earliest weekday morning slot shown anywhere in the list | Reaches Booked with a weekday morning slot that is not the earliest | Does not reach Booked, or books an afternoon |
| 3 Tomi's check-up | Reaches Booked with Tomi's name in Your details and an afternoon slot | Books under their own name, then notices and corrects it | Books under their own name and does not notice |
| 4 Reminder | Says they will get a reminder and how (for example "a text the day before") | Says they will get a reminder but not how | Says they do not know, or expects something the flow does not do |

Hints count against success because a real patient at home has no facilitator. Note in the plan what a hint is allowed to be ("You can try anything you would try at home") and what it is not (pointing at the screen, naming a button).

If you are writing this as a table in a document that does not show tables well, use one heading per task with the three lines underneath. The format matters less than every task having all three.

Check it worked

Every task in your plan has a success, partial and fail line, and each one names something you could check on screen or hear the participant say. There is one stop rule written down, with a time limit in minutes.

Step 5 · Choose what you will record and set up the sheet

Pick a short list of measures. Five participants will not give you statistics, so choose things that help you see problems and decide what to fix, not numbers to put in a slide.

A sensible set for a first test:

  • Outcome per task: success, partial or fail, from the rule you just wrote.
  • Wrong turns: each time the participant goes to a screen or taps something that does not move them towards the goal. Count them and note where.
  • Ease rating: after each task, ask "On a scale of 1 to 7, where 1 is very difficult and 7 is very easy, how was that?". A low score on a task they completed points to friction the outcome hides.
  • Quotes: what they say at the moments they hesitate.

Leave out time on task unless speed is one of your questions. Participants who think aloud are slower than they would be at home, so the times mislead.

Create a spreadsheet called "Larkfield usability test: observations". Make one tab per participant, named P1 to P5, with these columns:

| Task | Outcome | Wrong turns (count) | Where they went wrong | Ease (1 to 7) | Quotes | Notes |

Add one row per task, with the task number and a short name already filled in. Add a final tab called Summary with one row per task and one column per participant for the outcome, so the note-taker can fill it in straight after each session.

Check it worked

Your spreadsheet has five participant tabs and a Summary tab. Each participant tab has one row per task, with the task names already typed in, and the columns above. Ask a colleague to open it and tell you what goes in "Wrong turns"; if they cannot say, add a one-line note at the top of the tab explaining it.

Step 6 · Write the session script

The script is what makes the test repeatable. Write it as the exact words the facilitator says, with the things they do (not say) in brackets. Aim for a session of about 45 minutes, with the tasks taking around 25 of them.

Use these parts, in this order:

1. Welcome (2 minutes). Who you are, what the session is for, how long it takes. Say plainly that you are testing the app, not them: "If something is confusing, that is a problem with the app, and it is exactly what we want to find."

2. Consent (3 minutes). Explain what you are recording and who will see it, then get consent before you start recording. Link to the consent form you will use. The tutorial on getting informed consent for user research covers the form and the conversation.

3. Think aloud warm-up (2 minutes). Ask them to say what they are thinking as they go. Practise on something unrelated, such as "Show me how you would find a bus time on your phone, and talk me through it." This is easier than explaining the idea.

4. Background questions (5 minutes). Two or three questions that help you read their behaviour later: "When did you last book a medical appointment? How did you do it?" Keep these short; this is not an interview.

5. Tasks (about 25 minutes). For each task, write: the scenario word for word, a line saying "[Hand over the card with the scenario printed on it]", the starting screen ("[Start on Home]"), the stop rule, and the ease question. Add two neutral prompts the facilitator may use when the participant goes quiet: "What are you looking for?" and "What did you expect to happen?".

6. Debrief (5 minutes). Two or three open questions: "What, if anything, was harder than you expected?" and "If you could change one thing, what would it be?". Then answer any questions they saved up.

7. Close (2 minutes). Thank them, explain any incentive and when it arrives, stop the recording.

Underneath the script, add a Facilitator rules box of four or five lines: do not help unless the stop rule is reached; answer questions with a question; do not say "good" or "right" after an action; reset the prototype to Home between tasks.

Check it worked

Read the script aloud with a timer. The total is within five minutes of your target length, every task in the script matches its scenario in the task section word for word, and each task has a starting screen and the ease question written out.

Step 7 · Hand it over and run a pilot

The brief for this plan was that someone else could run it. Test that directly.

Give the document and the spreadsheet to a colleague who has not worked on the flow. Ask them to read the plan for ten minutes and then tell you, without looking: what the test is trying to learn, what counts as success on task 2, and when they would stop a task. Wherever their answer does not match the plan, the plan is unclear; rewrite that part.

Then run one pilot session with someone outside the project, such as a colleague from another team. Your colleague facilitates; you take notes in the P1 tab. Treat this as a test of the plan, not of the flow. Watch for:

  • a scenario the participant asks you to repeat, which means it is too long or ambiguous
  • a task they finish in seconds because the scenario still points at the answer
  • a stop rule that cuts in too early or never triggers
  • a prototype link that is missing, so the participant hits a dead end you did not mean to test

Fix what the pilot shows, clear the P1 tab, and add a short Changes after pilot section at the bottom of the plan listing what you changed. Then the plan is ready for recruiting.

Check it worked

Your colleague answered the three questions in line with the plan, or you have rewritten the parts they got wrong. The plan ends with a "Changes after pilot" section with at least one entry, and the P1 tab is empty and ready for the first real session.

Next steps

You now have a test plan for the Larkfield booking flow that another person can run. Put it to use:

Need this done for you?

Padding does this work for growing companies, from the first diagnosis to a system that keeps working.

Digital Transformation

Get new tutorials

New tutorials, about twice a month.