Start with the failed task
Fictional brief: students cannot find study rooms, yet staff report empty rooms with active reservations. One student needs to release an unwanted booking; another needs to reserve a room that is actually available. The reports identify a problem to investigate, not its cause.
Before designing, compare bookings with observed use and inspect the cancellation path. Empty rooms could reflect forgotten cancellations, confusing controls, inaccurate records or a delay before released slots become bookable. Each explanation suggests different work.
Compare options by the problem they solve
Do not choose a waitlist just because it adds an obvious feature. First establish how unused capacity becomes available.
| Option | When it could help | What could make it the wrong choice |
|---|---|---|
| Reminder | Students forget to release bookings | They remember but cannot find or complete cancellation |
| Clearer cancellation | The existing cancellation task is difficult | Cancellation already works; room records are inaccurate |
| Waitlist | Released capacity exists but is hard to redistribute | Few slots are released, or availability updates are unreliable |
| Investigate first | The cause is still unclear | Evidence gathering has no owner or finish condition |
Conditional recommendation
Describe behavior an engineer can test
A requirement such as make cancellation easy leaves important behavior unspecified. The first version needs to handle ownership, retries and failures as well as the successful button click.
| Situation | Required behavior |
|---|---|
| Authorized user cancels their eligible booking | The reservation changes state and the slot is released |
| The user retries the same cancellation | The booking remains consistently cancelled; the slot is not released twice |
| The service rejects or cannot confirm the request | Do not show success; provide a recoverable error |
| A user attempts to cancel someone else's booking | Reject the change unless that user has the relevant authority |
More cancellations do not prove better room use
In the fictional task test, A produces 40 completions among 200 eligible users and B produces 50 among 200. Completion rises from 20% to 25%: five percentage points, or 25% relative to A. That is not a 25-percentage-point gain.
The original problem is unused booked room time. Cancellation completion is a diagnostic measure, not proof that other students obtained usable rooms. Check room use, delayed availability updates and accidental cancellations. Also examine assignment, repeated users and logging before attributing the difference to the change.
The worked exercise below asks what the supplied numbers establish. You do not need a complicated statistical claim to give a good answer; you need the correct denominator and an outcome connected to the problem.
Present the choice and the fact that could reverse it
Use the complete product-brief.md example to inspect the reasoning, requirements and test plan. For your own case, lead with the problem and your recommendation, then show the alternative you rejected. Clearly distinguish observations you made from research or tests you propose.
Check the recruiting route as well as the case format. Atlassian's APM program consists of two 12-month rotations; it is not a summer internship. Follow the vacancy and assessment invitation for the role you are actually pursuing.
Sources[1]