Choose one ticket with a date that can go wrong
Build a small appointment confirmation for a fictional repair shop. The customer reserves a pickup for 6:30 p.m. in Los Angeles on January 15, 2026. An employee opens the same confirmation in London, where the instant falls on January 16. Should the ticket say Thursday or Friday? Before writing a test, write the product decision. For this exercise the printed ticket promises the shop’s local pickup time, even when the person printing it is traveling. Another product could reasonably promise the viewer’s local time. Each promise needs different tests.
Keep the fixture small. Store the instant as 2026-01-16T02:30:00Z and the pickup location as America/Los_Angeles. Give the ticket a quantity of items and a short pickup instruction. Use invented order details. A reviewer should be able to understand the entire case in a minute and repeat it without an account.
Change one setting at a time
First fix the display language and change the device time zone from Los Angeles to London. The promised pickup day and hour should remain January 15 at 6:30 p.m. local to the shop. Next hold the device zone fixed and change the display locale. The ordering of the date fields or the hour convention may change while the appointment stays the same. JavaScript’s Intl.DateTimeFormat accepts both a locale and an explicit time zone. Record the options you pass and the runtime you used so a reviewer can repeat the comparison. Date and time formatting options.
Include a deliberately broken version that formats the instant using the device’s default zone. Save the inputs and the incorrect local day beside the expected ticket. Then show the corrected result under both device settings. Check the year, month, day, hour and minute as fields where practical. A test that insists on one exact punctuation pattern can fail even when the pickup promise is intact.
Give the translator a whole sentence
Add a second case using “Collect one item” and “Collect three items.” Hand a language reviewer the complete message, where it appears, and what the quantity counts. ICU’s guidance on plural messages explains how a translator can choose among complete alternatives for the target language. Let the reviewer own natural wording. A program that selects the expected plural branch supplies evidence about the code; it still needs someone qualified to judge the sentence. ICU guidance on formatting messages.
If you cannot obtain a language review, label that part of the sample pending. You can still demonstrate a layout check with a deliberately expanded test string. Show the ticket at a narrow width and inspect whether the time, location and instruction remain visible. Describe this as a layout experiment. An invented string provides no evidence of translation quality.
Let the reviewer change the appointment
Package the fixture, a command or short sequence that runs both versions, expected results and one defect report. Lead the report with the customer consequence: someone printing the ticket abroad could arrive on the wrong day. Identify the owner of the fix, the smallest input that reproduces it and the checks that should pass before release. A screenshot can show what happened once. The fixture lets another person challenge your explanation.
Invite a hiring manager to move the appointment by one hour and rerun the checks without your help. Ask them to judge whether the intended behavior is clear, the failure repeats and the correction preserves the promise. Use that same task and evidence standard for every applicant. This follows the Vitaver Playbook’s Signal vs. Noise practice: keep the observations that change the release decision, including the language review still owed.