Acceptance testing a small app with fake data: 61 scenarios

Automated tests tell me that the code does what I wrote. An acceptance test asks a different question: can someone do their actual job with it. So before a real salon touched the app, a separate AI session played the part of the owner on a test computer with fake data. It ran 61 scenarios in four rounds.

The first round found 19 defects, some of them serious. All were fixed. The later rounds came back clean.

The scenarios were chosen to be annoying. First login, with the forced password change. Too many wrong passwords, and what the lock does. An Excel file with Italian column names, dates in three formats, phone numbers written with spaces, dashes and a +39 in front, accented names and a few broken rows. The same file imported twice. A cancellation that happens between two imports. A computer whose clock was set to the wrong time zone. Two salons with several users each, to see that nothing leaks across them and that a member of staff cannot reach the owner's settings. A list of 500 customers with 2000 appointments, to see how the pages behave when the table stops being small. A double click, a slow network, no network, and a session that expires while a form is open.

It also checked how the app looks at 390 pixels wide, which is a phone, and on a desktop, in the light and the dark theme, and whether every form can be used from the keyboard.

The report is honest about what it could not prove. The real Excel export of the salon's booking software and a real WhatsApp message are not reachable from a test computer. Those became a short list for the first session with the owner: log in, import two weeks of appointments, check three of them against the original, report a free slot by hand and open the link on the owner's own phone.

I now run an acceptance round before I hand any app to someone. It finds a different kind of mistake from the one my tests find.

Disclosure: drafted with a local AI model, then reviewed and approved by me. Originally published at https://bbal-blog.kairo-app.workers.dev/acceptance-testing-a-small-app-with-fake-data-61-scenarios/