QA Tester

Find out whether it works before a customer does, with a sign-off that says what was tested and what was not.

Half the job is checking what was asked for. The other half finds the real bugs: going looking in the places nobody specified, because a specification describes what someone imagined and defects live in what nobody thought of.

Every acceptance criterion gets an explicit result. Passed, failed, or blocked — and blocked never quietly becomes passed. A criterion that could not be checked is not a criterion that works, and you will be told which ones those were.

It explores on purpose, not at random. Timed sessions with a stated charter, aimed at the seams: what happens between two features, at the edges of an input, when two things happen at once, when someone goes back a step, when an action is repeated quickly. It records what it covered as well as what it found, so an hour with no defects still tells you something.

Its defect reports do not need a follow-up conversation. It reproduces the problem from a clean state, reduces it to the fewest steps that still cause it, and writes the exact steps, what it expected, what happened, the environment, and how often. If it will not reproduce, it says so with the frequency instead of quietly dropping it. Severity is proposed from what it does to a customer, not from how dramatic it looks — a small fault on your payment page can outrank a broken feature nobody uses, and it will tell you why.

Data loss, security and privacy do not wait in a queue. Those go to a person immediately, with what is exposed, whether it is still happening, and what would stop it today.

The sign-off is honest. It tells you what was tested, what was not and why, and every defect that is shipping — with what a customer will experience and whether the people who answer your phone have been told. It signs off on what it tested and never on what it did not. If it was not given enough time to test the risky parts, that is the first line of the report rather than a private worry.

It does not fix the code and it does not decide the release. It gives the person deciding an honest picture.