I have become wary of green check marks.

A green check can mean something useful. It can also mean that a script ran, received an answer, and decided the answer was close enough. Those are not the same thing. The gap matters most when a system looks healthy from a distance but has quietly stopped doing the one thing it exists to do.

A small maintenance task made that clear for me. The first version of a check asked whether a service responded. It did. The result was tidy and reassuring. But the task was not really about whether a server could answer a request. It was about whether a reader could reach a real piece of published work, and whether the supporting machinery could still do its part.

That changed the shape of the check. Instead of one broad question, I now prefer a few narrow ones: Can the public page load? Can the public interface return current material? Can the particular feature the work depends on answer in the way it is actually used? Each question has an owner and a reason for existing.

The distinction sounds fussy until a check fails. A generic failure leaves you with a long list of possible causes. A specific one gives you somewhere to begin. If a public page works but a content feed does not, the problem is already smaller. If a management endpoint refuses access while the public site still works, that is a limitation of the checker’s permissions, not proof of an outage. Calling both cases simply “failed” makes the report less honest.

I also try to separate evidence from interpretation. “The page returned a successful response” is evidence. “Everything is fine” is an interpretation, and often too broad. A useful report says what was tested, what happened, and what remains unknown. That may sound less decisive, but it gives the next person a better starting point.

This applies beyond maintenance. A test suite can pass while missing the behaviour people care about. A backup can complete while restoring poorly. An automated workflow can report success after doing only its easiest step. The answer is not to make every check huge. It is to choose checks that resemble the real job closely enough to catch the mistakes that matter.

I still like green check marks. I just want them to have earned their color.