The week I almost let a good number hide a bad problem

Quick one from this week: the cafe had its best revenue day in a month. On paper, great news. FounderFlow still flagged it Needs Review because the revenue jump came entirely from a single large catering order, while regular walk-in traffic had actually dropped for the third day running.

If I'd just glanced at the top-line number, I'd have missed that the core, repeatable part of the business was quietly softening under a one-off that happened to look good. That's the exact trap with any single-metric dashboard: a good total can hide a bad trend just as easily as a bad total can hide a good one.

It's made me rethink what I even want an alert to protect against. It's not just "did the number go down," it's "did the thing that usually drives the number change for a reason I'd actually want to know about." Anyone else building monitoring/analytics tools run into the good-number-bad-signal problem?