Metrics also share some problems with automated tests - they can be green for years, they don't test anything useful, and the real problems are elsewhere.
But it's not always clear what is "green"? For example, New Users going up can be a good thing but New Users going up by a lot can represent a fraud attack. Metrics always require some interpretation to provide value which is why we have dashboards instead of alerts.
There are some metrics that are binary good/bad and I agree that in those cases you should just have an alert.
You can thus have thousands but hardly ever will you have thousands all red.
And when you create a new metric, TDD style, the company will not be able / designed / changed to make it green.