Proving it, not just knowing it

A hunch that turns out correct is still a hunch

You will often be right early. The problem with being right early is that you cannot tell it apart from being wrong early, and neither can the person you are about to ask to deploy something.

Evidence is what closes that gap. Not more confidence — a specific artefact that would look different if you were wrong.

The test of a piece of evidence is not whether it fits your theory. It is whether it would have failed to fit a different one.