Pattern recognition is useful in operational work. It is also dangerous when a plausible pattern turns into certainty too quickly.

I have started liking a simple test:

What evidence would make me abandon this explanation?

If the answer is “nothing”, I am probably no longer investigating.

That matters when diagnosing a process because the fix follows the diagnosis. A confident wrong diagnosis can create a very efficient wrong solution.