How do you tell the difference between a bad set and a bad room for that set?

I’m trying not to use “weird room” as an excuse every time something dies, but I also don’t want to rewrite decent material because one audience wasn’t into it. If a bit has worked before and then gets nothing, what makes you treat that as useful evidence rather than just one rough night? Do you wait for a pattern across a few gigs, look at whether the same exact section keeps dropping, or change something small and test it again?