Personally if I was going to stake my professional reputation on a set of results, I'd expand from an exploratory study with a 20% chance of being wrong to a more rigorous study with, say, a 0.1% chance of being wrong.

Of course, that's much easier for people in STEM than it is for social scientists.