You make it sound so straightforward and simple. > and then you do the experiments and see if the idea works or not The problem is "if the idea works or not"

You make it sound so straightforward and simple. > and then you do the experiments and see if the idea works or not The problem is "if the idea works or not" can be very fuzzy. With a little p-hacking, anything can work! If nothing in your original data/experiment supports your hypothesis, you can just cherry pick some part of it that gives "significant" results.