[6510]

You look at it again, you again scream inside your head: So, what is the user actually trying to do? Then you hate on your previous solution until you are convinced it is terrible. The only thing it has going for it self is that people are now used to your terrible thing. If you are going to change it it better be dramatically better. You have 100 ideas that should all be considered bad. You prototype a few of them and try to convince yourself you've made it worse. If this fails you try to convince yourself the improvement is not big enough. If this fails you make more prototypes. If you are going to screw with peoples muscle memory you want larger gains out of the mistake.

Then you put it live, if enough people hate it you revert it back to the original version. If more people complaint how the new version was actually better you go back to the drawing board.

If no one cares however, then you've done a good job.

Listen to feedback but don't involve the users outside live testing, they don't have the hate one can project on ones own work. If they could give useful feedback it is something obvious that you convince yourself you should have seen.

edit: Dogfooding is good but mind the learning curve. Dogfooding might actually be required to approach perfection.