One of the most important ideas behind Lean UX is that we don’t assume that a new design or feature is automatically better than what came before it. We test everything. When we ship a new feature, we test to see if it made a significant impact on important user behaviors. When we change text or the timing of an email, we test to see which version performs better. When we make a change to a user flow or a navigational structure, we make sure we haven’t inadvertently made things worse for users.
It's kind of presumptuous to do things any other way, no?