The applications I developed get clicked thousands of times per hour, so everything comes down to efficiency. I like to think of it as a sort of currency with which you can generate lots of useful axioms:
If it saves a click, +1.
If it costs a click, -1.
If the user wastes time figuring out how it works, -1. (Most custom UI elements)
If it is fundamentally intuitive, +1. (Coverflow)
If it saves first-time users a click, +1. (Wizards)
If it costs long-term users a click, -1. (Wizards)
(Thus why you have to make sure your keyboard shortcuts and tab orders make sense.)
Etc.
Etc.
Everything gets subjectively weighted and tallied and you compromise where you have to. Ultimately, it might be a naive philosophy, but its served me fairly well. Extrapolate as you see fit.