
Because it saves time, and you won't end up doing all those long forgotten (for good reason) books you can buy for 1 cent.
My criteria has always been that if "enough" people want to read it, it's worth doing. Where "enough" is a pretty small number -- if people collectively spend more time reading it than I spend making it, it's probably a "success." Because I'm not putting in the effort in the first place unless [at least on some masochistic level] I'm enjoying the effort. Surprisingly, to me at least, looking at first 30-day download counts of recent additions it seems like just about everything meets this test. "Weak" performers would seem to be the less-common foreign language books, but I don't think people reading this thread would agree that would be a reason not to "want" them.