For the last twelve years I've been building web applications. During that entire time I haven't ever worked at a software company that hasn't, at one time or another, been concerned with prematurely optimizing the application we've been working on.
I think that we can all learn an important lesson from this study of gym users that was recently written up in the Financial Times (via Paul Kedrosky).
Two Californian academics, Stefano DellaVigna and Ulrike Malmendier, studied three US health clubs and discovered that 80 per cent of members used the gym so infrequently that they would have been better off paying the $10 fee for each individual visit. Many also left substantial gaps between their last visit and cancelling their membership.
This tracks well with my experience with subscription based software products. People inevitably sign up for more access than they need, and the difference between what they pay to use and what they actually use ends up going straight to your bottom line. Optimizing for expected usage never really pays off, because usage never lives up to expectations.
Also, get people on an auto-renew plan:
Some clubs attempt to limit the attrition rate by telephoning inactive members and asking whether there is anything the club can do for them. Mr Ratner never thought this was a good idea as it might remind them that they were wasting their money.
Lesson: People almost always use your software less then they, or you, expect. This is a good thing (for you, perhaps not so much for them).