Be it a new menu item in a fast chain or a new pricing scheme for your product lines before rolling out sweeping changes experiment first on a limited scale for a limited time to test your theories. New York City Mayor, Bloomberg has proposed dropping fare prices on certain bus routes all the way to $0. His argument, based on models built by his team, is that most people commute for free now using their transfer passes and there is going to savings by eliminating delays by people searching for exact fares. The NYTimes article points out the problems in the argument, the most significant flaw is the possibility that a large number of people would commute by the busses just because it is free.
Whether or not the models are true or the claims of increased uptake are true need not be debated, they can simply be tested by rolling out a limited experiment for a limited time.
Experiments will go a long way to test your hypotheses and make course corrections, a far cheaper and effective method than managing unexpected consequences from a full-fledged roll-out and rolling back the changes.