What Mitch Kapor writes in the context of Chandler is relevant for us also:
Implementation must be sequenced: It’s not a real project until commitments are made to defer some capabilities. Doing everything at once is not an option.
First, provide core functionality users expect. In order for the product to be adopted, it must have the critical mass of features people expect.
A few killer features are needed from the outset to make adoption worthwhile. Killer feature ideas must be sufficiently well-developed as ideas to be willing to place bets on them.