I would always have a playable prototype as the first part of development.
All papers, essays, concepts are crap, if the team can not
make a first running version.
Developing story, levels, grafix should come after! a prototype.

If the prototype is no fun to play, it might be better to cancel or restructure the
project, before spending much efforts in a complete campaign and custom grafix.


Aslo very bad: if the prototype (with much less content than the finished game)
runs slow already.
Games get slower, not faster in the later stages.