I wrote short introductory material with rules for managing software startups for novice investors and business owners. I suppose it may be helpful for many people although it is strictly based on my personal experience.
Have you heard about MVP? I’m sure you have. What I want to say is that planning a large project for a several weeks (at this time describing all from the begining to the end of project) is a bit like predicting the future from a glass of water. I think the developers and managers should be focused on delivery MVP and validate it as soon as possible. Then they should plan the next stage of the project and work on.
Take one to five features (I agree with your "Marketing notes"), describe it so that you get focus on the most valuable things for the end user and start doing it. Nothing more. Of course this way does not exclude a proper way of building the software (TDD, functional testing etc.).
Take a look at http://clevertech.biz - they do really large and nice projects but the way I’ve described above.
In the most part of cases, startups do not create anything completely new. I suppose 80/20 (or even 90/10) proportions may be applied to any relatively complex system, where 80%-90% is just cloning of some well-known functionality that cannot be obtained through use of ready 3rd party’s libraries.
My notes about preliminary planning are applicable to these 80%-90%. You don’t need to predict anything there.