Brian Welcker, writing for his own blog Direct Reports, had a good post about knowing when a project is headed for trouble. The lens for his post is his experience on Microsoft's file system project named WinFS. The best part:
I suspected from early on that the project was doomed to failure. What made me think this? Because when I would ask anyone involved with the project the question "What is it?", I would get a very different answer.
​If your project's goal can't be answered relatively consistently or relatively easily by the people involved in the project, then you're probably in trouble. This is indicative of either having a project that is too big/unfocused/gnarly to be completed successfully, or a project that is too ill-defined for everyone involved to be effective in delivering on its goal. I've been involved in both types of projects, and trust me it isn't pretty.