Something that's kicked around here a lot: fail quickly.

If a project doesn't fit success criteria I give it about a month before it's pivoted or done.