If you actually resolve all ambiguity perfectly, you end up with: The program itself. The most important thing is clarifying the nature of the problem people wa

If you actually resolve all ambiguity perfectly, you end up with: The program itself. The most important thing is clarifying the nature of the problem people want to solve; if you know that much, you can make some kind of headway with minimal churn and burn. Then get feedback, rinse, repeat. I have seen cases - watching one right now, in fact - where mgmt tries to push a no-problem project through, just "well get the data and make some screens and things and whatever" and the only possible outcome is failure.