I do have to agree with Nielsen for once. When you're working on a two month project then there is very limited scope for contextual enquiry/analysis. You've gotta move. So the best way is just to make something, sit people down, ask what they think, and iterate upon that. RAD is rapid application development; a geeky term. Let's have something like QUC - quick user consideration, or DFA - damn fast analysis. As 37 Signals are repeatedly proving, the users come first. Actually, 37 Signals are proving that hype is all important. But users are good too (via).