Truly novel projects are hard to predict what requirements are actually valuable

Excess work in models?