Truly novel projects are hard to predict what requirements are actually valuable
Excess work in models?