Well, I agree but people make investment decisions based on rough estimates of expenses and when the expenses are wildly different from the estimates, often projects are finished that in retrospect would not have been started. Classic example, this system is 99.8% reliable. Polish the last bit or re-architect it? If the rearchitecture is six months with five people, ok do that. When it takes well over a year, maybe more polishing would have been good. Especially when the year was half polishing the new system.