There were two things that resonated with people whenever we talked about software schedules. The first was the notion of predictability. Over and over again, managers have lamented to me, "It's the unpredictability I can't live with." As Roscoe points out, it is having results come in wildly at variance with the plan that throws everything off. If people were systematically late, we could deal with that.
This led to the second resonance, the idea of calibration. In order to achieve some measure of predictability, you really do need to calibrate, almost at the individual developer level, what the historical record says about the quality of the estimate. If you can know who is systematically optimistic and who is systematically way off, well, that's information.
My guess is that all good managers do this qualitatively. What Roscoe introduces here is the idea that we should gather data as we go, calibrating and recalibrating our people's estimates all the time. By better understanding how much credence to attach to their estimates, we can perhaps achieve greater predictability in our schedules.
It is interesting how compact this chapter is. I attribute that to Roscoe's ideas' being short and sweet. He proves once again that you don't have to be long-winded to have a good idea.
In the next chapter, I examine a curious phenomenon. All successful projects seem to have an internal rhythm that governs their progress. Where does this come from? I propose a model that perhaps explains the effect.