The title says "Stop wasting time getting estimates right", then outlines a number of classic ways in which estimates can be improved: give a range, collect data, break down estimates (which is problematic by the way[0]).

The process described actually sounds a lot like the Personal Software Process that Watts Humphrey described in the 1990s, right down to the use of proxy values for estimation purposes.

[0] obligatory self-promoting link -- http://confidest.com/articles/how-accurate-was-that-estimate...