Every time I've had a cost plan prepared, by a quantity surveyor, or sometimes an estimator, but I expect more from QSs, after all, I went to Uni with some of them, I don't fail to be disappointed.
The cost plan process starts early in the design cycle, typically, and to accommodate uncertainty, there are usually contingency allowances made: a 'design contingency', sometimes a 'project contingency' if the project scope is in flux, and a 'general contingency' to allow for architectural fantasies and client wishful thinking.
The contingencies are applied in a seemingly rote fashion: 10% here, 5% there, and so on.
I once asked a QS if he reviewed completed projects to see how the contingency allowances bore up to the real world. The answer was a staggering "no"! Here we are, with a great source of data to allow good calibration of many building and project types, and it's let blow in the wind.
The other thing I notice with project cost plans is that the estimate tends to increase as the project advances and more detail becomes available.
But, I think that it should go the other way. In the early days of the project, the margin for uncertainty should be quite high (worked out by the statistical history of similar projects), and as more detail is produced, it should drop, with the margins for uncertainty assigned quite clearly to particular elements of established risk. Then we'd be getting what we paid for!