Project

I was writing a note to someone about the use of PERT and Monte Carlo simulation in Microsoft Project and it made me think about how willing people are to accept simple models.

As Box and Draper said "All models are wrong; the practical question is how wrong do they have to be to not be useful*" (frequently paraphrased as "All models are wrong. Some models are useful"). The margin of error in almost any schedule is going to be in the range of several percent. Building a probabilistic model of the schedule will help give an idea of the range, but even that sort of model has a margin of error.

My point here is to not get carried away with increased precision in the model beyond the point where it exceeds the limit of accuracy**. Schedules are probably only accurate to 2 significant figures so there is no need for five digits beyond the decimal point for % complete.

The biggest holes in the schedule are not going to be from rounding, but rather from things which were not considered or hand-offs which were not properly made. Modeling those events is difficult if not impossible, the answer is to avoid them in execution through proper planning and keeping track of dependencies.

* Ref: George Box and Norman Draper, Empirical Model Building and Response Surfaces, John Wiley, 1987, pg. 74

**Accuracy refers to how close a measurement is to the "true" value. Precision is how finely we can discern differences between values. You can have a highly precise measuring device which is always inaccurate. In a pinch always choose accuracy over precision.