Search This Blog

Wednesday, May 31, 2017

Method to create a Risk adjusted estimate - PERT

While estimation discussions come up among project managers or students pursuing project management, many never consider PERT. The discussions revolve around top down, analogous, bottom-up and parametric estimation. While each of these techniques have a place in estimation, I wonder why PERT is never considered. In fact, in one of the chapter meetings organized by PMI MassBay, one of the founders of PMI, Jim Snyder, considered the "Father of the Project Manangemet Institute, expressed how even PMP certifed people cant explain thr basics of estimation techniques like PERT.

I prefer PERT to any estimation technique and only prefer because it builds into a risk adjustment quantitatively. Perhaps I am biased but my bias comes from the fact that the analogous has too much uncertainty and the bottom up estimation drains people time. So, one of the approach that I have taken is to get information using historical data amd expert judjment and extract details for bottom up without spending too much time on the bottom up.  That is, my approach is to get the data required for PERT without asking people. 
For instance, if project archives tell some work took 100 hours when projected estimate was 70, then, I know 70 was a guess at the time it was given. Similarly, if someone said 70 hours but historical projects have taken up 100 hours for that level of effort, I know the 70 hours was the same guess. Now, put them in the normal or bionomial curve. If 100 hours is the regular median, then apply the standard deviation calculations which says rough order of magnitude may go from 25% to 75%. The optimistic is 25% (17.5 less from 70 hours, i.e 52.5). The pessimistic is 75% more than 70 (i.e, 52.5 more from 70, 122.5). Since 70 itself was a guess, I apply a slack based on complexity, unknown assumptions, level of expertise of the person giving, and add 10 to 15% more to 70. So, using 10%, I have optimistic (52.5), optimal (77) and pessimistic (122.5). So, (52.5 + 4*77 + 122.5)/6 = 80.5 PERT.

Therefore, use the statistical ranges amd historical data combined with expert judgment to realistically even out the uncertainties by creating a risk adjusted estimate using PERT.

How do you think you can apply this approach?