Using Monte Carlo Simulation in Project Estimates by Mr. Akram Najjar
Mr. Akram Najjar was the speaker for the month of July and delivered the lecture on Monte Carlo Simulation “Using Monte Carlo Simulation in Project Estimates”
Why are single point estimates unreliable and what is the alternative?
What are distributions and how do we extract random samples from them (using Excel)? Two costing examples.
How to setup a Monte Carlo Simulation model in a spreadsheet?
Two PM examples (in detail)
How to statistically analyze the thousands of runs to reach reliable estimates?
A Project Manager usually knows how certain parameters (such as duration, resource rates or quantities) behave. However, the PM can almost never define reliable single point estimates for these parameters. The result: many projects fail due to unreliable estimates. The alternative? The PM has to use his/her knowledge of how specific parameters behave statistically. For example, the PM knows that a specific task’s duration is distributed according to the bell shaped curve OR that another is uniformly distributed (flat variation), or triangular, or Beta-PERT, etc. The PM can then use Monte Carlo Simulation (MCS) to arrive at statistically significant and robust results. Monte Carlo Simulation (MCS) is a technique that relies on two processes. Process 1 aims at developing a spreadsheet model that calculates the critical path or the total cost, etc. The calculation is setup in a single row (or Run). This row is then duplicated a large number of times (thousands). Process 2 aims at inserting Excel functions in each of the parameters (durations, costs). In each row (or Run), such functions will provide a sample drawn from a statistical distribution that properly describes the behavior of that parameter. For example, a specific duration follows a Normal (Bell) distribution with an Average A and a Standard Deviation S. The model will then generate for each run and for that duration a different value that conforms with the bell shaped curve as defined (A and S). Each of these thousands of runs will provide the PM with a different “simulation” of the duration or the total cost, etc. By statistically analyzing the thousands of results, the PM can arrive at a robust and reliable estimate. Proprietary Add On’s for Monte Carlo Simulation in Microsoft Project are available. However, it is easy, free and more flexible to use native Microsoft functions to carry out the full simulation. The talk covered all the steps needed for such simulations giving several examples.