When faced with uncertainty about making an economic decision it is usually always best to assess the probability of the various outcomes. A recent problem came about for a project cost estimate where there was uncertainty about the level of accuracy on the estimate and also uncertainty with respect to engineering costs which were a major cost factor for the project. The cost engineer had produced a table of 6 possible outcomes representing the cost associated with each level of uncertainty. Because of the uncertainty it is difficult to determine the proper cost to communicate to management.

Management has to recognize one premise when given a project cost estimate. There is always uncertainty. In fact if things are fair then 50% of the time a project will complete on time and under budget and the other 50% of the time a project will exceed its budget or estimate value. Want to know why project manager turnover is so high. Sometimes project managers are just the recipients of bad luck and held accountable for exceeding a projects estimate even when there was nothing that could have been done differently.

What information should really be important to management with respect to the cost of a project? How many times when given an estimate does that figure lock into a managers mind when in fact there is considerable variability early on in a project about the true cost of the endeavor. These concepts apply to engineering as well as software projects. There are really two important figures that need to be communicated.

1) The expected cost of the project and the accuracy of that point estimate.

2) The risk exposure on the project. What is the cost with a 90% probability that we will be under.

The best way to answer both these questions is by using Monte Carlo simulation methods. Several manufacturers produce software that does a very good job of this. @Risk by Palisades and CrystalBall are two well known examples. It also can be done easily in Microsoft’s Excel with a little knowledge of the randomization functions. This blog entry is not to teach you how but to let you know it’s possible.

The output of a typical simulation looks as above. Here we rolled the dice on 5000 parallel imaginary project universes. As it turned out the median result on this project was P(50) of $97.5M and a P(90) value of $109M dollars.

These two numbers provide a great framework for communicating the cost of a project. The expected cost is $97.5M and we can say we are 90% confident that the project should complete for less than $109M dollars. This gives management a good idea about their financial risk with regard to the capital expenditure.

Management has to recognize one premise when given a project cost estimate. There is always uncertainty. In fact if things are fair then 50% of the time a project will complete on time and under budget and the other 50% of the time a project will exceed its budget or estimate value. Want to know why project manager turnover is so high. Sometimes project managers are just the recipients of bad luck and held accountable for exceeding a projects estimate even when there was nothing that could have been done differently.

What information should really be important to management with respect to the cost of a project? How many times when given an estimate does that figure lock into a managers mind when in fact there is considerable variability early on in a project about the true cost of the endeavor. These concepts apply to engineering as well as software projects. There are really two important figures that need to be communicated.

1) The expected cost of the project and the accuracy of that point estimate.

2) The risk exposure on the project. What is the cost with a 90% probability that we will be under.

The best way to answer both these questions is by using Monte Carlo simulation methods. Several manufacturers produce software that does a very good job of this. @Risk by Palisades and CrystalBall are two well known examples. It also can be done easily in Microsoft’s Excel with a little knowledge of the randomization functions. This blog entry is not to teach you how but to let you know it’s possible.

The output of a typical simulation looks as above. Here we rolled the dice on 5000 parallel imaginary project universes. As it turned out the median result on this project was P(50) of $97.5M and a P(90) value of $109M dollars.

These two numbers provide a great framework for communicating the cost of a project. The expected cost is $97.5M and we can say we are 90% confident that the project should complete for less than $109M dollars. This gives management a good idea about their financial risk with regard to the capital expenditure.