Cost is always the consideration in making decisions for systems development. How much will it cost to provide the needed capabilities? Quantifying the cost estimate and the economic risks associated with the development effort is core to any business decision to proceed with the development.

Here you can see how to apply probability to modeling costs, measuring and managing risks associated with project cost. The fundamentals of probability theory, using examples and case studies is provided.

Garvey is Chief Scientist for Economic and Decision Analysis Center, MITRE Corporation.

This book must be on your shelf and you must have read it if you're going to have any chance of making credible estimates on projects or any management endevor. Huff shows how to *cook the books* with statistics. The most common in our project management world is selection bias for sampling. *I know a guy that ...* Really what the sample population. This is the basis of the bogus statistics used by Standish to show that IT project fail a lot. They have a *bias selection* since *ALL* the IT projects on the planet were not sampled.

The second way is to *roll up* the variances from underly process to hide them. These is the *most likley temperature* in Trinadad and Cody Wy are nearly the same. So without knowing the variance, those numbers are meaningless

This is a good second level book about statistics. It has solid mathematics but without the details in more advanced books. Polls, surveys, economics data, cost forecasts, planned completion dates are all random numbers drawn from some underlying probability distribution. That distribution is created by the statistical processes. Projects have statistical processes that drive them. When a *BAD* manager askes for a commitment date and doesn't ask for a confidence interval around that date that's a problem. No single point estimate can be credible without the variance on that number.

And by the way, when it is heard that *you can't estimate software projects* it is utter nonsense. You can estimate anything on the planet. You just have to include the confidence level of that estimate. This is another example of why we need to read and become informed about probability and statistics.

Where someone says *we can't estimate,* or *we don't want to estimate*, or better *I don't know how how to estimate* read this book first. It shows you how to estimate anything from the size the of mulch for your garden to the number of hours needed to deliver a piece of software.

We make estimates all the time, *how long will it take to get to the airport*?* How long will it take us to reach the summit of Long Peaks with our hiking trip*?

The process of estimating a software deliverable is simple. Here's how. Learn to do it.

When there is talk of *making decisions* in the absence of estimating, remember you can't assign a value to something unless you know its cost. You can't get a Return on Investment without knowing its cost. You can know the cost in the future without some means of estimating that cost in a probabilistic manner. You can produce the probabilistic forecast of a cost without knowing something about the underlying statistics of the *drivers* of the cost.

In the end you've go to be able to do the math for non-trivial project cost management. Small team, small cost, small duration no one cares. Big project, high risk, high reward, mission critical, the math is important.