1. Introduction

There are six risk management processes according to the 5^{th} Edition PMBOK® Guide, five in the Planning Process Group and one in the Monitoring and Controlling Process Group. The fourth out of five planning processes is 11.4 Perform Quantitative Risk Analysis. Based on the guidelines set up in the Risk Management Plan (output of process 11.1), risks were identified (process 11.2), and then give a risk rating based on the three factors of a) likelihood of the risk, b) impact of the risk, and in some cases c) the urgency of the risk (how early in the project it is likely to appear). These risks are ranked based on the risk rating from those with high or moderate severity, to those low-level risks which are simply accepted and put on a watch list for further monitoring.

In the process 11.4 Perform Quantitative Risk Analysis, these risks which were qualitatively analyzed are now quantitatively analyzed, which in practical terms means that the probability and impact of these risks are specified, and the product of these is called the expected monetary value, which gives a more precise measure of the impact of a given risk on the product weighted by its likelihood of occurring.

In order to make this transition from quantitative to qualitative analysis, many tools are used in the course of the process. This blog post covers the first of the tools & techniques listed for this process, namely the data gathering*
* of interviewing and the data representation technique of probability distributions.

2. Data Gathering techniques–Interviewing

One of the most common ways of refining cost and schedule estimates is to do a three-point estimate that gives the low, most likely, and high estimate for a particular cost or duration. Let’s take an example: if I enter an address into my iPhone map application, it will give me an estimate for how long it will take to get there. Right now, it’s the middle of the afternoon and it is saying that it will take 45 minutes for me to get to downtown Chicago from the southern suburb where I live. I know that if I tried to go at 8:00 AM, it would take longer because that’s during rush hour. It’s also possible that since the estimate is based on current traffic conditions and the conditions improve on the way to Chicago, that it might even take less time than the original estimate. So the low estimate might be 35 minutes, with the most likely estimate being 45 minutes, and the high estimate being 1 hour 15 minutes. In each case, I used my experience to figure out the assumptions under which these three estimates are derived. The low estimate has to do with the assumption that traffic conditions may change for the better from what they are now, and the high estimate is based on the assumption that I am attempting to make the same trip during rush hour.

On a project, in a similar way you ask experts through the technique of interviewing to give a range of estimates and to list the assumptions behind them, which supports the *credibility* of their analysis. If they have worked on similar projects in the past, their experience and the historical data from those projects support the *reliability* of their analysis.

3. Data Representation Techniques–Probability Distributions

There are two basic types of probability distributions used when discussing risks on a project.

A continuous probability distribution would be more appropriate for representing ranges of values such as cost and duration of an activity. There are several types of continuous probability distribution, one of the most common being the “normal” distribution, what is sometimes called the “bell curve” for its characteristic shape. However, a “triangular” or “beta” distribution which has a larger “tail” on the high end of the curve is probably going to be more appropriate for cost and duration estimates. Just to go back to my example of driving to downtown Chicago, the probability that there will be a delay that increases my driving time is greater than the probability that the traffic will magically clear up along the stretch of road I happen to be driving along and thereby decreasing my driving time. So the distribution of driving times will be skewed towards the high end (i.e., delays), and if I want to reduce the risk of being late to an appointment, I will have to leave earlier to account for that risk.

A discrete distribution would be more appropriate for discussing what will happen if particular events do or do not occur. For example, what would happen if the project manager in a company building a car is requesting to make a prototype before going into mass production? To justify the cost of building such a prototype, a decision tree could be constructed which shows that building such a prototype will reduce the cost of defects that might occur during mass production if the prototype were not done. If the project manager can show that the savings in terms of reduced defects is greater than the cost of making the prototype, then the expense can be justified.

Of course, it is best to find an expert that is not only experienced with the particular types of risks on the project in order to create the data on risks, but who also has experience on what type of data representation would most likely fit those data.

The next post will deal the next set of tools & techniques of quantitative risk analysis, those of modeling techniques such as sensitivity analysis, expected monetary value, and modeling simulation. These tools take the data that was gathered in this first set of tools & techniques and does some “number crunching” to put an actual dollar figure on the amount of impact that a particular risk may have on a given project.

Filed under: Uncategorized |

## Leave a Reply