Project Management Institute

Risk management for dummies, part 2

three computer-based approaches to schedule risk analysis



Harvey A. Levine

You can't analyze all the risk out of a project—but you can try! These methods and tools will arm you with the information you need.

In last October's column, I discussed several non-statistical approaches toward schedule and cost risk avoidance and management. These included practical, non-mathematical, common-sense approaches, such as time contingency, earned value analysis, and management reserve (cost contingency). I invited (and received) reader feedback, which confirmed both the importance of risk management and support for our pragmatic approaches.

My tongue-in-cheek introduction brought out a few defenders of traditional statistical methods. David Hulett noted that my claim of a 50 percent probability for traditional CPM calculations was highly optimistic. This was fully borne out by my experiments with the risk analysis software, discussed in this month's column. If fact, when run through the Risk+ for Projects program, the CPM calculated finish in our test model was shown to have less than a 5 percent probability. Readers are advised to review Mr. Hulett's article on Risk Management, published in the March 1995 issue of the Project Management Journal. Also look for a down-to-earth treatment of risk analysis (using Risk+ for Project) by David Hulett in an upcoming issue of PM Network.

Further support for Risk Management came from the Angel Group, a Dallas, Texas, firm specializing in project control and risk management methodologies for IS and high-tech manufacturing. They report continued success in helping firms to set realistic schedule and cost targets and then to meet those targets. One of their tools is a home-grown methodology called “Risk Averse Project Management Process.” The process centers on improving estimates and “having the good common sense to create Managed Contingencies and Risk Reserves”—just what I talked about in October.

In this month's column I will look at three simple, computer-based methods of working with schedule risk.

All three methods are based on what is often called the “PERT” method. This concept relies on three time estimates per task, rather than a single estimate. The latter two examples (both risk management programs) allow the user to either specify discrete values for each of the three estimates or to apply a defined formula for the optimistic and pessimistic values. While calling this the PERT approach may not be technically correct, I will yield to this popular label for the various three-time-estimate methodologies.

The Three-Time-Estimate (PERT) Approach

You won't find the three-time-estimate approach to be in great demand. After all, if we have such a terrible time arriving at a reasonable single time estimate, won't the PERT approach just give us a very precise error? This is certainly possible, and we have to evaluate the justification for either estimating mode on a case basis. Let's look at some of the advantages and disadvantages of each mode.

First of all, project managers seem to agree that the most common weakness of project schedules is the task estimates. We have trouble estimating the duration of tasks, as well as the effort required to execute the tasks. There are volumes of writings on the problems of task estimating, and there would be considerably more published on the subject if anyone had any really good solutions. Given the softness in our base estimates, what do we gain from the triple-estimate approach?

Figure 1. Project Scheduler 6 showing added PERT Duration column and Risk Analysis dialog box.

Project Scheduler 6 showing added PERT Duration column and Risk Analysis dialog box

First, we are more likely to gain precision in the time estimates. When we ask a performer to estimate the duration of the task, we often get a biased estimate. The performer may be overly optimistic, assuming that everything will go well (Murphy is on someone else's job). Or the performer may be afraid to make a commitment based on a best guess, so a little time is added as a safety factor. So just what does ten days mean? Is it ten days if everything goes well, but more likely to be 13 days? Or is it most likely to be eight days, but we'll add a couple of days as a cushion? With the PERT approach, we can ask for three distinct time estimates. An optimistic estimate is usually a duration that would be achievable about 10 percent of the time. Likewise the pessimistic estimate is usually a duration that would occur about 10 percent of the time. The third estimate is the most likely, which we are now able to obtain without deliberate bias. The traditional PERT formula for calculating task durations is A+4B+C over 6, where A is the pessimistic, B is the most likely, and C is the optimistic. But there are other options, as we will soon see.

Other advantages are: (1) we gain a range of task and project durations, (2) we can adjust weight factors to generate schedules with higher or lower confidence factors, and (3) we can evaluate the potential for achieving any selected project end date. We also expand the capability for performing “what if” analyses. We can use this increased information about durations in our analyses of the schedule, whether performed by simple observation or via computerized probability analysis.

It's time to look at our three computer-based approaches to duration analysis. In my October column, I argued that traditional CPM schedules produced a project end date calculation that could have a 50-percent-or-less possibility of being met. These programs address this issue, in varying degrees of sophistication. They are easy to understand, even if you are Sigmaphobic. But ease-of-learning and ease-of-use do increase with the level of sophistication. I don't necessarily recommend them for everyone, or every project. But when meeting a schedule date is important, and especially when there are dire consequences from missing schedule deadlines, these three programs will generate better estimates and an understanding of the potential (or improved confidence) for achieving the end dates.

Using Project Scheduler 6

Example one features a general purpose CPM program that also offers support for the three-time-estimate (PERT) approach. Several products have a PERT capability, but we will look at Project Scheduler 6, from Scitor, because of some of its special PERT features.

In PS6, we can activate the PERT mode and add a PERT DUR column to task the table (Figure 1). We can enter the three time estimates, and (in a special PS6 feature) adjust the weighting factors. In Figure 1, we can see that (using the traditional 1/4/1 weighting) the calculated project duration is 21.33 days (vs. the 20 days using the single estimates).

By adjusting the weight factors we can calculate various degrees of optimism or confidence. Using a 1/0/0 weighting we calculate the project using only the optimistic durations. In this case we can say that the best possible project end date is 10/16/95, or 16 days. By changing to a 0/0/1 weighting (all pessimistic) we can project a worst case of 11/1/95, or 32 days.

Another option is to deliberately add time contingency by giving more weight to the pessimistic dates. In the model above, setting the weight factors to 1/4/5 will generate a 25.6-day schedule. While PS6 has this unusual capability to select the weighting factors, it does not provide support for statistical analysis and probability of various end dates (normally expected in a PERT-type program). Our next two examples do.

Using Risk+ for Microsoft Project

When the PERT approach was first developed in the late ’50s, it was customary to perform a statistical analysis (Whoops! … here come the Sigmas) to calculate the confidence factor for each potential project completion date. Today, there are several add-on programs that work with popular PC-based CPM programs. We'll look at two of them here, starting with Risk+ for Project, from Program Management Solutions Inc. Most of what Risk+ does is virtually transparent to the user, and no expertise in statistics is needed to operate the program or to understand the results.

Risk+ attaches itself to Microsoft Project when installed. Then, when you boot up MS Project, the Risk+ functions are directly available from the toolbar. A Risk Gantt View is automatically added to the View Menu. Users can enter their own (triple) estimates, or can have the system spread the optimistic and pessimistic values based on a set of user-defined preferences. Where Risk+ departs from the capabilities of PS6 is in the ability to compute the probability of meeting any schedule date. This program (and Monte Carlo, our next example) do not try to average the three time estimates (as in the prior example) but rather make several passes through the network, using random selection of the (defined) possible task durations.

When we run the Risk Analysis routine in Risk+ (selecting 1,000 iterations), the results are similar (actually slightly more conservative) to those with PS6, except that, with just a single analysis run, we obtain a complete set of dates and probabilities. In Figure 2, we can see that there is less than a 10 percent chance to achieve the CPM completion date of 10/20. We have calculated a 50 percent probability of meeting 10/23, and a 90 percent probability of meeting 10/26. If we really wanted to play it safe (100 percent probability) we would not want to commit to anything earlier than 10/30.

A second capability, not handled by simple PERT solutions, is the problem of merge bias. It may not be readily apparent that the probability of a schedule being met will be adversely affected by the number of paths that converge at a single point. At least, this phenomenon is not taken into consideration in traditional CPM calculations. But merge bias will have considerable impact on project end date confidence.

For instance, in the example above (Figure 3), we extend the durations of tasks C and E so that there are three equal paths. Each of the tasks might have a 50 percent probability, and each path might have a 50 percent probability. But it just takes a delay in any of the tasks, in any of the paths, to delay the project end date. When we run the risk analysis on the model with the three equal paths, we find that the 50 percent probability date has moved out to 10/25, and the 100 percent date to 11/4. In fact, the risk analysis says that we have less than a 5 percent chance of meeting a 10/22 date, which is two days later than the simple CPM project calculation. Would you want to bet the store on 10/20?

Figure 2. Microsoft Project with Risk+ showing Risk Analysis Gantt view and risk analysis results.

Microsoft Project with Risk+ showing Risk Analysis Gantt view and risk analysis results

Figure 3. Microsoft Project with Risk+ showing Risk Analysis Gantt view and risk analysis results. Note relative criticality figures for each path in the Gantt.

Microsoft Project with Risk+ showing Risk Analysis Gantt view and risk analysis results. Note relative criticality figures for each path in the Gantt

Note also that our bar chart displays the degree of criticality of each path. A simple bar chart would indicate that all three paths are equally critical (because they all have zero float). The Risk Analysis bar chart indicates that the upper path was critical for 68 percent of the iterations and the lower path for 30 percent. (In our earlier illustration, Figure 2, the lower path was shown to be critical 98 percent of the time. Yet, we didn't change the duration of that path or the project.) With Risk Analysis, we gain both greater sensitivity to the effect of converging paths, and additional information to direct our attention to the tasks that are at the greatest risk to extend the project.

Risk+ can also perform risk analysis for costs as well as time. This column is deliberately limited to schedule analysis.

Using Monte Carlo

Our next risk management software example is Monte Carlo, from Primavera Systems, Inc. Monte Carlo sits atop Primavera Project Planner (P3). When Monte Carlo is loaded, it can be accessed from the P3 menu. You can use Monte Carlo on any project file saved in the P3 format (from P3 for Windows, SureTrak for Windows, or P3 for DOS). As with Risk+, Monte Carlo can analyze both schedule and cost risk, and generate date and cost probability graphs. Where Monte Carlo goes beyond Risk+ is in its ability to handle probabilistic and conditional branching and several other statistical functions. You can skip that part, if you're Sigmaphobic. I had to read the manual several times before the process sank in. The basics are not really that complex, but many of the advanced features can challenge the statistically disadvantaged.

Basic Risk Analysis in Monte Carlo. Basic schedule and cost risk analysis in Monte Carlo is similar to that in Risk+. Click on Monte Carlo and you get a chance to specify risk parameters and number of iterations. Initial results are shown in Primavera Look (an output reader). Monte Carlo employs ReportSmith (bundled with P3) for detailed reporting. Several predesigned reports are furnished.

Advanced Analysis in Monte Carlo. Up to now we have looked at routines for analyzing schedules based on a fixed work flow. We defined potential variables for the time of each task. However, all tasks were assumed to be performed, and in the defined sequence.

A frequent planning concern is the scheduling of alternative strategies. Monte Carlo deals with this in two ways. The first is Probabilistic Branching. You would use Probabilistic Branching when you want to define the percentage likelihood that a specific activity will occur after another activity.

Conditional Branching goes one step further, adding a condition that will determine which branch should be taken. You would use Conditional Branching when the choice of the successor task is tied to the status of a defined “condition” task.

Another feature of Monte Carlo, applicable to all modes, is the ability to define Probabilistic Calendars. Here, you can select dates that might not be available for work, and specify the probability of that unavailability. Monte Carlo also has a provision for defining three resource usage values (Opti/ML/Pess), extending the risk analysis into the resource usage and cost area.

While Monte Carlo offers several capabilities for advanced risk analysis, it raises the learning curve for both understanding the technology and for following a set of involved routines.

Working with probabilistic and conditional relationships might not be for everyone or every project. Yet, when there is significant sensitivity to missing dates and budgets (high risk), it's nice to know that there are simple, PC-based software products available to facilitate the analysis of such risks.

Other Products and Resources

The three products used here to describe three software-based methods of schedule risk analysis are not the only ones available to perform such tasks. Readers seeking information on such products might also consider MS Project (Microsoft), which can operate on PERT calculations, using macros and/or Excel. SuperProject (Computer Associates) provides actual probability analysis, based on three time estimates and standard deviation.

Palisade Corporation's @Risk is an alternate to Risk+ for Project. Welcom Software Technology offers OPERA, which works with their Open Plan (DOS version). The next release (Summer 1996) of Open Plan Professional for Windows, from WST, will have OPERA embedded in the base product.

Consulting, mentoring and training in various aspects of risk management and avoidance are available from numerous sources, including most risk analysis software providers.


Risk management requires a proactive culture. You cannot afford to wait until something bad happens to first consider a response. It's usually too late to prevent serious damage by then. You must evaluate and anticipate possible failure and plan for alternatives in advance. You must evaluate the “windows of opportunity” for alternatives and check for satisfactory performance while there is still time to act (or while shifting strategies is still economically viable and within an acceptable time frame).

In last October's column, I discussed several non-statistical approaches toward schedule and cost risk avoidance and management. Those, and the simple, program-based statistical methods discussed this month, will arm the project manager with important information needed to support this proactive activity. Obviously, it is not practical to consider every possible problem that could adversely effect your project. Risk analysis helps you to focus on those areas that have the highest sensitivity to things not going as planned. When schedule and cost risk is an issue, the prudent project manager will take advantage of methods and tools such as these. ▄

Harvey A. Levine, principal, The Project Knowledge Group, Saratoga Springs, New York, has been a practitioner of project management for over 30 years and is a past chairman of the Project Management Institute.

This material has been reproduced with the permission of the copyright owner. Unauthorized reproduction of this material is strictly prohibited. For permission to reproduce this material, please contact PMI.

PM Network • April 1996



Related Content

  • Project Management Journal

    Team Collective Intelligence in Dynamically Complex Projects member content locked

    By Hansen, Morten Juel | Vaagen, Hajnalka | van Oorschot, Kim In concurrent engineering projects driven by short delivery times, team performance rests on the team’s capability to quickly and effectively handle different, emergent issues. We conducted an…

  • PM Network

    High-Wire Act

    By Parsi, Novid Seven years ago, government leaders in Alberta, Canada vowed to take a major step toward addressing the energy needs of the growing province. The resulting transmission line, the longest of its kind…

  • PMI White Papers

    The Impact of the COVID-19 Crisis on Project Business

    By Project Management Institute | Project Business Foundation This is a report on the results of a survey jointly conducted by PMI and the Project Business Foundation. The intention was to replace observations and opinions with reliable data. The mission was…

  • PM Network

    Árboles de la vida

    By Hendershot, Steve El mundo necesita más árboles, muchos más árboles, para mitigar el daño causado por la deforestación masiva. Brasil está destruyendo el equivalente a tres canchas de fútbol por minuto en la selva…

  • PM Network

    Riesgos elevados

    By Nilsson, Ryan Desde que los humanos han construido ciudades, han migrado hacia las costas en busca de alimento, facilidad de transporte y un sinnúmero de beneficios ecológicos.