Conducting a Schedule Risk Analysis
Schedule risk analysis relies on statistical simulation to randomly vary the following:
activity durations according to their probability distribution;
threats and opportunities according to their probability and the distribution of their effect on affected activities if they were to occur; and
the existence of a risk or probabilistic branch.
The objective of the simulation is to develop a probability distribution of possible completion dates that reflect the program plan (represented by the schedule) and its quantified uncertainties and risks. From the cumulative probability distribution, the organization can match a date to its degree of risk tolerance.32 For instance, an organization might want to adopt a program completion date that provides a 70 percent probability that it will finish on or before that date, leaving a 30 percent probability that it will overrun, given the schedule and the risks as they are known and calibrated. The organization can thus adopt a plan and promise completion on dates that are consistent with its preferred level of confidence in the overall integrated schedule. A schedule risk analysis can provide valuable information to senior decision makers, as shown in case study 15.
Risk analysis should not be focused solely on the deterministic critical path—that is, the critical path as defined by the initial or current set of inputs in the schedule model. Because the durations of activities are uncertain, with risk considered, any activity may potentially affect the program’s completion date. Hence, the path that is most likely to determine the finish date is uncertain.
If the analysis is to be valid, the program must have a good schedule network that clearly identifies the work that is to be done and the relationships between detailed activities. The schedule should be based on a minimum number of justified date constraints. It is important to represent all work in the schedule, because any activity can become critical under some circumstances. Complete and correct schedule logic that addresses the logical relationships between predecessor and successor activities is also important. The analyst needs to be confident that the schedule will automatically calculate the correct dates and critical paths when the activity durations change, as they do thousands of times during a simulation.
If time or resources are insufficient to simulate the full program, or if detail in the future is unclear, perhaps because of rolling wave planning, the simulation can be performed with a summary version of the schedule. The summary schedule is a condensed form of the schedule that rolls detail activities into long-duration activities. By reducing the number of activities in the schedule, analysts reduce the time spent collecting data about and assigning risks and probability distributions to detail activities.
However, if a summary schedule is used for a schedule risk analysis, it is important that the schedule show enough detail to yield practical results. A summary schedule that is condensed too much will not convey the effort in very long activities, the activities that should have assigned risks, or how total float is distributed among key activities and milestones. For example, activities in the summary version of the schedule should show critical hand-offs. If an activity is 4 months long but a critical hand-off is expected halfway through, the activity should be broken down into separate 2-month activities that logically link the hand-off between activities. Finally, condensing the schedule may hide merging paths. As discussed in the previous section, merging paths are the source of much risk.
GAO performed a schedule risk analysis on the construction schedule for Phase IV of the Department of Veterans Affairs’ (VA) new Medical Center Complex in Las Vegas, Nevada. The project executive identified 22 different risks in an exercise preliminary to this analysis. Using these risks as a basis for discussion, GAO interviewed 14 experts familiar with the project, including VA resident engineers, general contractor officials, and architect and engineering consultants.
In these interviews, GAO identified 11 additional risks. During data analysis, some risks were consolidated with others and some were eliminated because data were too few. Finally, 20 risks were incorporated into the Monte Carlo simulation. They included 18 risk drivers, 1 schedule duration risk, and 1 overall system commissioning activity that was not included in the baseline schedule.
The schedule duration risk was applied to each activity duration to represent the inherent variability of project activities and inaccuracy of scheduling. Of the 6,098 activities in the schedule, GAO assigned risk drivers to 3,193. Some activities had one or two risks assigned, but some had as many as seven.
Beyond applying 20 risks to the schedule, GAO was interested in identifying the marginal effect of each risk. Therefore, GAO identified the risks that had the greatest effect on the schedule, because they should have been targeted first for mitigation. Marginal effect translates directly to potential calendar days saved if the risk is mitigated.
GAO’s analysis of the medical center construction schedule concluded that VA should have realistically expected VA’s acceptance between March 1, 2012, and May 17, 2012, the 50th and 80th percentiles. It was determined that the must-finish date of August 29, 2011, was very unlikely. The analysis showed that the probability of achieving VA’s acceptance by October 20, 2011, was less than 1 percent, given the current schedule without risk mitigation.
VA’s actual acceptance was December 14, 2011, approximately 4 months later than had originally been expected. Delays stemmed from issues with steel fabrication and erection, as well as changes to equipment requirements. At the time of GAO’s original analysis, December 14, 2011, fell within the 5th to 10th percentiles.After the risk information is developed, the statistical simulation is run and the resulting cumulative distribution curve, the S curve, displays the probability associated with the range of program completion dates. The results of risk analysis are best viewed as inputs to program management rather than as forecasts of how the program will be completed. The results indicate when the program is likely to finish without the program team’s taking additional risk mitigation steps. The high-priority risks can be identified and used to guide further risk mitigation action.
A schedule risk analysis may show that a given schedule has more risk than is acceptable. In this case, a review of the activities, dependencies, and network might help derive a shorter schedule. In some cases, the scope may need to be reduced. However, the initial estimates of effort and duration should not be changed without sufficient justification. Changing durations simply because an earlier finish date is preferred is likely to increase the risk of delaying a project.
A cumulative distribution sums all the probabilities of values less than or equal to the value of interest. The cumulative probability increases from 0 to 1 as the value of interest increases. Hence, a selected finish date from the cumulative probability distribution represents the probability of finishing on that date or earlier.↩︎