7-9 Uncertainty in Project Schedules
On some projects, it is easy to estimate durations of activities with confidence. On others, so many uncertainties exist that managers have far less confidence in their ability to accurately estimate. However, project managers still need to tell sponsors and clients how long they believe a project will take and then be held accountable for meeting those dates. One common strategy for handling this potential problem is to construct the best schedule possible and then manage the project very closely. A different strategy is to estimate a range of possible times each individual activity may take and then see what impact that has on the entire schedule. PERT and Monte Carlo are two methods sometimes used for this approach.
7-9a Program Evaluation and Review Technique
Program evaluation and review technique was developed during the 1950s to better understand how variability in the individual activity durations impacts the entire project schedule. To use PERT, a project team starts by sequencing the activities into a network as described above. However, instead of creating one estimate for the time to complete each activity, they would create three estimates: optimistic, most likely, and pessimistic. For example, the first activity, “Determine new product features,” will most likely take five days, but it could take as little as four days if everything works well and as long as 12 days if a variety of things interfere. The person scheduling the project then calculates the estimated time to perform each activity as shown in Exhibit 7.16 using the following equation:
Estimated time =

Optimistic + 4 (Most likely) + Pessimistic

6

Therefore, for the first activity, the estimated time =

4 + 4(5) + 12

6

= 6
The primary advantage of PERT is that it helps everyone realize how much uncertainty exists in the project schedule. When people use single time estimates, sometimes there is a tendency to believe that the estimates foretell exactly what will happen. On many projects, a great deal of uncertainly exists, and PERT helps to make this visible. In addition to making the overall uncertainty more visible, calculations often show that the expected time is actually longer than the most likely time. That is because if many things go very well on an activity, generally only a little time can be saved, but if many things go terribly wrong, a great deal of time can be lost.
EXHIBIT 7.16 PERT TIME ESTIMATE EXAMPLE

ACTIVITY	OPTIMISTIC	MOST LIKELY	PESSIMISTIC	EXPECTED
Determine new product features	4	5	12	6
Acquire prototype materials	16	20	30	21
Produce prototype	8	10	12	10
Design marketing campaign	9	10	14	10.5
Design graphics	6	10	20	11
Conduct marketing	28	30	50	33
Perform sales calls	20	25	30	25

However, using PERT involves difficulties. First, it is often hard enough to create one estimate of how long an activity will take, so it takes even more effort (and therefore money) to create three estimates. Second, there is no guarantee on how good any of the three estimates are. Third, PERT can underestimate the risk of a schedule running long because it does not accurately address when two or more activities both need to be accomplished before a third one can begin.27
Since PERT highlights uncertainty in project duration, its logic is useful to project managers. However, since it has some problems, only a few project managers actually use it to fully calculate and monitor project schedules. Some project managers informally use three time estimates for a few key activities on the critical path to get a sense for the amount of uncertainty and to better understand the activities that need close monitoring. Other project managers who want to understand the potential variation use Monte Carlo simulation. Students of project management need to be aware that both PERT and Monte Carlo simulations are sometimes used to help understand uncertainly in project schedules.
7-9b Monte Carlo Simulation
Monte Carlo simulation is “a process which generates hundreds of thousands of probable performance outcomes based on probability distributions for cost and schedule on individual tasks. The outcomes are then used to generate a probability distribution for the project as a whole.“28 Monte Carlo is more flexible than PERT, in that an entire range of possible time estimates can be used for any activity. The project schedule is calculated many times (perhaps 1,000 or more), and each time the estimate for a particular activity is generated based upon the likelihood of that time as determined by the project manager. For example, suppose a project manager estimated that for a particular activity there was a 10 percent chance of taking five days, a 30 percent chance of taking six days, a 40 percent chance of taking seven days, and the remaining 20 percent chance of taking eight days. Then for each 100 times the computer generated a project schedule, when it came to that activity, 10 times it would choose five days, 30 times it would choose six days, 40 times it would choose seven days, and 20 times it would choose eight days. The output from the computer would include a distribution of how often the project would be expected to take each possible length of time. Many other possible outputs can also be generated from Monte Carlo simulations.
One advantage of Monte Carlo analysis is the flexibility it provides. This allows more realistic estimates. Another advantage is the extent of information it can provide regarding individual activities, the overall project, and different paths through the project that may become critical.
A disadvantage of Monte Carlo is the amount of time necessary to estimate not just a most likely duration for each activity, but an entire range of possible outcomes. Another disadvantage is that special software and skill are necessary to effectively use Monte Carlo. This disadvantage is not as large as it once was because more software is available and most students are learning at least the fundamentals of simulation in statistics or operations classes.
A project manager needs to decide when some of the more specialized techniques are worth the extra effort for a project. The old saying that a person should spend $100 to save $1,000, but should not spend $1,000 to save $100 applies. If the savings on a project from using techniques such as learning curves, PERT, or Monte Carlo are significant, project managers should consider using one of them. If not, they should create the best estimates possible without the specialized techniques, incorporate risk management by carefully identifying and planning for specific risks as discussed in Chapter 10, and manage the project schedule very carefully as discussed in Chapter 14.
EXHIBIT 7.17 NEW PRODUCT DEVELOPMENT SCHEDULE IN CHINA EXAMPLE
Week one—Request is received from the customer for a product that is darker than anything we have in our current offering. Our sales manager forwards the request to our VP sales and our R&D department. A quick review of the potential price versus cost of materials is completed by the VP sales (with finance input), and the product is deemed saleable at an acceptable margin.
Week two—A trial cook in our “baby cooker” is conducted by our R&D department. Within two attempts, a product that is within the customer-requested specs is produced. An additional trial is conducted to quickly check repeatability. The trial product is express shipped to the customer and to our China facility for comparison purposes.
Week three—The formulation and related instructions for cooking are communicated to our China operations with a “red sheet” process. China has anticipated the receipt of this red sheet, and is able to schedule time in production within a week.
Week four—The initial red sheet production is successful and passes the specification tests in China and in Louisville.
Week five—Customer confirms purchase order, and the first shipment is sent. The product contributes significantly to the revenues and profitability of the China facility. Success!
Key factors—Strong communication between all the players and a clear understanding of the customer expectations up front.
Source: Elaine Gravatte, D. D. Williamson.
These specialized techniques are sometimes used in research and development (R&D) projects. However, some R&D projects do not need this level of sophistication. Exhibit 7.17 shows an actual R&D project schedule used by D. D. Williamson of Louisville, Kentucky, when a Chinese customer asked it to develop a new product somewhat different from any it had previously developed. Once D. D. Williamson decided to take the job, it developed and communicated the project schedule to all stakeholders both in its company and the customer’s company within the first week.
Australian researchers have discovered that two primary causes of late delivery of IT projects are variance in time to complete individual work activities and multiple dependencies for some activities. Suggestions for overcoming these two problems are shown in Exhibit 7.18.