Recently, I was thinking about the Program Evaluation and Review Technique (PERT) approach to duration estimation. (Weren’t you too?) According to Timothy J. Kloppenborg in Contemporary Project Management (Third Edition), PERT “was developed during the 1950s to better understand how variability in the individual activity durations impacts the entire project schedule” (p. 190). Stated more simply, it is a way to account for duration risk by using a weighted average formula. That calculation is:
Minimum Duration + 4(Most Likely Duration) + Maximum Duration
Let’s consider a task entitled, “Shipment of Material from London to New York via Barge.” The “Minimum Duration” is the least amount of time the task could take to complete if there were no weather or mechanical delays, and every aspect of the journey went according to plan. Let’s say that’s six days. The “Maximum Duration” is the amount of time the task might take if a variety of things went wrong. In the case of our task, although this duration could theoretically be months in duration – dock worker strike, serious problem with ship equipment etc. — the estimator would be wise not to use an unlikely number. A meteor could destroy all of the shipper’s boats too, but that doesn’t mean we should plan for that situation; the “Maximum Duration” should not account for a cataclysmic event. So let’s use 20 days. The “Most Likely” duration could be an average of the amount of days it’s taken to ship the material in the past, or just a good guesstimate of the time required based on current conditions. In our case, let’s call it 10 days.
Using the calculation above, the task owner would assign a duration of 11 days [(6 + 4(10) +20)/6]. There’s one day of safety included in that duration to account for risk, which is the point of using the PERT. Without it, the task owner would have input the Most Likely duration (10 days) into the schedule.
As a student and supporter of the Critical Chain approach to planning, I’ve read the informative (and entertaining) books of Dr. Eliyahu Goldratt, the business management guru who wrote various novels included The Goal and The Theory of Constraints. Thus, a question I must ask is: does this task require that safety? Not all tasks are created equal – some are more important than others in that they contain more technical, schedule, and/or cost risk. If the shipment of material from New York to London has been ordered well in advance of need, is that safety of one day really required?
You may think the above is a long way to go to dispute the addition of one day to a task’s duration, but consider what would happen if the PERT was used to derive all durations in the schedule. That would apply safety against every task – not all of them need it – and the result is a forecast project Finish date that may well exceed contractual requirements.
What’s a better way to represent duration risk in the schedule? There are a couple of options:
- Apply the PERT calculations to critical path tasks. Safety is used strategically, where the project needs it most.
- Apply the PERT calculations to tasks with a lot of technical risk. Again, safety is applied where most needed.
- Use a simulation software package.
Of course, approach #1 has some problems in that the critical path may change after every status and it’s possible the PERT calculation will have been applied to tasks that are non-critical after the first month of status. There has also been some research showing that PERT is much less effective at modeling duration risk. In a scholarly paper entitled, “PERT Completion Times Revisited,” Fred E. Williams writes, “PERT approximations of project duration – and the associated probability statements derived from those approximations – are, at best suspect and at worst, probably downright misleading.” Bottom line: To avoid the problems associated with PERT, it may be a worthwhile investment of time and money to either create a simulation spreadsheet in MS Excel or purchase one of the readily available packages on the market that specialize in predictive modeling, forecasting, simulation, and optimization.