Many researchers, most recently Richard Schönberger (49), have exposed some interesting anomalies in PERT/CPM networks which show that deterministic times are optimistically biased. In essence, the effect comes about when one or more paths in the network have times that are close to the critical path time. If the noncritical path is delayed and becomes critical, it may extend the average completion time for the network, as we noted earlier. Schönberger develops a simple example to illustrate this finding. His critique extends to the three-time estimate method, which allows for activity time variance on a path-by-path analysis, but does not consider delays caused by path interaction. (Remember that the ability to calculate path variance as the sum of the variances of individual activities is based on the assumption of independence between activities—and paths.)
Several possible conclusions may be drawn from Schonberger's insight:
• Projects will probably be late—relative to the deterministic critical path.
• Network simulation is probably not worth the added expense.
• Deterministic time estimates should be used in place of the three time esti- ' mates,
• The network developed from these deterministic time estimates should be subjectively reevaluated for any path interaction factors that would tend to make the project late.
• Critical or near-critical activities should be intensively managed, the usual practice of project managers.
While it is helpful to be aware of these issues, we are not entirely in agreement with Schonberger's conclusions. The costs of simulation techniques are decreasing rapidly, and there are several inexpensive simulation computer programs readily available. GERT is an example of a simulation-based technique that is quite valuable if the required information base and computational power are readily available. GERT and other computer programs come in sufficient variety to have a great many applications in project management.
A stronger area of disagreement is our belief that three time estimates are far more informative to the PM than deterministic time estimates. The degree of activity variability is a clear indicator of the need for adaptive planning. In any case, for most project managers, the use of deterministic times does not mean that they estimate the variance of each and every activity to be zero, but rather that they assume that optimistic and pessimistic times are symmetrically distributed around the most likely times and cancel out. The distinction between a deterministic time and an average time is easily ignored by the unsophisticated. Further, this error is compounded because it leads to the false assumption that the expected time approach is the same as the deterministic approach. Clearly, the critical times would be the same in either case, which leads the unwary to the error.
As we illustrated earlier with Lotus 1-2-3®, readily available and inexpensive computer software makes the additional computational cost of three time estimates a trivial matter. Perhaps more serious is the advice to increase the calculated net work time by subjective evaluation of the effects of path interaction. This will extend the network time by some arbitrary amount, usually in the form of network slack, a time-consuming but nonresource-consuming "activity" added to the end of the project, which automatically adds duration to the critical path. (In Chapter 9 we will define a similar activity as a "pseudo-activity" )
This practice makes the operation of Parkinson's Law a clear and present danger. The work done on project elements is almost certain to "expand to fill the additional time," as Schönberger himself and many others have observed. We favor a different way of handling the problem. If three time estimates are used, and if various completion dates with their associated probabilities are calculated, and, finally, if the PM and senior management can agree on a mutually acceptable completion date for the project, the PM can be held accountable for on-time delivery. If an additional allowance is needed for path interdependence, some mutually acceptable network slack can be added. It should not, however, make the PM any less accountable for project performance.
Was this article helpful?
What you need to know about… Project Management Made Easy! Project management consists of more than just a large building project and can encompass small projects as well. No matter what the size of your project, you need to have some sort of project management. How you manage your project has everything to do with its outcome.