CRITICISMS OF NETWORK METHODS
Network methods as described in this chapter have been criticized because they
incorporate assumptions and yield results that are sometimes unrealistic. For example, the
methods assume that a project can be completely defined upfront in terms of identifiable
activities with known precedence relationships. In many projects, however, not all work can be
anticipated, and not all activities can be clearly defined at the start. Rather, the project
“evolves” as it progresses. But this problem actually relates to scope planning, scope definition,
and work definition, not scheduling. A related problem is that the schedules sometimes require
regular modification of activities and timelines; this happens when there are too many activities
in the network or the activities are not well defined. The problem can be addressed by initially
creating only a rough schedule, then developing more-detailed schedules in a phased approach
as discussed in Chapter 4, and by avoiding “proliferation” of activities, i.e., keeping the number
of activities in the plan to the essential minimum as prescribed in the work definition guidelines
in Chapter 5. Another criticism relates to the fact that in real projects it is sometimes difficult
to demarcate one activity from the next, and the point of separation is more or less arbitrary.
This means that successors can sometimes be started before predecessors are finished,
and the two “overlap” in the sequence. But again, this is not really a problem. PDM allows for
overlap of activities, and hand-over points treat activities as if they did overlap. A further
criticism is that precedence relationships are not always fixed, and that the start of an activity
may be contingent upon the outcome of an earlier one that might have to be repeated. The
results of a test activity, for example, may require redoing the prior activities of analysis and
design, which in the network would be a “loop back” from the test activities to the activities
that preceded it. The GERT method discussed in the next chapter deals somewhat with this
inadequacy. In summary, the shortcomings of networks are actually shortcoming in any project
planning scheme. It can be argued (and innumerable project managers will attest) that the
methods, though not perfect, offer a good approach for analyzing and creating project
schedules.
AOA Diagrams Besides
AON, the most common method for diagramming networks is the activity-onarrow
(AOA) or arrow diagramming technique. The major feature that distinguishes AOA from AON
is the way activities and events are denoted. Figure 6-30 shows the AOA representation for
one activity and its events. Notice that in the AOA method the activity is represented as a
directed line segment (called an arrow or arc) between two nodes (or circles). As shown in
Figure 6-30, the nodes represent the start and finish events for an activity, and the arrow
between them represents the activity. The number inside each node merely identifies the
event. The number need not be in any particular sequence, however it must be unique for the
event (each event must have its own number). The direction of the arrow indicates the flow of
time in performing the activity, but like AON the length of the line has no significance (unlike
Gantt charts where it is proportional to the activity duration). The number over the line is the
activity duration. As in AON networks, an AOA network should have only one origin event and
one terminal event. All arrows must progress toward the right end of the network and there
can be no doubling back or loops. As with the AON method, the activities follow the order of
precedence as defined by their immediate predecessors. When an activity has more than one
immediate predecessor, the network must show that it cannot be started until all of its
immediate predecessors have been completed. This is the purpose of a special kind of activity
called a dummy
Near-Critical Paths
The PERT procedure has been criticized for providing overly optimistic results, a
criticism that is well justified since it does not account for the effect of mergepoint bias.6
Notice in the example in Figure 7-12 that two paths are “near critical” in length. The variance
of these paths is large enough that either could easily become critical by exceeding the 29 days
of the original critical path. In fact, as you may wish to verify using the statistical procedure
described previously, the probability of not completing Path A and Path E within 29 days is 33
and 28 percent, respectively. So there is more than a slight chance that these paths could
become critical. The warning is: Putting too much emphasis on the critical path can lead to
ignoring other paths that are near critical in length, paths that could themselves easily become
critical and jeopardize the project completion date. Furthermore, the 50 percent probability of
completing the project within 29 days (as presumed with the normal distribution) is overly
optimistic. Because all activities in the network must be completed before the project is
finished, the probability of completing the project within 29 days is the same as the probability
of completing all five paths within 29 days. Although the probability of completing Paths B and
D within 29 days is close to 100 percent, the probabilities of completing Paths A and E within
that time is 67 and 72 percent, respectively, and the probability of completing C, the critical
path, is only 50 percent. So the chance of completing all paths within 29 days is the product of
the probabilities, (1.0 1.0 0.67 0.72 0.5), or less than 25 percent.
