Probabilistic Branching
In addition to standard schedule risk and sensitivity analysis, typical events in programs require adding some new activities to the schedule. This is called “probabilistic branching.” One common event is the completion of a test of an integrated product (for example, a software program or satellite). A schedule often assumes that tests are successful, whereas experience indicates that tests may fail and that their failure will require the activities of root cause analysis, plan for recovery, execution of recovery, and retest. This is a branch that happens only with some probability.35
In the house construction example, the SRA accounts for two scenarios that could occur after owner walkthrough. The plan assumes that in 70 percent of the cases, deficiencies identified during walkthrough can be addressed by the general contractor within a work week. However, in 30 percent of the cases, the owner and the general contractor dispute the deficiencies for 15 working days, and, unable to resolve their differences, enter mediation for 30 working days. The Gantt chart in figure 42 shows the probabilistic branch associated with owner walkthrough for this example.
Figure 42: Probabilistic Branching in a Schedule
In figure 42, “deficiencies are corrected” will occur 70 percent of the time, resulting in no delay for owner acceptance. In 30 percent of cases, “deficiencies are disputed” occurs, leading to the successor activity “dispute is mediated.” This results in a delay of 40 working days to owner acceptance and, ultimately, in a 40-working day delay to owner occupation. The resulting probability distribution of dates for the entire project can be depicted as in figure 43, as applied to the 3-point duration risk simulation.
Notice the bimodal distribution with the corrected deficiencies scenario on the left of figure 43 and the dispute scenario on the right. In this case, if the homeowner demanded an 80th percentile schedule, it would be April 15.36
Figure 43: Probability Distribution Results for Probabilistic Branching
Probabilistic branching analyses may need to be conducted on a copy of the IMS file if the IMS is baselined or represents only required scope.↩︎
Probabilistic branching is used to model the random choice between two alternatives. An advanced technique known as “conditional branching” is also available in certain SRA software packages. With conditional branching, an action is determined by some scheduled event rather than by randomness. That is, it is modeled as an “If…then…else” statement rather than a probability of occurrence. For example, if a design activity takes 2 weeks, then execute Plan A, otherwise (else) execute Plan B.↩︎