Schedule risk analysis: Monte Carlo, merge bias and contingency
A schedule risk analysis runs a statistical simulation, usually Monte Carlo, on a sound critical path schedule. It predicts the confidence of meeting the finish date, sizes the schedule contingency needed for a chosen confidence level, and identifies the risks that matter most.
Why the critical path date is optimistic
GAO says that unless a statistical simulation is run, a completion date calculated from schedule logic and duration estimates tends to underestimate the critical path duration.
One reason is merge bias. Where parallel paths merge, the milestone is late if any path is late. GAO's example: two test paths each have a 60 percent chance of finishing on time, so the milestone has only a 36 percent chance (0.6 x 0.6). Each path is probably on time, but the milestone is probably late.
Reading the results
The simulation produces an S curve showing the probability of each finish date. GAO says the results are best viewed as inputs to program management, not forecasts. They show when the project is likely to finish if the team takes no further mitigation, and the high-priority risks they reveal guide that mitigation.
Risk data should come from a quantitative risk assessment, not arbitrary percentages. If the schedule carries more risk than is acceptable, review the activities, dependencies and network. Do not shorten duration estimates just because an earlier date is preferred.
Schedule contingency
Contingency is sized by comparing the deterministic finish date with the simulated date at the chosen confidence level. If the organization wants an 80 percent chance of finishing on time, the contingency is the gap between the 80th-percentile date and the schedule's finish date.
GAO prefers to hold contingency as one activity just before the finish milestone. Spreading it before key milestones makes it less visible, can let it be consumed early, and encourages work toward late dates. It should not be shown as a lag. Contingency is not total float: float comes from network logic, while contingency comes from the risk analysis.
GAO says to rerun the analysis periodically, and also when contingency is being consumed faster than expected or by risks that were not on the register.
Where the exam catches people
Believing that if every merging path is on time the milestone is safe: merge bias means the chance of an on-time milestone is the product of the paths' chances.
Believing the simulated date is a forecast to report: GAO treats it as an input for deciding what to mitigate.
Believing total float is schedule reserve: float comes from network logic, while contingency is sized by a schedule risk analysis.
Believing contingency is best spread across risky activities: GAO prefers one contingency activity just before the finish milestone.
Believing a too-risky schedule is fixed by cutting durations: GAO says to review activities, dependencies and the network instead.