A standard P6 or MS Project schedule produces a single finish date, calculated from single-point activity durations. This deterministic calculation ignores a well-known statistical fact: when several paths in an activity network converge on the same milestone, the probability that the milestone is met on the calculated date is almost always below 50% — a phenomenon known as "merge bias."

The principle: replace a fixed duration with a range

Monte Carlo simulation replaces, for each activity (or at minimum for critical and near-critical path activities), a single duration with a three-point estimate: optimistic, most likely, and pessimistic. The software (Primavera Risk Analysis / Pertmaster, Safran Risk, or add-ins like @Risk) then runs thousands of schedule iterations, randomly drawing a duration from each range on every iteration, and records the resulting finish date each time.

The result is no longer a date but a probability distribution: you can then read directly the date corresponding to a 50%, 80% or 90% confidence level — P80 often being the reference used for a prudent contractual commitment.

The two mistakes that invalidate a simulation

A poorly built Monte Carlo simulation gives a false impression of scientific rigour while its results are biased. The two most common mistakes:

  • Ignoring correlation between activities — activities exposed to the same risk factor (the same crew, the same supplier, the same weather-exposed area) shouldn't be simulated as statistically independent. Without correlation, the simulation systematically understates the real spread of outcomes.
  • Underestimating uncertainty ranges through optimism — optimistic/pessimistic estimates squeezed too tightly around the "most likely" duration defeat the purpose of the method; the quality of the simulation depends directly on how honest the input ranges are, usually derived from workshops with package owners.

Cost and schedule: an integrated simulation is more revealing

A Monte Carlo simulation applied separately to cost and schedule understates real risk, because the two are often linked: a delay extends resource mobilisation, which increases variable cost. An integrated simulation, which propagates the impact of a duration delay onto associated variable costs, gives a more accurate picture of actual project exposure — particularly useful for sizing a credible contingency in front of a finance department or a client.

A well-run Monte Carlo simulation doesn't replace a well-structured schedule — it complements it, by revealing the uncertainty a deterministic schedule conceals. It's an exercise we build into assignments where the delivery commitment carries significant contractual weight.