Monte Carlo Simulation

Definition

Monte Carlo Simulation is a quantitative risk-analysis technique in which a project model — typically schedule, cost, or both — is run thousands of times with key inputs drawn from probability distributions. The output is not a single forecast but a distribution: a range of possible outcomes with associated probabilities. Instead of "the project will finish on July 14," Monte Carlo answers "there is a 50% chance of finishing by July 14, an 80% chance by August 28, and a 95% chance by October 11."

The technique converts a deterministic plan into an honest one. Every estimate hides uncertainty; Monte Carlo surfaces it.

History

The method takes its name from the Monte Carlo casino. It was developed during the Manhattan Project in the 1940s by Stanislaw Ulam, John von Neumann, and Nicholas Metropolis to solve neutron-diffusion problems too complex for analytical methods. The technique spread to operations research, finance, and engineering. Application to project schedule and cost risk began in earnest in the 1980s as desktop computing made the iteration counts affordable. Today, AACE International Recommended Practice 57R-09 and similar guidance codify the technique for capital projects.

How It Works

1. Build a deterministic model — schedule network or cost estimate.
2. Identify uncertain inputs — typically activity durations, productivity factors, unit rates, quantities, and risk-event triggers.
3. Assign probability distributions to each uncertain input — triangular (min, most-likely, max) is most common; beta, uniform, and lognormal also used.
4. Model correlations between inputs — uncorrelated assumptions produce optimistic results because risks rarely move independently.
5. Run the simulation — typically 1,000–10,000 iterations.
6. Analyse outputs — distribution curves, P-values (P50, P80, P90), and tornado diagrams ranking inputs by impact.

Reading the Outputs

• P50 — the median; 50% chance of finishing by this date or under this cost.
• P80 / P90 — the values commonly used for contingency setting. A P80 cost is a budget with an 80% confidence of not being exceeded.
• S-curve of outcomes — cumulative probability against date or cost. The steeper the curve, the lower the risk.
• Tornado diagram — ranks the contribution of each input variable to overall variance. The variables at the top of the tornado deserve the most management attention.
• Criticality index — for each activity, the percentage of iterations in which it lies on the critical path. Activities with high criticality but low management attention are typical risk surprises.

Real-World Construction Example

A USD 1.2 billion offshore platform programme presented a deterministic completion date of 42 months. Monte Carlo analysis using triangular distributions on activity durations and correlations between weather-dependent marine activities showed P50 at 46 months, P80 at 51 months. Management originally believed the schedule was robust. The tornado diagram revealed that 60% of total variance came from three input sets: weather windows for offshore lift, vendor delivery of the main module, and commissioning duration. Mitigation focused there: a second contracted lift vessel as backup, a vendor-incentive on early delivery, and a phased commissioning approach. Eighteen months later a fresh Monte Carlo run showed P80 at 47 months — four months of risk eliminated by surgical attention to the right three inputs. The deterministic date was always going to lie; Monte Carlo revealed where to act.

Real-World IT / Agile Example

A platform migration team used Monte Carlo on a release plan with 14 inter-dependent workstreams. Each workstream's duration was modelled as a beta distribution from historical velocity. The simulation showed that the deterministic plan was at the P30 confidence level — i.e., 70% likely to slip. Management used the analysis to negotiate a realistic external commitment date at P75 and to focus contingency on the three workstreams contributing 55% of variance. Without the simulation, the team would have committed to a date with one-in-three odds and a near-certain executive crisis at month nine.

Project Controls Perspective

Controls teams use Monte Carlo for three primary purposes: contingency setting (justifying budget and schedule reserves at P80 or P90 confidence rather than gut feel), risk-driver identification (the tornado diagram tells management where to spend management attention), and stage-gate decision support (P-value forecasts at each gate inform the go/no-go decision better than single-point estimates). The technique should be applied at three points minimum: initial baseline approval, after every major scope change, and at the start of each new tranche of a programme.

Common Mistakes

• Treating triangular distributions as truth — they are convenient approximations, not reality.
• Ignoring correlations between inputs; uncorrelated assumptions systematically underestimate risk.
• Too few iterations; below 1,000 the tail of the distribution is unstable.
• Running Monte Carlo on a baseline nobody believes; garbage in, garbage out.
• Presenting only P50 to management, hiding the upside-risk tail.
• Failing to identify risk-event triggers separately from activity durations.
• Re-running after every minor change; the simulation becomes a treadmill rather than a decision tool.

Expert Tips

• Use three-point estimates collected in workshops — discipline leads and field staff have better calibration than office planners.
• Model risk events explicitly as triggered impacts, not as widened activity durations.
• Always publish the tornado diagram alongside the P-curve. The curve tells you the answer; the tornado tells you what to do about it.
• Commit at P80 externally, manage to P50 internally. The gap is the management reserve.
• Document the distributions and assumptions. Future audits and the next project will need them.

Key Takeaways

• Monte Carlo converts deterministic forecasts into honest probability distributions.
• Triangular distributions, correlated inputs, and at least 1,000 iterations are the operational basics.
• P80 is the typical external commitment level; P50 is the internal management target.
• The tornado diagram, not the P-curve, tells you where to act.
• Apply at baseline, after major changes, and at every stage gate.

Related Concepts

Monte Carlo Simulation interlocks with Risk Management, Uncertainty Analysis, PERT, CPM, and Cost Control. Worked schedule-risk-analysis examples are available at PMMilestone.org.