SCHEDULE ANALYSIS · Updated November 4, 2026
Monte Carlo schedule risk analysis: a practical guide
A schedule gives you a planned finish date. Monte Carlo analysis explores how that finish might change when remaining activity durations vary. XER Viewer turns those repeated simulations into finish-date distributions and confidence dates.
What is Monte Carlo analysis?
The method samples uncertain inputs repeatedly and records the resulting outcomes. In a schedule model, duration samples are passed through activity dependencies to produce a different possible finish in each iteration. The collected outcomes describe a range, rather than a single promised date.
Project teams use this approach to discuss completion confidence, contingency and mitigation priorities. GAO includes schedule risk analysis in its scheduling best practices. Source: GAO Schedule Assessment Guide.
A short history
Computer-based Monte Carlo methods emerged at Los Alamos in the 1940s, with contributions from Stanislaw Ulam, John von Neumann and Nicholas Metropolis. Ulam's 1946 reflections on solitaire helped inspire the sampling approach. Its early applications involved physics; the same idea of repeated random trials now supports many fields, including project analysis. Metropolis's historical account and Los Alamos account of Ulam and von Neumann.
Run it in XER Viewer
- Open an XER schedule. Check the data date, remaining durations and relationships before simulating. Use the DCMA assessment to investigate structural issues first.
- Choose Schedule Check in the left navigation. The Schedule Assessment screen offers Monte Carlo Risk; select Run Monte Carlo to open and run it. Existing results can be reopened using View results.
- Review the inputs. Set Iterations, choose PERT/Beta or Triangular, and adjust the Optimistic and Pessimistic percentages. Choose ranges you can explain from experience or evidence.
- Choose Run Simulation after changing the settings. Wait for the progress indicator to complete.
- Inspect the results. Review confidence dates, the histogram, cumulative curve and activity criticality. Compare alternative assumptions by changing the inputs and running again.
- Share the assumptions with the result. Use the Monte Carlo toolbar's print control to preview the report and print it or save it as PDF.
How to choose uncertainty inputs
The app applies the same percentage ranges globally to remaining activity durations. For example, a 10-working-day activity with a 10% optimistic shortening and 30% pessimistic lengthening has bounds of 9 and 13 working days, with 10 as its most-likely value. PERT/Beta places more weight around the most-likely duration; triangular uses straight sides between the three estimates.
More iterations can stabilize sampled percentiles, but they cannot fix unrealistic assumptions or missing schedule logic. Compare a reasonable base case with a more adverse case and record why you chose the ranges.
Understand the results
| Result | What it tells you |
|---|---|
| Deterministic finish | The model's finish using most-likely durations without uncertainty. It may differ from the imported P6 finish because the model is simplified. |
| P50 / P80 / P90 | The finish date at the 50th, 80th or 90th percentile of simulated outcomes. P80 means about 80% of model runs finish on or before that date, conditional on the inputs. |
| Histogram | How often simulated finishes fall into each date interval. |
| Cumulative curve | The share of runs finishing by each date. |
| Contingency | The modeled working-day gap between a confidence date and the deterministic finish. |
| Criticality index | The share of iterations in which an activity is critical. High criticality warrants attention, but is not by itself a measure of how much mitigation would improve the finish. |
Know the model limits
- It forecasts remaining work from the data date; completed work is fixed.
- Uncertainty ranges are global, not activity-specific risk-register inputs. Durations are sampled independently, without correlation between activities.
- Date mapping uses a Monday–Friday working week. Calendar holidays, exceptions and hard date constraints are not modeled.
- Milestones, summary and level-of-effort activities carry zero duration and no duration uncertainty in this model.
- The network uses FS, SS, FF and SF relationships with lag, but this is not a full reproduction of the Primavera scheduling engine.
Use these outputs for exploration and discussion. A formal risk study may need validated activity-level estimates, common risk drivers, correlation and more complete calendar or constraint handling.
Use the forecast to make a decision
Investigate frequently critical activities, discuss the assumptions with the scheduler, and identify practical actions such as resolving approvals or protecting procurement dates. Update the source schedule and rerun after changes. Keep the data date and input settings with each saved report.
Next: check the quality of your schedule with the DCMA 14-point guide →