Project Management guide · Risk, quality and control
Monte Carlo Simulation in Project Management: A Worked Example
Monte Carlo simulation is a quantitative risk analysis technique that runs a project model thousands of times, each time picking random values from the range of possible durations or costs. The results show how likely the project is to finish by a given date or within a given budget.
Key takeaways
- Monte Carlo simulation turns estimate ranges into probabilities for dates and costs.
- Most-likely plans often have a low chance of being met.
- Use P50 and P80 results to set commitments and contingency.
- Good inputs and sound schedule logic matter more than the software.
Why use Monte Carlo simulation
A single-point estimate hides uncertainty. If each task has a range of possible durations, adding up the 'most likely' values usually gives a date the project has only a small chance of meeting, because overruns tend to be larger than underruns. Simulation shows the full spread of outcomes instead.
- Answers questions such as 'how confident are we of finishing by 30 September?'
- Helps size schedule and cost contingency reserves; see contingency plan.
- Shows which tasks or risks drive most of the uncertainty.
- Is one of the main techniques for quantitative analysis in project risk management.
How it works
- Build a model: a schedule network or cost breakdown. See critical path method for schedule logic.
- Replace single estimates with ranges, often a three-point estimate (optimistic, most likely, pessimistic), and choose a distribution such as triangular or PERT (beta).
- Optionally add discrete risks with a probability and impact, for example a 30% chance of a 10-day delay.
- Run the model many times, typically thousands of iterations. Each run samples a value for every task and risk.
- Collect the results into a histogram and cumulative curve (an S-curve).
- Read off confidence levels, such as P50 and P80, and review which inputs drive the result.
Worked example
A small project has three tasks in sequence. The estimates are in working days:
| Task | Optimistic | Most likely | Pessimistic |
|---|---|---|---|
| A: Design | 8 | 10 | 16 |
| B: Build | 15 | 20 | 30 |
| C: Test | 4 | 5 | 9 |
| Total | 27 | 35 | 55 |
Adding the most likely values gives 35 days. Running 100,000 iterations with triangular distributions gave the following results (your figures will vary slightly from run to run):
| Measure | Result |
|---|---|
| Chance of finishing in 35 days or less | About 14% |
| P50 (median) | About 39 days |
| P80 | About 42 days |
| P90 | About 44 days |
So a plan built on 35 days has roughly a one-in-seven chance of being met. If the sponsor wants 80% confidence, the team should commit to about 42 days and hold the difference as schedule contingency.
Reading the results
| Term | Meaning |
|---|---|
| P50 | A 50% chance of finishing at or below this value |
| P80 | An 80% chance; often used for commitments and reserve sizing |
| S-curve | Cumulative probability plotted against date or cost |
| Tornado chart | Ranks inputs by how strongly they influence the result |
| Criticality index | How often a task appears on the critical path across all runs |
Common mistakes
- Using ranges that are too narrow; people are usually over-confident.
- Ignoring correlation, for example when the same cause would slow several tasks at once.
- Modelling a schedule with poor logic or missing dependencies.
- Presenting results as precise facts rather than estimates based on the inputs.
- Running the analysis once and never updating it.
Tools
Simulation can be run in a spreadsheet with random-number formulas, with spreadsheet add-ins, or with dedicated schedule risk analysis software. Some project management tools and Agile forecasting tools apply the same idea to throughput data to forecast when a backlog will be finished; see throughput.
Free templates for this topic
Frequently asked questions
What is Monte Carlo simulation in project management?
A technique that runs a project model many times with random values drawn from estimate ranges, to show the probability of meeting dates or budgets.
What is a P80 estimate in Monte Carlo simulation?
A value with an 80% chance of not being exceeded according to the simulation. It is commonly used for commitments and reserve sizing.
How many iterations does a Monte Carlo simulation need?
Usually thousands. Results stabilise as the number of runs grows; check that repeated runs give similar answers.
What distribution should I use for Monte Carlo simulation?
Triangular and PERT (beta) distributions are common for three-point estimates because they are simple and intuitive. Use historical data where you have it.
Is Monte Carlo simulation qualitative or quantitative?
Quantitative. It produces numerical probabilities for outcomes such as finish date and total cost.
Can I do Monte Carlo simulation in Excel?
Yes. You can sample from distributions with formulas and repeat the calculation with a data table, or use an add-in.
Why does Monte Carlo simulation give a later date than the plan?
Because estimate ranges are usually skewed towards overruns and parallel paths merge, so the most likely total is rarely the sum of most likely task durations.
What is a tornado chart in Monte Carlo analysis?
A bar chart ranking inputs by their influence on the result, used to find which tasks or risks matter most. It is a form of sensitivity analysis.
How does Monte Carlo help with contingency?
The gap between the planned value and a chosen confidence level, such as P80, gives a defensible size for schedule or cost contingency.
Do Agile teams use Monte Carlo simulation?
Yes. Many Kanban and Scrum teams sample historical throughput to forecast how many items they can finish by a date or when a set of items will be done.
What are the limitations of Monte Carlo simulation?
Results are only as good as the model and estimates. Poor logic, narrow ranges or ignored correlations give falsely optimistic results.