How to use the contingency reserve simulator
- List each risk with its chance of happening in percent and its cost if it does: low, most likely and high, separated by commas.
- Choose the confidence level you want the reserve to cover, the number of simulated projects and a seed.
- Read the reserve for your confidence level, the expected value, the simulated average and the table of totals at other levels.
- Check the expected cost of each risk in the second table, then copy or download the CSV.
Formula
each simulated project: for each risk, it happens with probability p; if so, its cost is drawn from a triangular distribution on [low, high] with peak at the likely value
total = sum of the costs of the risks that happened
reserve at confidence c = total at position c in the sorted list of project totals (nearest rank)
expected value = sum of p x (low + likely + high) / 3
The simulated average should land close to the expected value, within a few standard errors, which the page reports. The reserve above the expected value is the extra cushion that the chosen confidence level asks for.
Five risks
The sample risks are a supplier delay (30%, $2,000 to $12,000), a scope change (40%, $5,000 to $30,000), a key person leaving (15%, $8,000 to $40,000), rework after testing (50%, $1,000 to $9,000) and a permit slip (20%, $3,000 to $15,000). The expected value is $14,983.33.
Simulating 10,000 projects with seed 2026 gives a median total of $12,791.35, a simulated average of $15,015.51, and an 80% reserve of $25,179.74, which is $10,196.41 above the expected value. In 13.7% of the projects no risk happened at all.
Expected value is not the reserve
The expected value is the average total, and if you set aside only that amount you will run short in roughly half the projects. A reserve for a higher confidence level covers more outcomes but sets more money aside. Moving from the median to 80% or 90% shows how much extra cushion each step buys.
The distribution of the total is skewed. Most projects have a few small costs, and a few have several large ones, which is why the high percentiles are far above the average.
Independence and estimates
The simulation treats risks as independent. In real projects risks are often linked, such as a delay that causes both rework and a permit problem, and linked risks make the spread wider than shown. The chances and costs are your estimates, and the output is only as sound as they are.
Keep the risks and owners in a Risk Register Template, simulate the schedule with the Monte Carlo Project Duration Simulator, and check one risk with the Expected Value Calculator.
Assumptions and limits
- Costs are in one currency. The low, likely and high values must satisfy low <= likely <= high and low must be 0 or more.
- Each risk happens at most once per project, with the chance you give.
- Up to 50 risks and 20,000 projects are read. The same seed gives the same result.
- Percentiles use the nearest rank. The simulated average has a standard error that is shown.
- Nothing typed here leaves your browser.
Frequently asked questions
Why a triangular distribution?
It needs only three numbers people can estimate: a low, a likely and a high value. It is a simple and widely used way to describe a cost whose exact amount is uncertain.
What confidence level should I choose?
It depends on how costly an overrun is and how much money you can tie up. Many teams look at the 70% to 90% range. The table shows the totals at several levels so you can compare.
Why does the average differ slightly from the expected value?
The simulation uses a finite number of projects, so its average has some noise. The page reports the standard error, and more projects reduce it.
Can I use it for hours instead of money?
Yes. Enter hours for the low, likely and high values. The money formatting will still show a dollar sign, so read the numbers as hours.