CIVIL_SYSTEMS
Tools · Risk analysis

Monte Carlo Risk Calculator

Cost & schedule ranges · P10 to P90 · Contingency · Tornado

Reviewed November 2026 · Civil Systems LLC

Monte Carlo simulation replays your project thousands of times to show how likely any budget or finish date really is.

In each replay, every uncertain cost or duration lands somewhere in its range and each risk either happens or does not. The spread of those results lets you choose how much confidence to pay for, instead of hoping the most likely number holds.

Start from an example or download the Excel template, fill in your own estimate ranges and risks, and open it here. The calculator runs in your browser and shows the spread, the confidence curve, the contingency at the level you choose, and which lines drive the range. Then practice reading the results on the same chart.

What is Monte Carlo simulation, anyway?

Give every uncertain line a range instead of one number. The computer picks a value for each line from its range, adds them up, and writes down the total. Then it does that again, 10,000 times. Sorted from lowest to highest, those totals show which outcomes are common and which are rare. That spread is the forecast.

The name comes from the casino in Monaco, because the method runs on chance. Scientists at Los Alamos in the 1940s, including Stanislaw Ulam and John von Neumann, used it for problems too tangled for equations; project teams adopted it for cost and schedule risk decades later.

Why not just use one number?

A single estimate is usually the sum of most likely values, and most likely is not the same as safe. Most costs and durations can overrun by more than they can underrun, and risks add when they happen. So the sum of most likely values is met far less than half the time: in the bridge example below, in about 7% of runs.

What does P80 actually mean?

The calculator builds your project 10,000 times. Each time the luck is different but plausible: in one run a supplier delivers early, in another a key task runs long. Then it lines up the 10,000 results from best to worst.

P80 is the point 80% of the way along that line: 8,000 runs came in at or under it. For the bridge example that is $1,613,178. In plain terms, if you ran this project 10 times, you would expect to stay within $1,613,178 about 8 times and go over about 2 times.

P50 is a coin flip: half the runs land above it ($1,482,898 here). P90 gives more cushion ($1,697,388 here), but that cushion is money or time you hold back and cannot spend elsewhere.

What P80 does not mean: it is not a guarantee, and it is not 80% of the work done. It is only as good as the ranges you entered, and it covers none of the risks you left out.

The decision: choosing P80 means accepting roughly a 1-in-5 chance of needing more. Contingency is the gap between your plan and that number. Here, $1,613,178 minus the point estimate of $1,355,000 is $258,178.

When should I not trust it?

When the ranges are guesses nobody has challenged, because the output is only as good as the inputs. When lines tend to move together, such as labor rates rising across every trade: this calculator treats lines as independent, so the real range is wider. And for schedules with parallel paths, because adding durations ignores network logic. Use it to compare options and size contingency, not as a promise.

01 / Your model

Estimate lines and risk events

An item always happens but its amount is uncertain, like the cost of concrete or the duration of a permit review. A risk is an event that may or may not happen; enter its probability and the impact if it does. Use negative impacts for opportunities.

Bring your own data Download the template, fill in the Model sheet in Excel, Google Sheets, or Numbers, then open it here. Up to 50 lines; .xlsx or .csv up to 1 MB. The file is read in this browser and never uploaded.

Model lines. Edit any cell; results update automatically.
#TypeNameDistributionMinMost likelyMaxProbability %Remove
02 / Results and practice

One simulation, four ways to read it

Simulation results

Point at the chart, tap it, or focus it and use the arrow keys to read any percentile from P5 to P95.

How the outcomes spread

Each bar counts the simulated runs whose total landed in that range. A long tail to the right is normal: most lines can overrun by more than they can underrun, and a risk event adds its impact only in the runs where it happens. Bars past your chosen confidence are shown in gray. The dashed line is the point estimate; the solid line is the value at your chosen confidence.

03 / Understand

Variability, risk events, and contingency

Quantitative risk analysis puts numbers on the combined effect of uncertainty. Two kinds of input feed it, and they behave differently in the simulation.

Item: uncertain amount

Always part of the total. Each run draws a value from its range.

value ~ range(min, most likely, max)

Risk: uncertain event

Part of the total only in the runs where it occurs. On average it adds its expected value.

EMV = probability × mean impact

Three shapes for a range

TriangularMean = (min + ML + max) / 3. Straight sides give the extremes real weight. Common when the range comes from judgment.
PERT (beta)Mean = (min + 4 × ML + max) / 6. Concentrated near the most likely value, with thinner tails.
UniformMean = (min + max) / 2. Any value in the range is as likely as any other. Use when you cannot name a most likely value.

Why the sum of most likely values is rarely likely

Add up the most likely value of every item and you get the point estimate most budgets start from. Here is what the simulation says about it for the model above:

    Contingency at a confidence level

    Contingency

    How much to add to the point estimate so the model meets it at the chosen confidence.

    Contingency(P80) = P80 − point estimate

    Chance of meeting a target

    The share of runs at or under the target. Read it where the target crosses the S-curve.

    P(total ≤ target) = runs ≤ target ÷ all runs

    PMI separates a contingency reserve, set inside the baseline for identified risks, from a management reserve, held outside it for unforeseen work within scope. This calculator estimates the first. Choosing the confidence level is a decision for the sponsor or owner, not a statistical result.

    04 / Work it through

    Five runs by hand

    The whole method fits in four steps. Here are the first five runs of the simulation above, line by line, followed by the arithmetic that turns thousands of runs into a percentile and a contingency.

    Common questions
    What is Monte Carlo simulation in project management?

    It is a way to turn ranges into a forecast. Each uncertain cost or duration gets a range, and each risk gets a probability and an impact range. The model draws one value for every line, adds them up, and repeats thousands of times. Sorting those totals shows how likely each outcome is. PMI describes it as using multiple iterations of a computer model to develop a probability distribution of possible outcomes.

    What do P10, P50, P80 and P90 mean?

    Each is a percentile of the simulated totals. P80 is the value that 80% of runs came in at or under, so the model gives roughly a 4-in-5 chance of not exceeding it. P50 is the median. P10 and P90 bracket the middle 80% of outcomes. They describe the model, so risks you left out are not covered.

    How much contingency should a project carry?

    There is no universal answer. Contingency at a confidence level is that percentile minus the point estimate, for example P80 minus the sum of most likely values. The confidence level is a decision for the owner or sponsor based on risk appetite and policy; some public owners set it, such as P65 or P80. In PMI terms this covers contingency reserve for identified risks. Management reserve for unforeseen work is set separately. See defensible cost contingency.

    Should I use a triangular or a PERT distribution?

    Both take a minimum, most likely and maximum. Triangular gives the extremes more weight, so its mean is (min + most likely + max) / 3. PERT, a beta distribution, concentrates around the most likely value, so its mean is (min + 4 × most likely + max) / 6. Use triangular when the range comes from judgment and you want a cautious spread, PERT when the most likely value is well supported.

    What does a tornado diagram show?

    It ranks the lines by how strongly each one moves with the total across the runs. This calculator uses rank correlation, the measure PMI describes for tornado diagrams. Long bars at the top are where better information or a risk response will narrow the range the most.

    How many iterations do I need?

    Enough that the answer stops moving. Run the model, change the seed, and run it again: if the P80 shifts by less than a percent, you have enough. 10,000 runs is usually plenty for a register of a few dozen lines. The What if? tab asks you to predict this for your own model.

    Is my spreadsheet uploaded anywhere?

    No. The workbook is opened and read by your browser. Nothing is sent to a server or saved, and the page cannot read anything except the file you choose. Download your model to keep it.

    Can I use this for schedule risk?

    Only for activities that run one after another, because the calculator adds the lines. When activities run in parallel, the finish date depends on the network: a delay on a path with float may not move the finish at all. A schedule risk analysis recalculates the critical path in every run. The CPM calculator shows how paths and float work.

    Model assumptions. The total is the sum of all lines. For durations, that treats every activity as if it sat on one path, one after another: parallel paths, float, and merge bias (the extra delay where several uncertain paths must all finish before the next activity starts) are not modeled. Lines are independent: the model does not correlate them, so if two lines tend to overrun together, the real range is wider than shown. Risks are independent events, each sampled with its own probability; combine overlapping or mutually exclusive risks into one line. Up to 50 lines and 20,000 iterations; values up to ±1 trillion. Results are reproducible for the same model, iterations, and seed. Percentiles use linear interpolation between sorted runs. Terms follow the PMI Lexicon of Project Management Terms (Version 5.0), the PMBOK Guide, Eighth Edition (Section 5, sensitivity analysis and three-point estimating), and The Standard for Risk Management in Portfolios, Programs, and Projects (X6.4).