Monte Carlo Simulation in Financial Modeling

What Monte Carlo simulation is, how it differs from scenario analysis, and how thousands of iterations turn a point forecast into a probability distribution.

Every forecast is wrong. The useful question is not 'what will happen' but 'how wrong could we be, and how likely is each outcome?' Sensitivity analysis flexes one input; scenario analysis tests three coherent stories. Monte Carlo simulation goes further: it runs the model thousands of times with inputs drawn from probability distributions, and hands you the full range of outcomes with a likelihood attached to each.

Definition

Monte Carlo simulation runs a model many thousands of times, each time drawing key uncertain inputs at random from defined probability distributions, to produce a distribution of possible outcomes rather than a single point estimate.

From point estimate to distribution

The three approaches answer progressively better questions. A point estimate says 'NPV is $9m' — precise and almost certainly wrong. Scenarios say 'between $4m and $16m depending which world we are in' — honest, but with no sense of which is likely. Monte Carlo says '$9m is the median, there is an 8% chance of a negative NPV, and a 90% chance of landing between $2m and $17m' — which is the answer a decision-maker can actually act on.

Method	What you flex	Output	Tells you likelihood?
Point estimate	Nothing	One number	No
Sensitivity	One input at a time	A table of outcomes	No
Scenario	A coherent set of inputs	3 cases (base/up/down)	No
Monte Carlo	All inputs, thousands of draws	Full distribution	Yes

Three ways to handle uncertainty

How it works

The mechanics are simpler than the name suggests. First, pick the handful of inputs that genuinely drive the answer — usually growth, margin, WACC and CAPEX. Second, give each a distribution rather than a value: revenue growth might be normally distributed around 10% with a 4% standard deviation; a launch date might be uniform across a range. Third, run the model thousands of times, each pass drawing a random value from each distribution. Fourth, collect every resulting NPV or IRR and plot them. Ten thousand iterations of a small model runs in seconds.

Reading the output

The output is a distribution, and you read it with percentiles: P10 (only a 10% chance of landing below this), P50 (the median), P90 (a 10% chance of exceeding it). The single most valuable figure is usually the probability of a bad outcome — the share of simulations where NPV is negative or cash runs out.

The median lands in the $5–10m band — but 8% of simulations produce a negative NPV. That downside risk is invisible in a single-point forecast.

That 8% is the whole point. A point estimate of $9m implies a comfortable yes. Knowing that roughly one run in twelve destroys value changes the conversation — you might still proceed, but you will size the downside, stage the investment, or negotiate protection first.

Correlation is where it goes wrong

The biggest technical trap is treating inputs as independent when they are not. In the real world a demand shock hits volume and price together; a recession lowers growth and raises your cost of debt at the same time. A simulation that draws them independently quietly cancels out the very scenario you most need to see, and produces a distribution that is far too narrow — a false sense of safety dressed up in statistics.

When it is worth it — and when it is overkill

Monte Carlo earns its keep on big, irreversible, long-horizon decisions where the downside genuinely matters: infrastructure and energy projects, large acquisitions, drug development, anything with a 20-year life and heavy upfront capital. It is overkill for a three-year startup plan, where the inputs are so uncertain that dressing them in distributions creates false precision. And remember the hard truth: garbage in, garbage out. A simulation cannot rescue bad assumptions — it just gives you thousands of variations of them, with a reassuring bell curve on top.

Start with sensitivity and scenarios

For most models, two-way sensitivity tables and an honest downside case answer the same question at a fraction of the effort. EasyFinancialModels builds sensitivity tables on WACC, terminal growth, revenue and margin automatically, and lets you flex a full downside by editing the assumptions. Build a model free for up to 3 years, find which assumptions actually move the answer, and only reach for Monte Carlo when the decision is big enough to deserve it.

→ Build your financial modeling model free with the Financial Model tool

More Financial Modeling guides

How to Build a Financial Model in Excel · Quarterly Financial Model: When to Use Quarterly Forecasts Instead of Annual Models · Industry Financial Model Templates: How to Choose the Right Revenue Drivers · How to Build a Startup Financial Model for Investors · The Three-Statement Financial Model Explained

About the author

Every model is built and reviewed by the project's Financial Advisor — a Fellow Chartered Accountant (FCA), Chartered Global Management Accountant (CGMA) and Associate Chartered Management Accountant (ACMA) with around two decades of corporate finance, audit and accounting experience, designing investor-grade financial models across industries. Full credentials and background are available on LinkedIn. More about the author →

← All guides · WACC calculator · DCF calculator · IRR calculator · Inside the 15-sheet model