Scenario Planning With Numbers: Monte Carlo For Leaders Who Decide Under Uncertainty

Almost every board plan carries the same three columns: best case, base case, worst case. It looks like an honest reckoning with uncertainty, and it is one of the weakest tools in common use. Three numbers tell you nothing about how likely each is, nothing about the wide territory between the base and the worst, and nothing about how ordinary variations combine. Worse, the format has a built-in bias: everyone plans to the base case and files the worst case as a scenario to worry about later. The uncertainty that should drive the decision gets flattened into a single hopeful number with two bookends.

There is a better way to reason about an uncertain plan, and it has been standard in insurance and finance for decades. Monte Carlo simulation replaces the three guesses with the full range of what could happen and, crucially, the odds attached to each part of it. It sounds technical. The idea is simple enough to explain in a paragraph.

What The Method Actually Does

Take your ordinary cash-flow model, the one that turns revenue and costs into a year-end number. Instead of typing a single figure into each uncertain input, you give it a range: revenue might land anywhere from a weak year to a strong one, storm losses might be nothing or large. The computer then plays the year out thousands of times. On each run it draws a random value for every uncertain input from its range, works the model through, and records the outcome. Ten thousand runs later you do not have one answer, you have ten thousand, and their shape is the real picture of where you might end up.

The output is a distribution. You can read the median, the middle outcome; the spread, how wide the plausible results are; and the tails, the good and bad extremes. Most valuably, you can read the probability of any threshold you care about. What are the odds we end below zero? What is the chance we miss the covenant? How bad is the worst five percent? Those are the questions a leader deciding under real uncertainty actually has, and three static scenarios cannot answer any of them.

A Worked Example: Cash Flow Under Storm Risk

Picture a business projecting year-end cash. Revenue is uncertain, sitting somewhere around a central estimate with a realistic spread. Costs vary. And once in roughly a hundred years a severe storm lands a large loss, with smaller storm losses more common, exactly the pattern from the expected-loss model earlier in this series. Feed those as ranges rather than fixed cells, run ten thousand simulated years, and the outcomes stack up like this.

Read what those three numbers do that a single forecast cannot. A point forecast of "about J$500,000 in cash" looks healthy and hides the whole problem. The simulation says the central case is indeed fine, and that there is a one-in-eight chance of ending the year in the red, driven mostly by the storm tail landing on top of a soft revenue year. That 12 percent is a number you can act on. It tells you how large a cash buffer to hold, whether the parametric cover from the risk model is worth its premium, and how much of the worst five percent you are willing to carry yourself. The plan stops being a hope and becomes a bet with known odds.

Where It Goes Wrong, And How To Keep It Honest

A simulation produces a clean, confident-looking distribution no matter how good the inputs are, and that is its trap. The output inherits every flaw in the ranges you fed it. Guess the revenue spread too narrow and you will understate the risk; assume the inputs move independently when they do not and you will badly understate the tail. The last point matters most in a storm economy. A hurricane does not politely hit only your property. It hits revenue as customers stay home, hits costs as you repair and restock, and hits collections as debtors struggle, all in the same weeks. A model that treats those as three unrelated draws will show a gentler tail than the real one. Building the correlation in, so a bad storm year drags several inputs down together, is the difference between a simulation that flatters you and one that protects you.

Used with that discipline, Monte Carlo is the most honest planning tool most businesses are not yet using. It does not remove uncertainty, and it does not predict which year the storm comes. It does something more useful for a decision-maker: it turns a vague fear into a probability you can size, budget against and, where it makes sense, insure. That is the whole series in one method. Govern the data, read the charts honestly, measure the risk, and then plan against the range rather than the hope.

Frequently Asked Questions

What is Monte Carlo simulation in plain terms?

It runs your plan thousands of times, each time drawing the uncertain inputs at random from their likely ranges, and records the outcome each time. Instead of a single answer you get a full distribution, so you can read the average, the spread, and the probability of the outcomes you care about.

Why is best-case, base-case, worst-case planning weak?

Three scenarios give three numbers and no probabilities. They do not tell you how likely the worst case is or what lies between, and they invite planning to the base case while treating the worst as an afterthought. Monte Carlo attaches odds to the whole range.

What can Monte Carlo tell a leader that a single forecast cannot?

The probability of specific outcomes, not just their possibility. Rather than "cash will be around J$500,000," it says the median is J$520,000, there is a 12 percent chance of ending below zero, and the worst five percent fall below minus J$150,000. That is a risk you can size and budget for.

How does Monte Carlo handle hurricane or disaster risk?

It includes rare, severe events in the inputs. Each simulated year draws whether a storm hits and how large the loss is, so a small share of runs carry a big storm loss. That produces the long negative tail and lets you read the probability that a storm year tips you into a shortfall.

Is Monte Carlo only for large companies?

No. It scales down cleanly. A small firm's spreadsheet cash-flow model already has the structure; the simulation replaces fixed cells with ranges and runs the sheet thousands of times. Knowing the probability of a shortfall matters as much to a thin-buffer small business as to a large one.

What are the limits of Monte Carlo simulation?

It is only as good as the input ranges and the relationships you feed it. Guess the ranges badly, or assume independence when inputs move together, and the tidy distribution is confidently wrong. A storm that hits revenue, costs and collections at once is worse than three separate shocks.

Related reading across the Caribbean AI network