Every budget is a single story about the future: one sales figure, one cost line, one cash balance at year end. It is useful for alignment, but it says nothing about how wrong it might be. When the year goes off plan, and it always does, the organisation has no shared view of how far off is still normal and how far off is a real problem.

Monte Carlo simulation replaces the single story with thousands of them. Each assumption becomes a range instead of a point, the model is run again and again with random draws from those ranges, and the results form a distribution: the most likely outcome, the bad case and everything in between, each with its probability.

The fan chart: a budget with its uncertainty

Consider a company with 20 million in cash and a plan that adds 0.3 million a month. Monthly cash flow varies, and in some years a six-month downturn hits demand. We simulated 2,000 two-year paths. The fan shows where cash will be with 50% and 80% confidence, against the plan and a 10 million covenant floor.

Simulated cash balance over 24 months: median, 50% and 80% bands, the plan and the 10 million floor. 2,000 simulated paths; illustrative assumptions.

Three things appear that the budget could never show. The median ends below the plan, because downturns weigh more than good months. The 80% band at month 24 spans roughly 20 million, so a miss of several million is normal, not a failure. And in about one scenario in ten, cash touches the floor at some point, which is a risk the board should decide about now, not discover later.

How to build one without a data science team

  1. Start from the budget model you already have. Monte Carlo does not replace it; it runs it many times.
  2. Choose five to eight drivers. Volume, price, key input costs, exchange rate, collection days. The rest can stay fixed.
  3. Give each driver a range and a shape. Use history where you have it and expert judgement where you do not; a low, likely and high value is enough to begin.
  4. Model the links. If a weaker currency raises costs and lowers demand at the same time, simulate them together, not independently.
  5. Report probabilities, not just averages. The chance of breaching a covenant, of missing guidance, of needing new funding.

The value of the exercise is not the precision of the numbers; it is the conversation it forces. When leadership sees the range, it stops debating whether the forecast is right and starts deciding how much risk it is willing to carry and what it will do if the bad case arrives.
BS

Bartor S.

Founder

All insights