Published on
A 68 Percent Probability. And Then What?
Monte Carlo simulation was part of our software years ago, and technically it was sound. Nobody wanted it.
Recently I have been thinking again about planning in ranges. The basic idea makes sense: away from the single planned figure, towards a corridor. Monte Carlo often enters at this point. The method takes uncertain assumptions, attaches probability distributions, and simulates thousands of scenarios. What comes out is a distribution instead of a single number: with 68 percent probability we achieve at least this EBITDA.
Our value driver models allowed distributions to be attached to driver values. Demos got consistently positive feedback, and we even had a partner who wanted to take the solution to market.
So why did nobody want to use it?
For a long time we blamed a lack of statistical knowledge. Eventually I had to ask myself the question: what do I concretely do with a 68 percent probability? Which decision do I make differently as a result?
Today I think what was missing was a rule defined in advance. At which percentile does which measure trigger? What is the next step after the result? Without that rule, every probability statement stays without consequence.
Methodologically, Monte Carlo is the better answer to uncertainty. I still very rarely see it used in steering.