Every forecast is, at its core, a probability distribution. The question is whether you show the distribution or collapse it into a single number before presenting it. Most organisations choose the single number, for understandable reasons: it is easier to communicate, easier to hold someone accountable to, and easier to put in a slide. It is also, in most cases, a less honest representation of what the analysis actually shows.

What a Point Estimate Hides ¶

A revenue forecast of £4.2 million does not mean the model predicts exactly £4.2 million. It means that £4.2 million is the central estimate of a distribution that might run from £3.1 million to £5.6 million under plausible assumptions. When that distribution is hidden, the board cannot assess whether a result of £3.8 million represents a failure or a normal outcome. They are making judgments without the information they need.

How Probabilistic Forecasting Works ¶

A probabilistic forecast presents a central estimate alongside a credible interval, typically the range within which the actual outcome is expected to fall with a specified probability. We usually present a 70% credible interval alongside the central estimate, with explicit notes on the assumptions that drive the width of the range. Wider ranges are not a sign of a weaker model. They are a sign of an honest one.

The Practical Objection: It Feels Less Certain ¶

The most common pushback we receive when presenting ranges is that they feel less certain than a single number. This is correct. They are less certain. The question is whether the certainty implied by a point estimate is real or performed. In our experience, boards and investors who are given ranges consistently make better decisions than those who are given single numbers, because they are forced to engage with the question of what would have to be true for the optimistic case to hold.

Presenting Ranges Without Losing the Room ¶

There is a skill to presenting probabilistic outputs in a way that is legible to a non-technical audience. We have found that the most effective approach is to anchor the presentation on the central estimate, introduce the range as a measure of the model's honest uncertainty, and then spend time on the assumptions that drive the upper and lower bounds. The goal is not to overwhelm with statistics but to give the decision-makers the information they need to make a judgment.

If your organisation currently presents forecasts as single numbers and you are curious about what a probabilistic approach would look like for your specific situation, we are happy to walk through an example. Get in touch with us to arrange a short call.