Show us a forecast that is a single number and we will show you a document that was not written to inform a decision. It was written to win a room.
This sounds harsh until you try the alternative. Pick any major number in your last planning cycle — next year's revenue, the cost of a build, the timeline on the integration, the headcount the new region will need. Ask the person who produced it what range they would actually bet on, with their own money, at two-to-one odds. Watch what happens.
What happens, almost always, is a pause. Then a shrug. Then a wider range than the one in the deck. Often much wider. Sometimes you discover that the number in the deck was the low end of the range the analyst privately believed, sharpened to a point because the CFO wanted "the number." Sometimes you discover the opposite: the deck number was the optimistic case, sharpened because that's what the CEO wanted to hear. Either way, what is now on the page is not the forecast. It is the result of a negotiation between the forecaster and the room.
That document is a marketing artifact. It may be useful as a marketing artifact — alignment is real, narratives matter, companies are run on shared stories. But it should not be confused with information you can plan against.
The cost of false precision
A real forecast — meaning, one written to help someone make a decision rather than to settle a meeting — has a range, and the range has been earned. Earned meaning: the person producing it has thought through what could go right, what could go wrong, what historical analogs suggest, how badly previous forecasts of this type have been off, and what their own track record says about their tendency to over- or under-shoot.
The reason this matters is not philosophical. It is operational. Decisions that depend on a forecast almost always have asymmetric payoffs. A revenue plan that misses by twenty percent on the upside is a pleasant surprise. A revenue plan that misses by twenty percent on the downside is a hiring freeze, a recut budget, possibly a layoff. The point forecast, by definition, does not tell you which of those is more likely. The range does.
A point forecast tells you what to expect. A calibrated range tells you what to prepare for. Companies that conflate the two end up surprised by things they had the data to anticipate.
We have watched a quarter-billion-dollar capacity decision get made on a single-number demand forecast that, when we asked the team to widen, turned out to have a fifty percent chance of being off by more than a third. The point estimate said: build the second factory. The honest range said: build the second factory and secure a flexible-capacity option, because the cost of being wrong on the high side and the cost of being wrong on the low side are different by an order of magnitude. The point estimate, in other words, contained none of the information actually needed to decide.
What a real forecast looks like
A useful forecast has three properties.
First, it has a range, expressed at a stated confidence — say, eighty percent. Not "high case, base case, low case" pulled from the same gut, but a genuine attempt to bracket the truth. The eighty-percent number means: if I produced ten forecasts like this, I'd expect eight of them to bracket the actual outcome. If your historical record says you brack at sixty percent when you claim eighty, your forecasts aren't calibrated, and the first thing to fix is the gap, not the forecast.
Second, it has named drivers. Not "various market factors" — actual variables, with directions. "If our second AE hire ramps on the historical curve, the base case holds; if they ramp at the median of our last four hires, we land at the bottom of the range." Anyone reading this can update on the world as it changes. Anyone reading a point estimate has to throw the whole document away the moment the world moves.
Third, it has a track record attached, even if implicitly. The team that produced it knows how often forecasts of this type, by this team, in this domain, have been right. This is the single most useful piece of information a CFO can have, and it is almost never collected. Three quarters of building a forecasting culture is just keeping score.
The cultural piece
The hardest part of moving a company toward calibrated forecasting is not technical. It is cultural. Calibrated forecasters say "I don't know" more often, hedge more on the record, and resist being pinned to a number. In a company that rewards apparent certainty, this gets read as weakness. In a company that rewards being right over time, it gets read as professionalism. The difference between the two cultures is, eventually, the difference between two stock prices.
If your team is still producing point forecasts and your decisions are still being made off them, that is a solvable problem, and a worthwhile one. Calibrated forecasting is the closest thing to a free lunch we know of in business — the inputs cost nothing, the discipline is teachable, and the returns compound. The only thing it costs is the comfort of pretending you knew.
Writes about decision quality at Bayeseon. Reach the team at hello@bayeseon.com.