A forecast that says "next quarter will be $4.2M" is not a forecast. It is the midpoint of one, with the useful part removed.
What was removed
Every forecast has uncertainty. The uncertainty is not a weakness in the method — it is a measured property of the situation, and it carries most of the decision-relevant information.
Next quarter will be $4.2M and next quarter will be between $3.9M and $4.5M, with 80% confidence look like the same statement with extra caution attached. They support completely different decisions.
If you are deciding whether to hire, the first tells you nothing about whether the downside is survivable. The second tells you the plan must work at $3.9M.
What to ask for
Four things. If a forecast cannot supply them, it has not been validated.
The range, and the confidence attached to it. Not "roughly." A stated interval at a stated confidence.
The measured error on data the model has not seen. Accuracy claimed on data the model was fitted to is not accuracy; it is memorization.
What it was compared against. The naive baseline, and by how much the model beat it. Without this you cannot tell whether the model added anything.
The conditions under which it stops being valid. A reorganization, an acquisition, a definition change, a shift in the mix. Every model has these, and a forecast delivered without them is being presented as more durable than it is.
Reading a range honestly
Two practical habits:
Plan against the unfavorable end, not the midpoint. The midpoint is the single least informative point in the interval for a decision with asymmetric downside.
Track actuals against the predicted range, not the predicted number. The right question after the fact is not "was the forecast right" — it will not be exactly right. It is did the actual land inside the range as often as the confidence level implied. A forecast claiming 80% confidence whose actuals land outside the band half the time is miscalibrated, and that is measurable.
Why the single number persists
Because it is easier to put in a board pack, and because a range invites the question "so you don't know?"
The honest answer to that question is: no, and neither does anyone else — what we can tell you is how wide the uncertainty is, how that compares to guessing, and what would narrow it.
Narrowing the range of things you have to plan for is genuinely valuable, and achievable. Certainty is not, and a supplier offering it is describing something that does not exist.
Practical check
A useful board-pack format shows the midpoint, the planning range, the confidence level, the naive baseline error, and the conditions that would invalidate the estimate. That is only a few more lines than a single number, but it changes the conversation from certainty to a decision made under measured uncertainty.
Predictive Intelligence delivers forecasts with measured accuracy, a stated range, and documented conditions under which the model should be distrusted.