Skip to content

Forecast resolution

Forecast resolution is the act of deciding, at the review point, whether a forecast held. It requires comparing the stated expectation with what happened and recording the result, including where the call was wrong and where the outcome was ambiguous.

A forecast that is never resolved is an opinion that was written down carefully.

Why it matters when the plan changes

Resolution is the step that converts a prediction into evidence. Without it a forecast carries all the cost of being specific and none of the benefit, because nothing is compared against what actually happened and nothing accumulates into a record. Forecasting itself is defined by exactly this comparison: a forecast can later be checked against actual outcomes, and that check is what resolution performs. It is also the step most often skipped, because by the review date attention has moved elsewhere and nobody owns closing the loop.

The tension is that resolution is frequently uncomfortable and sometimes genuinely hard. Some outcomes are ambiguous, some were affected by the forecast itself prompting action, and some were right for the wrong reason. Scored forecasting programmes such as the Good Judgment Project depend on this step happening consistently, since improvement over time only shows up once outcomes are recorded and compared. Recording that complexity honestly is more useful than forcing a binary, and it takes longer than anyone budgets.

In practice

A forecast said a milestone would slip unless an owner was assigned by March. The owner was assigned in February and the milestone held. Resolving it honestly means recording that the condition was met and the prediction therefore untested, which is a different result from the forecast having been wrong, and a more useful one.

Evidence

What it cannot tell you

Resolution tells you whether a stated expectation matched what happened, not why the forecast was made or whether the reasoning behind it was sound. A forecast can resolve correctly for the wrong reason, or incorrectly despite good reasoning, and resolution alone does not distinguish a good process from a lucky or unlucky outcome.

Questions

The party who made the forecast, with the party who acted on it, since both hold information the other lacks. The Good Judgment Project scored forecasters against outcomes for exactly this reason; resolution by only one side loses either the generosity or the reasoning. Doing it together produces the usable record.

It is recorded as ambiguous, with what made it so. That is information about how the forecast was specified, usually that the expected outcome was too vague to resolve, and it is more useful than forcing the result into a binary that nobody actually believes.

That is recorded too. A forecast that prompted the action which prevented the predicted problem is a success of the process and an untested prediction, and those are different things. Collapsing them makes the track record look better and teaches nothing.

Immediately, because memory and context decay fast and attention has already moved on. The Wikipedia entry on forecasting, updated in 2026, notes that forecasts can later be compared with actual outcomes; a comparison made months late is reconstructed rather than observed, shaped by already knowing how things turned out.

Because by the review date the situation has moved on, nobody's objectives include closing the loop, and the exercise can only produce an uncomfortable result. It is unowned administration with a downside, which is a reliable recipe for something not happening.