A judgement with no review date cannot be wrong, which is why it also cannot improve anything.
Why it matters when the plan changes
Without a date, a judgement is revisited when someone remembers it or when it has already failed. Both are too late to be useful and neither produces a record. A forecast, by definition, is a prediction that can later be compared with actual outcomes, and that comparison needs a fixed point to happen against. Setting the date at the moment the judgement is made is the cheapest possible action, and the one that makes everything downstream, tracking, scoring, learning, possible.
The tension is that a review point invites being judged. A supplier or a leader who fixes the date is agreeing in advance to a moment when their call will be checked, in the same way the Good Judgment Project scores predictions once outcomes arrive. That is uncomfortable, and it is the entire mechanism: scored, feedback-driven judgement improves, and judgement left unscored simply repeats itself, however wrong it was.
In practice
A judgement says an unowned interface will delay a milestone unless an owner is assigned within six weeks, and sets the review for eight weeks out. At the review, the owner was assigned in week five and the milestone held. That result goes into the record, and the next similar call is made against something observed rather than remembered.
Evidence
Forecasts are predictions that can later be compared with actual outcomes, which requires a point at which the comparison is made.
Forecasting, Wikipedia (2026)Judgements that are scored against outcomes improve, and the scoring depends on the prediction having been fixed beforehand.
Superforecasting and the Good Judgment Project (2026)
What it cannot tell you
A review point tells you when a judgement will be checked, not what it will find. Setting one does not make the underlying call more accurate, and a badly specified judgement will produce an ambiguous or unscoreable result at the date rather than a wrong one that at least teaches something.
Questions
From when the evidence that would settle the call becomes available, not from the reporting calendar. If the judgement is that a milestone will slip without an owner, the review sits shortly after the point where the owner would have had to be in place.
The original judgement is compared with what happened, in the way the Good Judgment Project scores predictions against outcomes, and the comparison is recorded even where the judgement was wrong. The original view is kept rather than replaced, because the gap between the two sharpens the next call.
Yes, and it should when any of the six events fires: a new strategy, ownership, operating model, pressure, dependency or material new evidence. Each changes what the plan asks, which means the prior judgement is now about a situation that no longer exists.
That is recorded as it stands rather than resolved in either direction. An ambiguous outcome is information about how the judgement was specified, usually that it predicted something too vague to score, which is worth knowing before the next one is written the same way.
Because a date chosen afterwards is chosen with knowledge of what happened, which destroys the comparison, the same principle that a 2026 account of forecasting notes when it defines a forecast as a prediction that can later be checked against actual outcomes. Fixing the date in advance is what separates a forecast from a description written after the fact.