A forecast that cannot be wrong is not a forecast

"Things will change a lot." "Technology will transform the sector." "There will be turbulence ahead." Such statements are safe, because nothing that happens could ever contradict them. We think they should not be called forecasts at all.

A real forecast states something that could turn out true or false, gives a time, and, ideally, a probability. "There is a 70 percent chance that X will happen by 2030" can be checked. Over many forecasts, it can be scored: a measure such as the Brier score rewards forecasters whose stated probabilities match how often events occur.

Research on forecasting, including work associated with the Good Judgment Project, suggests that some people are consistently better than others and that practices such as breaking questions into parts, updating often and keeping score can improve accuracy.

The objection is that the most important futures are not predictable, and a rule that insists on numbers pushes people to ignore them. We agree that numbers are no substitute for thinking, but vague claims that avoid scrutiny teach nothing.

What would change our mind: evidence that precise, scored forecasts are no better than vague ones at supporting decisions.