A Forecast With No Number Is Only a Mood
Sherman Kent wrote "a serious possibility" and meant 65 to 35. His board approved the same words and meant everything from 20 to 80. Words feel safe because they cannot be proved wrong.
In March 1951 the CIA published an estimate on whether the Soviet Union would attack Yugoslavia that year. Its key judgment said an attack "should be considered a serious possibility." Readers took that to mean one thing. The author of the phrase meant another.
The man behind it, Sherman Kent, wrote about what happened in a 1964 article in the CIA journal Studies in Intelligence. A few days after the estimate came out, the chairman of the State Department Policy Planning Staff asked him what odds he had in mind for "serious possibility". Kent said his own view was about 65 to 35 in favour of an attack. The chairman was jolted. He and his colleagues had read the words to mean odds very considerably lower.
Kent then asked the members of his own board what they had meant. They had all agreed to the same wording, and they had not agreed on anything behind it. One had about 20 to 80 in mind. Another had 80 to 20. The rest sat somewhere between.
Think about that. A group of experts reviewed a sentence, approved it, and each walked away with a different belief about what it said. It looked like agreement because the words were the same.
Most of us do this every week. We say a project is "likely" to finish on time. We say a risk is "unlikely". We tell a client something "should" work. Each listener hears their own number, and nobody finds out until something goes wrong.
Kent proposed a fix: choose a small set of words and tie each to a range of odds. Some of his colleagues pushed back hard. He called them the poets, and he called himself a mathematician. Their worry was that numbers give a false sense of precision, and that is a fair worry. But the alternative is a sentence that can mean 20 percent to one reader and 80 percent to another.
The habit is simple. When you write "likely" or "unlikely" in a report, add a number beside it, even a rough one. If I cannot defend a number, that is useful to know before the decision is made, not after. And if two people who agree on a word give me different numbers, we have found the real disagreement early.
Words feel safe because they cannot be proved wrong. That is the reason to be careful with them.