One Big Number Hides the Argument. Break It Into Small Ones.

Confidence: hypothesis. The failures below are drawn from published research on forecasting and clinical development, not firsthand deployment. The pattern at the end is an argument, not a shipped system. Argue with it.

Someone puts a number on the wall. This program has a 65% chance of success. The room nods or frowns, but nobody can really argue with it, because there is nothing to argue with. It is one number. You either believe it or you do not. That is not a discussion. It is a vote.

The problem is not the 65%. The problem is that a single number has no seams. You cannot see which part of it is solid and which part is a wish.

The clinical why

In drug development, the probability of success is never really one thing. It is several bets multiplied together. Does the drug's mechanism actually work in people. Does the trial, as designed, read out clearly. Does the regulator accept the endpoint and the evidence. Does the product scale in manufacturing and reach the market. Each of those is its own question, with its own evidence and its own track record.

When a team reports one blended number, all of that collapses into a single digit. The number might be 65% because every factor is a solid 90%, or it might be 65% because three factors are strong and one is a hope someone rounded up. From the outside, those two look identical. The blend hides exactly the thing a reviewer needs to see.