How to say what you do not know

Not knowing comes in three forms and they need different sentences

There is the number you could measure better with more time or more money. There is the fact that exists and is disputed, where two credible sources give different answers and nothing will reconcile them by Thursday. And there is the thing nobody can know yet at any price, because it depends on a decision another party has not taken.

Board material flattens all three into the word “estimated”. They are not the same problem. The first is a budget question and the board can buy its way out. The second is a weighting question, and it belongs to the board rather than to the adviser. The third is a design question: structure the decision so it survives either outcome, and say that this is what you have done.

A single number is a claim about precision

When the underlying work does not support a point estimate, writing one is a statement the evidence cannot back. We write ranges, and next to the range we write how we produced it. A range with no method behind it is a point estimate wearing a coat.

Boards handle ranges better than advisers expect. What they handle badly is a number that looks solid in one pack and moves in the next with no explanation of what changed.

Say who is confident, and on what

“We are confident” is a sentence about us. The board needs a sentence about the evidence.

“Our confidence here rests on two sources, both supplied by the operator, neither independently verified” is a statement the board can act on. It tells a director how much weight the sentence carries under a test, and who could check it. It also protects the adviser, though that is the smaller reason to write it that way.

The sentence people find hardest to write

“We do not know. Here is what it would take to find out, and what that would cost. Here is what we would do in the meantime.”

Advisers expect that sentence to damage their standing in the room. In my experience the meeting gets easier after it, not harder. The board stops testing the adviser and starts deciding what the missing answer is worth, which is a decision only the board can take.

Confidence expires

A statement of confidence without a date is unfinished. The data was good in April. The regulator’s position was stable when we last checked it, and the competitor had not moved by then either.

We date the evidence in the text, not in a footnote, so the reader can see the age of the ground she is standing on. It takes a few words. It saves the argument where one person is quoting a fact that has changed and nobody in the room can tell.

The audit reads the limits first

Two years later, the reader is not checking whether you were right. She is checking whether you knew what you did not know, and whether you said so at the time.

A document with a clean, confident surface fails that reading. The reader can see in hindsight the uncertainty that was present and never made the page. A document that names its limits survives even when the decision turned out badly. That is the part clients find counterintuitive. The record defends the reasoning, not the outcome. Reasoning can be sound and still lose.