The deck reached the slide everyone had been waiting for, and the number on it had two decimal places. Not a quarter, not somewhere north of a fifth, but a figure carried out past the point where any input in the building could have justified it. The room read the decimals the way a congregation reads a line of Latin: not quite understood, but plainly the work of someone who knew what they were doing. The questions that had been circling the table for the previous half hour went quiet. A number that specific does not invite interrogation. It invites note-taking.
Nobody asked the one question that mattered, which was never how the number had been calculated but whether the thing it described could be known to two decimal places at all. The arithmetic was impeccable. The spreadsheet balanced. The figure had survived three rounds of review by people checking that the formulas pointed at the right cells, and not one round by anyone asking whether the cells pointed at the right world. The precision was real. The accuracy was undiscussed. In the gap between those two words, which the room had long since started treating as synonyms, the decision got made.
Precision and accuracy are not the same property, and the distance between them is exactly where this kind of mistake lives. Precision is a fact about the number: how many decimal places it carries, how tight its stated range. Accuracy is a fact about the relationship between that number and reality, whether it is actually anywhere near the truth. A figure can be exact to the eighth decimal and wrong by an order of magnitude, and nothing about the way it looks on the slide will tell you which one you are holding. Precision is manufactured inside the model. Accuracy has to be earned outside it, against a world that was never in the room and did not get a vote.
Uncertainty goes in soft and comes out hard.
The false precision is rarely a lie. It is the output of a process that quietly loses its honesty on the way through. Every input to that model started life as a guess, a judgment call, a figure someone was fairly but not fully sure of. Each one got entered as a single confident value, because spreadsheets have no column for roughly, and the caveats that lived in the analyst's head did not survive the paste into cell B7. The model then multiplies these point estimates together, and the result inherits all the decimal places of the arithmetic while shedding every doubt that went into the arguments. A stack of maybes walks in the front and a fact with a decimal point walks out the back, and the decimal point is now load-bearing in a way that no one in the building actually authorized.
There is a reason the precise number wins the room, and it is not that it is more likely to be right. A range invites argument; a point estimate ends it. The honest analyst who says the answer is somewhere between two uncomfortable figures, and frankly could be worse than either, sounds like a person who has not finished the work. The colleague who states one number to two decimals sounds like a person who has. The room, which wants to stop deliberating and start deciding, reaches gratefully for the one who will let it. We tell ourselves we reward rigor. What we actually reward is the willingness to hide the rigor's uncertainty behind a decimal point, and we have been promoting for that willingness, quietly, for years.
Garbage rounded to four decimal places is still garbage. It just no longer looks like garbage. It looks like diligence.
The cost is not the occasional wrong number. Organizations absorb wrong numbers all the time. The cost is that the precise figure launders a decision nobody actually examined, and then stands in the record as proof of a scrutiny that never happened. When the forecast misses, the post-mortem studies the model, hunts for the broken formula, refines the methodology, and resolves to add another decimal place next time. It almost never finds the real failure, because the real failure was not in the arithmetic. It was in the room's inability to tell a confident number from a knowable one, and that particular defect does not show up in any cell.
You hired for the nerve to round.
Here is the part that should sting. The number was not the problem. The decimals were not the problem. The problem is that the organization had built a culture that cannot separate confidence from competence, and had been selecting for the first while congratulating itself on measuring the second. Every meeting that rewarded the clean point estimate over the honest range taught the building the same lesson, that the way to be trusted here is to state what you cannot know as though you know it. You did not hire for judgment. You hired, promotion by promotion, for the nerve to round, and then you wondered why nobody in the room would tell you how uncertain they really were.
Rebuilding this does not require less quantification. It requires numbers that are made to carry their own doubt. A point estimate should not be allowed to travel alone; it moves with the range around it or it does not move at all. Every consequential figure should arrive with the one sentence the model cannot generate on its own: what would have to be true for this to be wrong, and how would we find out before it was too late to matter. The operative question stops being what is the number and becomes how much of this do we actually know, which is a harder question, a slower one, and the only one that was ever worth the meeting.
- Ban the naked point estimate. If a figure cannot state the range around it, it is not ready to be decided on, and a range that is embarrassingly wide is information, not weakness.
- Ask every forecast what would have to break for it to be wrong. A number whose author cannot answer that has not been thought about. It has only been computed.
- Reward the honest range over the confident point out loud, in front of the room, until naming your uncertainty stops being the thing that loses you the meeting.
This is precisely the reflex the organizational crisis simulation is built to expose, because it is invisible inside any real meeting. Put a leadership team into a cascading crisis, hand them numbers that look authoritative and are quietly built on sand, and watch who grabs the precise figure to close the discussion and who stops to ask what the figure actually rests on. Under a running clock, with a decision owed, most leaders do not interrogate the clean number. They inherit its confidence and move. Which of your leaders questions an authoritative figure when questioning it is the slow and unpopular thing to do is not something a quarterly review will ever surface, for the plain reason that the review is reading the same laundered numbers the meeting did.
A number no one in the room can question is not a measurement. It is a decision that has already been made, wearing the clothes of one that hasn't.