“The map is not the territory.”
— Alfred Korzybski
In a quarterly review, the Head of Marketing puts up a dashboard: customer sentiment at an all-time high. Minutes later, the Head of Support puts up another: churn and ticket volume climbing fast. Both charts are mathematically correct. Both leaders call themselves data-driven. And they flatly contradict each other, while the room sits there with no way to choose. So the CEO reaches for the obvious fix. Give me a single source of truth. That instinct, reasonable as it sounds, is the mistake.
Here is what we get wrong. When a number walks in, it arrives dressed as objectivity, so the room treats it as a referee, a neutral party that has blown the whistle and named the winner. But a number was never a referee, and it never can be. The reason is not that data is biased, which would imply a clean number waiting underneath the bias, a flaw you could scrub off. There is no clean number underneath. A number is not found in the world. It is built. We say data, from the Latin for a thing given, when the honest word would be capta, a thing taken. The world holds phenomena, heat, anger, movement, loyalty, and a number is what you get when you trap one of them in a container someone designed. Every side of that container is a human choice.
That container has four walls, and each one bends what the number can mean. The first is definition. To count a thing, you have to draw a hard line around a fuzzy idea, and active user means one growth curve if it counts anyone who opened the app, and a much smaller one if it counts anyone who finished a task. Same word. Different world. The second is the instrument. A sensor often reports its own state as much as the thing you care about, the way a thermometer left in the sun tells you about the sunlight, not the air. The third is the sample. You never see the whole, only the hands that were raised, and a satisfaction score built from the customers who stayed is deaf to everyone who already left in disgust.
The fourth wall is the dangerous one, because it changes the very thing being measured. Measure a human system, and the system pushes back. When a measure becomes a target, it stops being a good measure. A call centre judged on calls per hour does not get more efficient. It gets agents who hang up on hard problems and pass difficult customers along to stop the clock. The number climbs while the service rots, and the dashboard glows green at the exact moment the experience falls apart, because the mess has been shoved just outside the frame the metric can see.
Play that out, and the metric quietly eats the strategy. We stop working for the mission and start working for the dashboard, and at scale the number becomes a shield. The data made me do it is the most convenient defence an executive has, because it launders a value judgment into an apparent fact. You chose to weight this quarter’s margin over next year’s culture, and then dressed that choice as arithmetic. A number can only describe what is. It can never decide what ought to be done. So the numbers decided is always a small lie. No number has ever decided anything: people decide, and then reach for a number to carry the blame.
So what is the fix? Not to distrust every number, which only swaps one error for another. The fix is to stop treating the number as a referee and start treating it as a witness, one you cross-examine before you rely on its testimony. You audit the body that produced it.
You might say this is a loaded gun. If every number is constructed and every wall distorts, then no number can be trusted, decision becomes impossible, and worse, we have just handed every leader a licence to cross-examine any finding he dislikes into silence and crown his own preference as the verdict. The danger is real, and it is the exact mirror of the error we started with. But cross-examining a number is not dismissing it. A measurement whose construction you understand is more trustworthy than a spotless one you do not, because you know precisely where it is sharp and where it is blind. You interrogate a number in order to judge with it, not to throw out whatever you would rather not see. The point is to put the judge back in the chair, not to empty the witness stand.
So before any number is allowed to close a decision, walk its body. Start with the recipe: the definition written as steps precise enough that two people applying it would land on the same count, because until we agree on the recipe, we are arguing about different things in the same words. Check the instrument: is the sensor measuring the thing, its own state, or the user’s performance for the sensor? Check the sample: does it represent the population, or only the hands that were raised, and who is missing by design? Check the incentive: a number tied to a bonus or a reputation is testimony from an interested witness, and you weight it as such. Check the reaction: once this number carries consequences, what behaviour will bend to move it, and will that hollow out the thing it was built to track? And check the reversal: what evidence would have changed our mind, because a number that was never allowed to be wrong was never evidence. It was decoration for a decision already made.
A chart is not a verdict. It is a witness statement, taken from one vantage point, through one instrument, about one slice of the world, shaped by whoever built it and whatever they were paid to find. So when the chart enters the room and the argument goes quiet, do not go quiet with it. Ask how the number was built. Then pick it up and judge with it, the way a fair judge weighs a witness, and never once let it judge for you.
Decision Rule — The Sensor Audit
Before a number is allowed to drive a decision, audit the body that produced it, because a number is built, not given:
- The recipe. How did we define the boundary of this count, and would two people applying our definition reach the same number?
- The instrument and the blind spot. Is the sensor measuring the thing, its own state, or a performance for it, and who is invisible by construction?
- The incentive and the reaction. Whose interest does this number serve, and once it carries consequences, what behaviour will bend to move it while the real good decays?
- The reversal. What would have changed our mind, and did we ever give reality the chance to?
A number that survives these has earned its weight. One that cannot should open the conversation, not end it.

Leave a Reply