One Number, One Definition, One Owner

Two dashboards show a different number under the same word, and a meeting gets spent deciding which one is correct. Almost always, both are. They compute two different quantities that were given the same name, and the room argues arithmetic when the problem is vocabulary.

This framework is three rules. Every number that matters has one definition, that definition has one named owner, and no two definitions may share a name.

Why this exists

The reconciliation instinct is the trap. Somebody volunteers to get to the bottom of it, spends two days diffing queries, and comes back with the answer: one includes canceled orders and the other does not. That answer is correct, satisfying, and changes nothing. Next quarter a third dashboard appears with a fourth treatment of cancellations, because nobody was ever told which treatment was supposed to win.

The durable fix is not picking a winner. Both numbers usually deserve to exist — finance needs one treatment, operations needs the other, and forcing either side to adopt the other’s rule only relocates the argument. Rename both, and retire the bare word. “Revenue” stops being something anyone may put on a chart. There is booked revenue and there is recognized revenue, each with its own sheet, and a panel titled only “revenue” becomes a defect you can point at without reading a line of the query behind it.

Ownership has a similar edge on it. The owner is not the person who built the pipeline. It is the person who will stand in front of whoever acts on the number and explain why it moved. Those are different people often enough that naming the wrong one breaks the rule quietly: an engineer who owns a definition will update it when the source system changes — correctly, reasonably, and without telling anyone — and the person making decisions on that number finds out months later, from a trend that was never real.

How to use it

Build the list from meetings, not from the warehouse. A number is in scope if it gets said out loud in a recurring meeting or if it triggers an action. That usually lands between twelve and twenty-five numbers. Everything else stays deliberately unowned, because a definition nobody consults rots faster than one nobody wrote.

Test a definition by handing the sheet to two people who have not spoken to each other and asking both to compute it. Then compare their row sets, not their totals. Matching totals is a weak test — two different definitions can land on the same figure for one period and separate the next. Matching rows means the sheet actually determined the answer.

Settle collisions with an arbitrary rule, on purpose. When a new number wants a name that is already taken, the newcomer renames and the incumbent does not move. It is unfair and it takes ten seconds, which beats a standing debate that recurs every time somebody builds a chart.

Where it goes wrong

It does not work on numbers that are under negotiation. Forecast, pipeline value, and anything with attribution in it are opinions in a numeric format. Give them an owner and skip the single-definition rule, or you will get one official definition and three shadow spreadsheets that people actually use.

It collapses under its own weight if you apply it to everything. At two hundred entries it is a catalog, and a catalog needs a review queue nobody is staffed to run. Hold the line at the meeting list even when somebody offers to do more.

The role dies quietly when the owner cannot say no. If anyone can ship a chart carrying a new definition without asking, you have documentation rather than governance. Watch for the wrong symptom: the failure signal is not conflict, it is silence. Six months with no change requests almost never means stability — it means the process got routed around.

And none of this makes a number true. A company can agree perfectly on a definition that measures the wrong thing. This framework ends the argument about which figure is right; it has nothing to say about whether the figure is worth having.