Everyone has been in that meeting. Marketing says 4,200 new customers last month. Finance says 3,800. The product team says 4,600. For twenty minutes people defend their spreadsheets, and then someone says they'll "align offline" and nothing changes.
Nine times out of ten, nobody is wrong. They are counting different things and calling them by the same name. The fix is dull and it works: write down what each metric means, once, in one place.
What goes wrong without a definition
"New customer" sounds simple. Is it someone who placed their first order? Someone who placed it and didn't return it? Someone who signed up? Does a guest checkout count? What about a customer who bought three years ago, vanished, and came back last month?
Each team picks the answer that suits the system they work in. Marketing counts from the ad platform, which counts a click as a conversion. Finance counts from the payment system, which only knows about completed payments. Product counts from the app, which knows about accounts. All three numbers are accurate, and all three are different.
A format that works
I keep a definition to about six fields. It fits on a card, and nobody has to read an essay.
- Name. The exact words people use in meetings. If people say "active users," call it that.
- What it measures, in one sentence. Plain language, no jargon.
- How it's calculated. The formula, written so that someone could reproduce it. Include what is excluded: test accounts, refunds, internal orders.
- Where the number comes from. The system or table. If there are two possible sources, pick one and say why.
- Who owns it. One person who answers questions and approves changes.
- Time window and refresh. Calendar month or rolling thirty days? Updated nightly or weekly?
An example, for a store:
New customers. The number of people who completed their first paid order in the period. Guest orders are matched to an existing customer by email address. Orders cancelled or fully refunded within thirty days are excluded. Source: the orders table in the warehouse. Owner: head of growth. Calendar month, refreshed daily.
You can argue with that definition. That's the point. An argument about a written sentence takes ten minutes and ends. An argument about a number takes a week and doesn't.
Getting people to use them
Writing the document is the easy part. A few habits help.
Put the definition next to the number. If your dashboard shows "New customers," make the definition appear on hover or in a note below the chart. People won't go looking for a wiki page.
Start with ten metrics, not a hundred. Choose the ones that appear in leadership meetings. You can add more later. A fifty-page dictionary that nobody has read is no better than none.
Have one place to change them. Decide who can edit a definition and how others find out. When a definition changes, note the date. Otherwise last quarter's chart quietly stops matching this quarter's, and you'll have the argument all over again.
Reconcile once, visibly. The first time you publish a definition, show how the old numbers map to it. "Marketing's 4,200 was counting form sign-ups. Finance's 3,800 matches the new definition." That one table does more to build trust than any amount of explaining.
When there are genuinely two definitions
Sometimes two definitions are both useful. Marketing may really need "sign-ups" and finance really needs "paying customers." That is fine. Give them two names. The trouble starts when both are called "customers."
The unglamorous bit
None of this is clever. A shared document with six fields and a person responsible for it. But I have seen it save more time than any new tool, and it's usually the first thing I do when a team tells me their data can't be trusted. Often the data is fine. It just hasn't been introduced to the people reading it.