Two teams present different numbers for the same thing, and the meeting becomes a discussion about whose report is right. The usual response is to buy something — a warehouse, a semantic layer, a BI migration — on the theory that the tooling will settle it.
It will not, because the disagreement is not technical. Sales counts bookings, finance counts recognised revenue, and the board wants ARR. All three are correct. Nobody wrote down which one the word “revenue” refers to.
Have the argument once, deliberately
The definition conversation happens either once, on purpose, in a room — or endlessly, by accident, at the start of every meeting. There is no third option where it does not happen.
Run it properly. Take one metric, get the people who use it in the same place, and make them state their definition out loud. The disagreements surface immediately and they are usually about legitimate differences in what each function needs, not about anyone being wrong.
Then decide. Not consensus — a decision, by someone with the authority to make it, recorded with the reasoning. Consensus-seeking on metric definitions produces a compromise nobody uses.
Name the variants instead of forcing one
The instinct is to declare one true definition. That usually fails, because the finance figure and the sales figure both genuinely need to exist.
The workable answer is distinct names. Not “revenue” three times, but bookings, recognised revenue and ARR, each defined once and each used consistently. The problem was never that three numbers existed; it was that they shared a word.
Once they have separate names, a report showing all three is informative rather than confusing.
Write down the parts people forget
A definition that says “number of active users” is not a definition. The detail that causes disputes lives in:
- The time window. Active in what period, and is it rolling or calendar?
- Exclusions. Internal accounts, test data, refunded transactions, cancelled-then-reinstated records.
- Timezone. A day boundary in UTC and one in Eastern produce different daily numbers, permanently.
- Restatement. When historical data changes, does the past number change with it? Both answers are defensible; only one can be true.
- The source of record. Which system wins when two disagree.
Most metric arguments are actually arguments about one of these five, conducted without either party realising it.
Give each metric an owner
Someone has to be accountable for the definition — able to approve a change, expected to review it periodically, and named on the documentation.
Without an owner, definitions drift silently. A field changes meaning upstream, a filter stops matching a new product line, and the number quietly becomes wrong while continuing to look plausible. Ownership is what turns a definition from a document into something maintained.
Put the definition where the number is
A glossary in a wiki is read once. The definition needs to be reachable from the number itself — a tooltip, a link beside the chart, a note in the report footer.
The test is whether someone questioning a figure in a meeting can find its definition in under ten seconds. If not, they will assume their own, and you are back where you started.
Expect to revisit it
Definitions are not permanent. Businesses launch products, change contract structures, enter markets that break an assumption baked into the calculation.
Review the important ones periodically and, when one changes, say so explicitly rather than letting the number shift and hoping nobody notices a step in the chart. An unannounced definition change costs more trust than the original ambiguity did.
We build reporting where the definition travels with the number — see data intelligence and analytics, or read why your dashboard is not being used.
Common questions
Why do teams report different numbers for the same metric?
Because nobody wrote down which definition the word refers to. Sales counts bookings, finance counts recognised revenue, the board wants ARR. All three are correct; the problem is that they share a word.
Should we force one definition for every metric?
Usually not. Give the variants distinct names instead — bookings, recognised revenue, ARR — each defined once. A report showing all three is informative rather than confusing once they are named separately.
What does a complete metric definition include?
The time window and whether it is rolling or calendar, exclusions such as internal or test accounts, the timezone of the day boundary, whether history restates when source data changes, and which system wins when two disagree.
