Your data is not messy. Your definitions are.

Finance says one number. Operations says another. Both pulled from the same system, on the same morning, for the same month. So somebody's data must be wrong.

Usually nobody's data is wrong. The two teams are answering different questions with the same word.

The word is doing too much work

Take something that sounds unambiguous. An "active client".

• Finance means: has been invoiced in the last quarter.
• Operations means: has work scheduled this week.
• The sales team means: has not formally told us to go away.

All three are reasonable. All three are used daily. And none of them is written down anywhere, because to each team their definition is so obviously the definition that it does not occur to them it needs stating.

Then someone builds a dashboard, picks one of the three, and now the dashboard is "wrong" to two-thirds of the organisation.

One word, three meanings

• Invoiced recently
• Work scheduled
• Not yet lost
• Nobody is sure

None of those readings is wrong. That is exactly why the argument never resolves.

Why this gets misdiagnosed as a data problem

Because the symptom looks exactly like one. The numbers do not match. Numbers live in systems. Systems are IT's problem. So a cleanup project gets commissioned, someone spends a quarter deduplicating records, and at the end of it the two teams still report different numbers, because nobody changed the thing that was actually different.

It is an expensive way to discover you had a vocabulary problem.

What to do instead

Write the definitions down. That is genuinely most of it.

One page. One line per term that appears in a report anyone acts on. Each line says what the term means, who owns that meaning, and (the part everyone skips) what it deliberately excludes.

A definition without an exclusion is a wish. "Active client means a client who is active" tells you nothing. "Active client means invoiced in the last 90 days, excluding retainer credits" is a decision.

Do that before you migrate anything, before you buy anything, and certainly before you build a dashboard. The tooling is easy once everyone means the same thing. It is impossible while they do not.