Finance and Sales report different revenue for the same quarter. Both are convinced they are right. You are asked to fix the data. Where do you start?

Assume neither number is wrong

The instinct is to look for a bug. In practice the overwhelming majority of these cases are two defensible definitions with no mechanism for choosing between them — and starting a data quality investigation wastes a week before arriving there.

Trace both figures to their definitions

For each report, establish: which tables, which filters, which time basis, which currency handling, which exclusions. The differences are usually mundane and individually reasonable:

• Timing. Booking date versus invoice date versus recognition date. At a quarter boundary these differ materially.
• Scope. Does revenue include shipping? Tax? Intercompany? Cancelled orders refunded after the period?
• Currency. Rate at transaction, at period end, or an average.
• Adjustments. Credit notes applied to the original period or the current one.

Present the reconciliation as a bridge — start at one number, apply each difference with its value, arrive at the other. That converts an argument about competence into a list of four decisions.

Get the decisions made, by the right person

Each difference needs an owner who decides — someone in the business, not in the data team. For revenue that is almost always Finance, and their definition is likely the one with regulatory weight behind it.

Sales may have a legitimate need for a different measure. That is fine, and the resolution is that it gets a different name — bookings, or gross sales — not that it stops existing. Most of these disputes are two metrics sharing one word.

Fix it structurally or it recurs

The reconciliation solves this quarter. What stops the next one:

1. Define each metric once in a semantic layer, with its filters, time basis and allowed dimensions. Both reports consume the definition rather than reimplementing it.
2. Name the owner in the glossary, and bind the glossary entry to the implementation so they cannot drift.
3. Certify the authoritative report so a reader can tell at a glance which one is quotable.
4. Search for other consumers of the same word. There are usually five, and three of them are spreadsheets.

What to tell the executives

Not "the data was wrong". Both figures were computed correctly from different and defensible definitions, the difference is explained by four specific choices, here is the reconciliation, and here is who now owns each choice. That is a governance outcome rather than a blame outcome, and it is the version that gets the structural fix funded.