Why two dashboards give different numbers for the same thing.

Write down, in one sentence each, what the two reports think they are measuring. In most cases the disagreement resolves here, before anybody opens a query.

Revenue is the classic example: booked, invoiced, collected, and net of refunds are four different numbers, all correctly called revenue by somebody.

If the definitions match, the next thing to compare is the filters, and specifically the ones nobody wrote down: excluded test accounts, an internal customer, a region that migrated systems last year.

Order date, ship date, invoice date and payment date all produce plausible and different monthly totals from the same transactions. Any report that does not say which it uses is one of the two reports in this argument.

Timezone moves everything that happened near midnight into a different day, and at month boundaries that becomes a different month. It is a small effect that produces persistent, unexplainable differences of a few tenths of a per cent, which is exactly the size that wastes the most time.

A report built at nine and one built at five are looking at different data, and if either caches, the difference persists after both appear to have updated.

Late-arriving rows compound this. Refunds, adjustments and corrections attach to earlier periods, so a figure for last month legitimately changes for weeks afterwards. A report that says when it was computed removes most of this class of argument permanently.

The instinct once the cause is found is to change one report to match the other. That is usually the wrong repair, because it leaves two definitions in the organisation and one of them now silently mirrors the other.

Write the definition down where both reports can reference it, decide which one is correct for which audience, and let them differ openly if they genuinely should. Finance and operations often need different revenue figures, and the problem was never that the numbers differed: it was that nobody could say why.