Why does one business have three revenue numbers?
Before reconciling totals, find out which question each number is answering.
Three reports, three purposes
A sales report may count orders when they are created. A management report may use completed orders after cancellations. An advertising report may attribute sales to an earlier click. Differences are not proof that one system is broken. Start by recording the event, date and status each source uses.
Write the definition before the query
For every revenue measure, specify the population of orders, timestamp and timezone, currency, and the treatment of tax, shipping, discounts and refunds. Name the business owner who can approve the rule. If the sales and finance teams need different measures, keep both with distinct names instead of forcing agreement on a misleading total.
A small worked example
Suppose a period contains 10 completed orders totaling 1,000 and a refund of 100. Under a completed-sales-less-refunds definition, net sales are 900. A created-order report could also include a cancelled order of 200 and show 1,200. The numbers differ because their populations differ. This example is a calculation pattern, not a client result.
Reconcile records instead of adjusting totals
Choose a short reporting period. Join exports using stable order identifiers and keep unmatched records. Classify differences as status, timing, currency, duplicated lines, missing orders or fee rules. Compare each group with source records. Do not add a balancing amount simply to make the totals agree.
Make the decision traceable
A report should include its date basis, refresh time and exclusions. Keep the reconciliation alongside the metric dictionary and repeat it after source or rule changes. The metric dictionary template provides a practical starting point for naming the formula, owner and validation check.
Put this into practice.
Prepare a metric dictionary, review a source export or discuss the report your team needs.
Continue reading
Five questions to answer before your first dashboard
A useful report starts with the decision someone needs to make.
Read the guideWhat does AI need to give useful answers about your data?
Schema, definitions and access boundaries matter as much as the question.
Read the guideA practical checklist before turning a CSV into a report
Catch simple file problems before they become persuasive charts.
Read the guide