4 MIN READ · PRACTICAL GUIDE
Two dashboards can show different revenue totals without either containing a calculation error. One may count orders, another issued invoices and a third cash received. Before redesigning the charts, agree on the question each number is supposed to answer.
A metric contract is a short shared definition of a measure. The term here describes an internal working document, not a legal agreement or a particular software product.
Define the decision before the chart
Ask what someone will do when the number changes. A cash-collection measure might guide a credit-control queue. A sales-order measure might guide capacity planning. Calling both “revenue” invites confusion, particularly when they use different dates and currencies.
Write down the population included, the event being counted, the date used, the currency treatment and the exclusions. Identify the source system and the refresh schedule. A dashboard should make late data visible rather than quietly implying that an incomplete period is final.
The FinOps Foundation’s Reporting and Analytics capability emphasises making cost and usage information useful to stakeholders. The broader practical lesson for a business dashboard is that a number must be understandable in its decision context, not merely available in a database.
An illustrative metric contract
Consider a business with UK and euro-area sales. The following example defines an operational invoicing measure. It is not a statutory accounting definition and should not be used to determine financial reporting or tax treatment without appropriate advice.
| Field | Example definition |
|---|---|
| Name | Issued invoice value, net of posted credits |
| Purpose | Monitor invoicing activity by operating entity |
| Included records | Posted invoices and posted credit notes |
| Excluded records | Drafts, cancelled documents and tax amounts |
| Period | Document posting date within the reporting month |
| Currency | Show original currency separately; use an agreed documented rate for consolidation |
| Owner | A named finance process owner |
| Freshness | Last successful load time displayed beside the measure |
Suppose one entity issues €100,000 of invoices and €8,000 of credit notes in the period, both excluding tax. The operational measure is €92,000. If it receives €70,000 in customer payments, that does not make the invoicing measure wrong: cash collection answers a different question.
Do not add a sterling amount to the euro figure without a defined conversion method. Keep the original amounts and the applied rate so that someone can reproduce the consolidated result. Consistency in the method matters more than the visual precision of showing extra decimal places.
Build checks that explain discrepancies
Reconcile the source record count, amount total and excluded records separately. This distinguishes missing data from a definition difference. If the dashboard is lower than the source system, check whether the refresh failed, whether the date boundary differs or whether credits were included on one side only.
Include a small set of deliberate edge cases in testing: a credit issued in the following month, a cancelled document, a late-posted invoice and a transaction in another currency. Ask the metric owner to explain the expected treatment before the data engineer implements it.
Show freshness and provisional status next to the number. A warning hidden in a separate technical report will not help a manager who sees only the dashboard. Decide what happens when the data is stale: keep the last valid value with a warning, suppress it or stop the downstream action, depending on the business consequence.
Give the definition an owner and a change history
A metric should not change silently when a developer improves a query. Record the reason for the change, its effective date and whether historical periods have been restated. Otherwise a trend can appear to improve because the definition changed rather than because the business improved.
Start with the few measures that guide regular decisions. A smaller dashboard with agreed definitions is more useful than a wall of charts whose owners cannot explain the inputs. Avoid collecting personal or sensitive detail merely because a drill-down feature makes it possible.
When buying a reporting platform, ask the supplier to demonstrate reconciliation and access controls using your sample data. The ERP selection guide applies the same evidence-first approach to broader system purchases.
Next step
Choose one disputed total and write its metric contract on a single page. Have the process owner and data owner agree on the edge cases. Only then decide which chart best communicates it.
Source and scope
Primary framework checked 14 November 2026: FinOps Foundation: Reporting and Analytics. The metric contract, invoice example and implementation checks are our own illustrative business-analysis framework.