Start with the definitions. A finance report and an acquisition cohort can differ legitimately because they group events by different dates, statuses or customer populations.
Write a metric contract
For each headline measure, record the event source, included statuses, period, time zone, currency and adjustment policy. Identify the team responsible for the definition.
| Measure | Boundary to resolve |
|---|---|
| Stakes and payouts | Accepted or settled activity, including treatment of voids |
| GGR | The exact calculation and product-specific adjustments |
| Bonus cost | Awarded, consumed, converted or another defined measure |
| Deposits and withdrawals | Requested, approved, processed or reconciled payments |
| Retention cohort | Entry event, qualifying return activity and observation window |
Do not force different business questions into one label simply to make two charts agree.
Decide how late corrections appear
Suppose an illustrative €100 settlement is corrected the following day. An event-date report may revise the original day, while a posting-date report may show the adjustment on the second day. Either approach needs a clear definition and a path between the two views.
Retain the original record, correction and relevant dates. A dashboard that silently changes a historical total can make reconciliation and investigation harder even when the latest figure is correct.
Accept reports against source records
Select a representative period and reconcile the headline figures to transaction exports. Include voids, late settlements, refunds and currency conversion. Check both totals and individual examples so that offsetting errors do not disappear inside a matching sum.
Verify role access and the detail an operator can inspect or export. Agree refresh expectations and how incomplete or delayed data are identified. A dashboard that looks current should not imply that every source has finished reporting.
Build the report around a decision
Choose a concrete use case: close a day, investigate a payment discrepancy, review a product change or compare a cohort. If the underlying concern is an unexplained result, a performance audit can define the investigation before dashboard work begins.
Send the reports that disagree and their definitions. Include the decision each report is intended to support and an anonymized example of the discrepancy.