Agree what the number means.
Two teams can report different revenue, stock or margin figures without either making a calculation error. They may be using different dates, statuses, entities or exclusions. A visual redesign will not resolve that disagreement. The first task is to define the question and the measure.
We work back from the disputed result. If sales and finance disagree about revenue, we compare the reporting date, transaction status, entities and exclusions before designing another dashboard. The output needs to answer a business question and give its owner a way to investigate a difference.
Build confidence in the reporting chain.
Definitions and reconciliation
Document the measure, filters, timing and treatment of exceptions. Agree a trusted comparison or reconciliation so the team can explain differences rather than debate screenshots.
Useful levels of detail
Give a manager the summary needed to see a pattern, with a route to investigate the underlying transactions. Avoid combining metrics that answer different questions simply because they fit on one screen.
Access and distribution
Consider who may see the information, how it is shared and whether exports create additional handling requirements. Sensitive data needs appropriate access and an accountable owner.
From a business question to a maintained view.
- 01
Frame the decision
Describe the action the report should support and the audience using it. Identify frequency, timeliness and the level of detail required.
- 02
Trace and validate the data
Review source records, definitions and transformations. Investigate missing or inconsistent inputs before treating a calculation as authoritative.
- 03
Design and test the view
Build an appropriate search, report, dashboard or analytical output. Test representative periods and exceptions with the people who understand the underlying business.
- 04
Assign ongoing ownership
Document definitions and change triggers. Review the output when processes, entities, modules or integrations change so it remains relevant.
Choose the reporting tool after the question is clear.
A role-specific operational queue may fit a saved search; a recurring financial view may belong in a report. A question that depends on several systems may need a separate analytical layer. We assess the sources, calculations, access and refresh requirements before choosing the presentation.
AI can make a well-defined dataset easier to explore, but a fluent explanation does not reconcile the number. A proposed analytical assistant needs questions with accepted answers, a route back to the source and a review of ambiguous requests. Where a conventional view is easier to verify, it remains a useful option.
Be precise about data freshness
- Can a dashboard show real-time information?
Freshness depends on the source, transaction timing and any integration or refresh process involved. State the actual update behavior and its limitations. A label such as real time should only be used where the complete data path supports it.