Define every metric
Revenue may mean paid invoices, issued invoices or completed orders. Agree on the meaning, currency, time zone and treatment of refunds. Name an owner who approves definition changes. Two charts can draw from accurate records yet contradict each other when their calculations or reporting periods describe different things.
Show freshness and missing information
Display the last successful update and flag delayed sources. Distinguish zero from unavailable data. If an integration fails, show an explanatory state rather than manufacturing a convincing trend. Let authorised users inspect underlying records. A useful dashboard makes uncertainty visible instead of encouraging decisions based on a falsely complete picture.
Choose charts around decisions
Use time series for trends and labelled comparisons for categories. Keep the selected period and filters near results. Avoid combining incomparable currencies or hiding small samples. Ask which action the viewer can take next. If a chart cannot support a useful question, simplify it or remove it from the overview.
Reconcile before publishing
Compare totals with source records for a known period. Test date boundaries, duplicate imports, cancelled transactions and permission restrictions. Document acceptable delays and how discrepancies are investigated. Use genuine data or explicitly labelled empty states; polished design cannot substitute for a clear definition and a reproducible calculation.
Example: why two revenue charts can disagree
One team may count invoices when issued while another counts money when received. Both can be valid views, but labelling both simply as revenue invites confusion. Add the definition, time period, currency and treatment of refunds near the metric or in an accessible explanation.
If a manager cannot trace a total to its source records, the chart needs investigation before it becomes a basis for action. This example explains a reporting issue; it is not a claim about a particular customer.
Create a metric card before creating the chart
- The question the metric answers.
- Its source system and calculation.
- Filters, time zone and refresh frequency.
- Known exclusions and handling of missing values.
- The owner who investigates discrepancies.
Keep definitions available to users. When a calculation changes, record the date and decide whether historical periods need to be recalculated so trend comparisons remain meaningful.
Review usability and data integrity together
Ask a user to answer a real business question, change the period and inspect an underlying record. Check that exports respect the same permissions and filters as the screen. Include empty, delayed and failed-source states in testing.
If a visually attractive chart hides stale data, the design has failed its purpose. Prefer an honest unavailable state and a clear next action over filling gaps with invented values.
