A dashboard metric definition turns a familiar label into an agreed calculation. Revenue, active customers and on-time delivery can mean different things to different departments. If those differences remain implicit, a polished chart may create disagreement instead of clarity.
For a hypothetical distributor, sales may count an accepted order while finance counts an issued invoice. Both views can be useful, but calling them the same metric hides the difference. A metric dictionary makes the intended decision and underlying calculation visible.
Start with a decision
Ask what someone will do when the value changes. A purchasing manager may need to investigate stock risk, while an executive may need a monthly trend. Their required detail, freshness and comparison period differ.
Write the metric's purpose in a sentence before selecting its visual form. If the team cannot explain an action or question the metric supports, consider whether it belongs on the first dashboard at all.
Record the calculation contract
Give each metric a stable name, plain-language definition, formula, unit and owner. Specify the included records and important exclusions. Define whether a cancelled order, partial delivery or later refund changes the original period.
| Definition field | Example question |
|---|---|
| Grain | Is each row an order, item or shipment? |
| Time boundary | Which timezone and business calendar apply? |
| Exclusions | How are test records and cancellations handled? |
| Denominator | Which eligible events are counted? |
| Owner | Who approves a changed definition? |
Keep example records with expected results so reviewers can check the definition without reading a large query.
Understand the model beneath the chart
Microsoft's star schema guidance explains modelling concepts relevant to Power BI. Our recommendation is to inspect the actual relationship between transactions and dimensions before adding totals; duplicated joins can create a convincing but incorrect result.
For instance, joining an order to several delivery events may multiply its value. Decide whether the metric requires order-level aggregation, delivery-level aggregation or a separate bridge. Document the choice in terms the business owner can validate.
Specify freshness and incomplete data
Show when the data was last updated and what period is complete. A day's sales at midday should not be compared with a completed prior day without explaining the difference. Late-arriving records may require a restatement policy.
Use a visible warning when a source feed is delayed or missing. Reporting zero activity for an unavailable source creates a false operational signal. Keep freshness status separate from the metric's business result.
Make the definition accessible
Readers should be able to find the formula and context near the chart. Avoid relying on colour alone for performance status. The W3C accessibility overview is a reference for reviewing how people perceive and operate the dashboard; the metric's wording still needs editorial judgment.
Govern changes deliberately
Version a definition when it materially changes. Record the reason, effective date and whether history was recalculated. Do not quietly change the denominator and present the new line as directly comparable with previous reports.
Work through a deliberately small example before release. In a hypothetical ledger, eight of ten delivered orders arrived on time, while two additional overdue orders remain open. A delivered-order measure gives eight out of ten; an eligible-order measure might include the open orders and give eight out of twelve. Neither should appear under an ambiguous shared label. Agree the decision each measure supports and preserve the sample as an acceptance check.
Read the data quality guide and executive KPI guide. Our data and dashboard team can help turn competing spreadsheet definitions into a reviewable reporting model.