Two teams can count the same activity differently without making an arithmetic mistake. One may report appointments scheduled this week; another may report appointments completed this week. Both labels can sound like “weekly bookings” in a meeting. A shared measurement definition turns that ambiguity into a decision the business can review. It also gives the data and design teams something concrete to implement and test.
On this page
Connect the dashboard to a business decision
Ask who will use the information, what they need to decide and how frequently they need it. Planning staff coverage calls for a view of upcoming demand; reviewing delivered services calls for completed activity. Those are different questions and may need separate measures. Specify the comparison that gives each number meaning, such as another equivalent reporting period. Also identify the detail users need when a total changes. This helps prevent a dashboard from accumulating attractive charts that do not answer the team’s questions.
- Name the audience and the decision.
- Separate upcoming demand from completed work.
- Define the reporting period and useful comparison.
Write a definition and test it with sample records
Illustrative measure: “Completed appointments this week” counts appointment records with a scheduled start inside the agreed local week and a current status of completed. It excludes cancellations, no-shows and appointments still waiting to happen. If the sample contains 12 records—eight completed, two cancelled, one no-show and one upcoming—the result is eight. This is an appointment count, not a count of distinct customers. If one customer attended twice, a unique-customer measure would answer a different question. Have the business owner approve that distinction.
- Source: the agreed appointment dataset.
- Time rule: use the appointment’s scheduled start in the business time zone.
- Week boundary: record the chosen start day and an explicit start-inclusive, end-exclusive interval.
- Correction rule: state whether later status changes update the historical total.
Check the path from source to report
Trace how those appointment records reach the reporting dataset. Can an import repeat a record? Can a status update arrive after the daily refresh? Which fields can be missing? Agree checks that match the definition: a stable appointment identifier, a valid scheduled time and a recognised status are relevant to this example. When a check fails, decide whether the report should flag the issue, exclude affected records with an explanation or wait for correction. Assign an owner instead of silently treating an incomplete feed as complete.
- Check sample totals against the source records.
- Distinguish a genuine zero from unavailable data.
- Record expected refresh timing and investigate late updates.
Make the number explainable in the interface
Label the measure precisely, show its reporting period and display the latest successful update. Provide a short definition near the total and a way to inspect relevant records where permissions allow. If filters change the population, make the active filters visible. Test the dashboard by asking an intended user to explain why the example returns eight and whether a late completion would change it. If the answer requires an analyst to reconstruct hidden rules, the definition or presentation still needs work.