Start with a decision inventory
List the recurring choices the team needs to make: whether to change targeting, repair a handoff, continue a program or investigate a conversion decline. If a metric does not inform a choice, explain why it needs to be collected and displayed.
Distinguish descriptive questions from causal ones. “Which sources appeared in won opportunities?” is not the same as “Which activity caused additional wins?” Attribution and experiments provide different evidence. A single dashboard should not blur that distinction.
Specify each metric
| Field | Example of the required precision |
|---|---|
| Decision | Whether a qualification rule needs revision |
| Definition | Accepted enquiries divided by eligible routed enquiries |
| Grain | One enquiry, with a defined duplicate rule |
| Cohort | Enquiries received in a stated period |
| Window | Time allowed for a disposition to occur |
| Source | Named system fields and status rules |
| Owner | Person who can explain and correct the number |
| Action | What evidence would trigger investigation |
The example is illustrative. Your organization should define the terms according to its process and preserve changes to those definitions over time.
Reconcile before visualizing
Trace a sample of records and compare system totals. Inspect duplicates, missing identifiers, late updates and exclusions. A clear chart can make an inconsistent calculation more persuasive, so validate the underlying data before investing in presentation.
Respect time and maturity
Recent cohorts may not have had enough time to progress. Compare equivalent observation windows or label immature results. Keep period totals distinct from cohort conversion: opportunities closed this month may have originated much earlier.
Build a balanced review
Use outcome measures, leading observations and quality indicators. For example, opportunity progression can be reviewed alongside response capacity and missing dispositions. A change in performance is easier to diagnose when the operational context is visible.
Document uncertainty and action
Show where data is incomplete or modeled. A missing value is different from zero, and an attributed value is different from incremental contribution. End the review with a decision, owner and next evidence request. The measurement system is working when it helps the organization act more intelligently, not merely when every tile is filled with a number.