Diagnose before you rebuild
When two reports disagree, resist the urge to rebuild both. Open them side by side and compare four things: the object being counted, the date property used for filtering, the filters and exclusions, and the date range boundaries.
In nearly every case the discrepancy is one of those four. Rebuilding without knowing which one produces a third number, and now three reports disagree.
The date property problem
Lifecycle stage date-stamp properties exist precisely for this and are consistently underused. If a report filters on create date when the business means stage entry, the number will be wrong in a way that looks plausible — the worst kind of wrong.
| Metric | Wrong property | Right property |
|---|---|---|
| Pipeline created | Deal create date | Date entered first sales stage |
| Win rate by month | Close date only | Cohort by created date, measured at close |
| MQLs | Contact create date | Date entered MQL lifecycle stage |
| Time to close | Create date to close date | First sales stage to closed-won |
The metric dictionary
Twelve to twenty metrics covers most mid-market businesses. Keep it in one shared document, link it from the dashboard description, and review it quarterly.
- Name — the exact phrase leadership uses in meetings.
- Definition — one sentence a new joiner could apply.
- Source — object, properties, date property.
- Exclusions — test deals, internal records, closed-lost-duplicate, specific pipelines.
- Owner — a person, not a team.
- Canonical report — a direct link to the one report that produces this number.
One canonical report per metric
Dashboards should be arrangements of canonical reports, not places where new logic gets written. When a board deck needs a variant, clone the canonical report, rename it with an explicit suffix and note the difference — do not silently create a competing definition.
Data quality is a reporting feature
Reports built on properties that are 60% populated are guesses with a chart. Before publishing a dashboard, check fill rates on every property it depends on and either fix the collection path or drop the breakdown. A dashboard with four trustworthy tiles beats one with twelve negotiable ones.
The monthly reconciliation habit
- 01On the first working day of the month, export the canonical numbers.
- 02Compare against finance's closed revenue and against last month's export.
- 03Investigate any variance over an agreed threshold before the leadership meeting, not during it.
- 04Log the cause. Repeat causes indicate a process problem, not a reporting one.
Common mistakes
- Building dashboards before agreeing definitions.
- Letting every team build their own version of pipeline created.
- Filtering on close date for anything that is not a closing metric.
- Twelve tiles where four would be trusted.
- No named owner, so nobody is accountable when a number drifts.