Definitions first, dashboards second
When two dashboards disagree it is almost never a bug. It is two teams using two definitions — one counting deals created, the other deals with a value and a close date; one counting MQLs by form fill, the other by score.
Write the definitions down in a one-page glossary with an owner per metric, and make it the source of truth. This one document resolves more reporting disputes than any rebuild.
| Metric | Define precisely | Common ambiguity |
|---|---|---|
| Lead | Lifecycle stage + qualifying property | Every form fill counted as a lead |
| Opportunity | Deal at stage X with amount and close date | Deals created at stage 1 with no value |
| Pipeline value | Open deals, weighted or unweighted — pick one | Two dashboards, two methods |
| Win rate | Won ÷ (won + lost) in a period, by close date | Divided by all deals including open |
| Source | First-touch or last-touch — pick one | Mixed across marketing and sales reports |
The four dashboards
- 01Executive: revenue closed, pipeline created, win rate, average sales cycle. Six tiles maximum, monthly cadence, owned by the revenue lead.
- 02Pipeline: deals by stage, stage-to-stage conversion, ageing deals, next-activity coverage. Weekly, owned by sales leadership.
- 03Marketing performance: sessions to conversion by channel, pipeline sourced, cost per opportunity where you have spend data. Monthly, owned by marketing.
- 04Data quality: records missing owner, deals with no close date, contacts with no lifecycle stage, duplicate rate. Weekly, owned by ops — this is the dashboard that keeps the other three honest.
Ageing and next-activity beat vanity metrics
The two reports that change behaviour fastest are deal ageing by stage and open deals with no scheduled next activity. Both surface work rather than describing history, and both are uncomfortable in the right way.
Pair them with stage-to-stage conversion rather than a single funnel percentage. A funnel number tells you something is wrong; conversion by stage tells you where.
Delete before you build
- Export the report list with last-viewed dates. Anything unopened in 90 days is a candidate.
- Archive rather than delete for one quarter, so the objection 'I need that' can be tested.
- Ban personal copies of shared reports — they are how definitions drift.
- Cap each dashboard at eight tiles. Scrolling is where trust dies.
- Re-run the exercise every six months; report sprawl regrows.
Attribution and what to expect from it
Multi-touch attribution on Professional and Enterprise is useful for directional channel decisions, not for settling budget arguments to the decimal. Treat it as evidence, not verdict, and always publish which model the report uses on the report itself.
If your data quality dashboard is red, attribution output is noise. Fix ownership and lifecycle-stage hygiene first.
Common mistakes
- Building dashboards before agreeing metric definitions.
- Letting each team keep private copies of the same report.
- Twenty-tile dashboards nobody scrolls through.
- Reporting on data quality nowhere, so bad data stays invisible.
- Treating attribution output as a settled fact rather than a model.