DataIntermediate
Spec a dashboard people actually use
Defines decisions, metrics and alerts before choosing charts, to avoid a decorative dashboard.
When to use it
When someone asks for "a dashboard" and nobody knows what decision it will support.
Copy Prompt
Copy the prompt and replace the placeholders in braces.Act as a data product designer. A dashboard without an associated decision is decoration.
Who uses it: {{user}}
Decisions they make: {{decisions}}
Available data: {{data}}
Usage frequency: {{frequency}}
Deliver:
1. THREE QUESTIONS the dashboard must answer in under 10 seconds.
2. METRICS: 5 main ones with an exact definition, formula and source. Mark the north-star metric.
3. SCREEN STRUCTURE: top block (status), middle (trend), bottom (detail). Which chart goes in each one and why that type.
4. ALERTS: 3 rules with a threshold and who they notify, including what to do when receiving one.
5. WHAT SHOULD NOT BE THERE and why (vanity metrics, everything on one screen, endless filters).
6. SUCCESS CRITERION at 30 days: how we will know it is being used.
Rules: 7 visualisations maximum. Each chart must answer a question from point 1. No pie charts with 9 categories.Variables to fill in
| Variable | What it means | Example |
|---|---|---|
user | Who uses it | Head of Sales |
decisions | What they decide | Where to focus sales each week |
data | What is available | CRM, invoicing, marketing |
frequency | How often they open it | Monday morning |
Tips
- Ask for the exact definition of every metric: that is where meeting arguments surface.
- Fewer charts and more clarity: 5 well chosen beat 20.
- If there is no action associated with an alert, the alert is unnecessary.
Editorial template. Test it with real data before using it in production.