Business
KPI Dashboard Spec AI Prompt Template
Write a KPI dashboard spec with this AI prompt. Layout, metrics, and requirements developers can build from. Includes variables, examples, and pro tips.
The prompt
ROLE: You are a product manager who writes dashboard specs engineers love to build. CONTEXT: Dashboard for: [AUDIENCE] (e.g., exec team). Purpose: [PURPOSE]. Key metrics: [METRICS]. Data sources: [SOURCES]. Refresh needs: [REFRESH]. Current pain: [PAIN]. TASK: Write a complete dashboard requirements spec. FORMAT: 1. Dashboard Purpose (one paragraph: who uses it, what decision it supports). 2. Layout: describe 3 sections (top: 4-5 headline KPI cards with trend; middle: 3-4 trend charts; bottom: breakdown tables), each widget specified (metric, visualization type, time range, comparison). 3. Interactivity (filters, drill-downs, date ranges). 4. Data Requirements (source per metric, refresh cadence, owner). 5. Alerts (which metrics trigger alerts, thresholds, recipients). 6. Non-requirements (what this dashboard will NOT do). End with acceptance criteria (5 testable statements). CONSTRAINTS: One dashboard, one decision; dashboards that try to serve everyone serve no one. Every widget must map to the stated purpose. Specify empty states and loading behavior briefly.
Variables to fill in
Replace each [variable] in the prompt with your own details:
[AUDIENCE] [PURPOSE] [METRICS] [SOURCES] [REFRESH] [PAIN]
Example input
AUDIENCE=exec team, PURPOSE=weekly revenue health check; METRICS=revenue, new customers, churn, CAC; SOURCES=Stripe, HubSpot, GA4; REFRESH=daily 6am; PAIN=numbers differ between tools
How to use this prompt
Agree the purpose sentence with stakeholders first, then spec widgets; purpose arguments later are expensive.
Tips for the best results
- Define each metric once in a shared doc; most dashboard arguments are actually definition arguments.
- Start with 8 widgets max; add more later, but cluttered v1 never gets cleaned.
- Put the data owner on every metric; ownerless metrics rot when sources break.
Frequently asked questions
How many dashboards does a company need?
One per key decision area (exec, sales, marketing, product, finance). More than 8-10 usually means unclear ownership.
Real-time or daily refresh?
Daily for decisions, real-time only for operations (support queues, incident response). Real-time exec dashboards create anxiety, not insight.