Overview
A dashboard isn't a place to dump every metric you have. It's an information radiator that answers a specific set of questions for a specific audience. If the viewer has to think about what they're looking at, the dashboard has failed.
Before You Start
Ask the user:
- Who's the audience? — Execs, PMs, engineers, ops, customers. Each needs different views.
- What questions should it answer? — 3-5 specific questions, not "show everything."
- What decisions does it support? — What will someone DO after looking at it?
- Data available — What metrics exist and what needs to be built?
- Refresh cadence — Real-time, hourly, daily, weekly.
Design Process
Step 1: Define the Information Hierarchy
Level 1 (Glance — 3 seconds): Am I on track? Is anything on fire? → 3-5 headline metrics with trend indicators (up/down/flat)
Level 2 (Scan — 30 seconds): What's driving the numbers? → Key breakdowns, top-line charts, segment comparisons
Level 3 (Analyze — 5 minutes): Where should I dig deeper? → Detailed tables, drill-down views, time series
Step 2: Layout Design
┌─────────────────────────────────────────────────┐
│ LEVEL 1: Headline Metrics (3-5 KPIs) │
│ [Metric] [Trend] [Metric] [Trend] [Metric] │
├──────────────────────┬──────────────────────────┤
│ LEVEL 2: Primary │ LEVEL 2: Secondary │
│ Chart/Breakdown │ Chart/Breakdown │
│ │ │
├──────────────────────┴──────────────────────────┤
│ LEVEL 3: Detailed Table / Drill-down │
│ │
└─────────────────────────────────────────────────┘
Step 3: Metric Selection per Audience
| Audience | Key Questions | Metrics | Cadence | |----------|-------------|---------|---------| | Exec | Are we growing? Are we profitable? | MRR, growth rate, burn | Weekly | | PM | Is the product healthy? What needs attention? | Activation, retention, feature adoption | Daily | | Engineering | Is the system healthy? | Uptime, latency, error rate | Real-time | | Support | What are users struggling with? | Ticket volume, resolution time, top issues | Daily | | Customers | Am I getting value? | Usage, ROI, team activity | On-demand |
Step 4: Visualization Principles
- Time series for trends (line chart)
- Bar charts for comparison across categories
- Single numbers for current state with trend arrows
- Tables for detailed drill-down data
- Heatmaps for patterns across two dimensions
- Never use pie charts for more than 3 segments
Color coding:
- Green = on track / improving
- Yellow = watch / declining
- Red = needs action / below threshold
Step 5: Alert and Threshold Design
Every headline metric needs:
- Green threshold: [metric] > [value] — all good
- Yellow threshold: [value] > [metric] > [value] — investigate
- Red threshold: [metric] < [value] — action required
Output
# Dashboard Design — [Name] for [Audience]
## Purpose
[What questions this dashboard answers]
## Metrics Inventory
| Metric | Definition | Source | Refresh | Threshold (R/Y/G) |
|--------|-----------|--------|---------|-------------------|
| [metric] | [how calculated] | [data source] | [cadence] | [thresholds] |
## Layout Specification
[Wireframe with metric placement and chart types]
## Interaction Design
- [Filters available]
- [Drill-down paths]
- [Date range controls]
- [Export capabilities]
## Technical Requirements
- Data sources: [list]
- Refresh frequency: [cadence]
- Access control: [who sees what]
- Tool recommendation: [Looker/Tableau/Metabase/custom]
Save as DASHBOARD-DESIGN-[name].md.