--- name: dashboards description: "Dashboard design principles" --- Design dashboards that communicate the most important information at a glance, grounded in visual perception science rather than decoration. "A dashboard is a visual display of the most important information needed to achieve one or more objectives; consolidated and arranged on a single screen so the information can be monitored at a glance." — Stephen Few A dashboard must fit on a **single screen** without scrolling. If it requires scrolling, it's not a dashboard—it's a report. The constraint forces prioritization. Information must be perceivable **at a glance**. This means: - No hunting for data - No mental calculations required - Immediate pattern recognition - Exceptions visible without interaction Reduce to the minimum necessary: 1. **Reduce non-data pixels**: Eliminate unnecessary ones, de-emphasize others 2. **Enhance data pixels**: Eliminate unneeded ones, highlight the most important Every pixel should earn its place. If removing something loses no information, remove it. Color is a **scarce resource**. Use it only for meaning: - **Default**: Grayscale for most data and UI elements - **Reserve bright colors**: Only for alerts, exceptions, highlights - **Consistency**: Same color = same meaning across all dashboards - **When everything is bright, nothing stands out** Match precision to need: - Executives need trends and exceptions, not decimal places - Analysts need exact values for investigation - Operations needs real-time status, not historical depth Don't display more precision than the viewer can use. Three primary types, each with different design requirements: | Type | Purpose | Update | Interaction | Focus | |------|---------|--------|-------------|-------| | **Strategic** | Long-term oversight | Daily/Weekly | Minimal | Trends, KPIs | | **Analytical** | Deep investigation | As needed | High | Drill-down, comparison | | **Operational** | Real-time monitoring | Seconds/Minutes | Low | Alerts, current state | Most dashboards serve ONE type. Mixing types creates confusion. Choose the right display for the data: | Data Type | Best Display | Avoid | |-----------|--------------|-------| | Single KPI vs target | **Bullet graph** | Gauge, pie chart | | Trend over time | **Sparkline** | Full chart with axes | | Distribution | **Strip plot, histogram** | Pie chart | | Part-to-whole | **Stacked bar** | 3D pie, donut | | Comparison | **Bar chart** | Radar chart | | Geographic | **Choropleth** | 3D globe | | Correlation | **Scatter plot** | Bubble chart | | Current status | **Indicator + value** | Animation | **Rule**: The simplest display that communicates the data wins. What are you designing? 1. **New dashboard** — Need to determine type, metrics, layout 2. **Reviewing existing dashboard** — Audit against Few's principles 3. **Specific component** — Bullet graph, sparklines, KPI cards, alerts 4. **Display media selection** — What chart type for this data? 5. **Color/layout guidance** — Visual design decisions | Topic | Reference | |-------|-----------| | Strategic/Analytical/Operational dashboards | `references/dashboard-types.md` | | The 13 common mistakes to avoid | `references/common-mistakes.md` | | Bullet graph design and implementation | `references/bullet-graphs.md` | | Sparklines for trends | `references/sparklines.md` | | Color rules and palettes | `references/color-guidelines.md` | | Visual perception science | `references/visual-perception.md` | | Layout and organization | `references/layout-organization.md` | | Choosing display media | `references/display-media.md` | Run through this for any dashboard: **Structure** - [ ] Fits on single screen without scrolling? - [ ] Serves ONE dashboard type (strategic/analytical/operational)? - [ ] Most important information in top-left quadrant? **Display Media** - [ ] Using bullet graphs instead of gauges? - [ ] Sparklines for trends instead of full charts? - [ ] No pie charts, 3D effects, or decorative elements? **Visual Design** - [ ] Mostly grayscale with color reserved for meaning? - [ ] Adequate white space between groups? - [ ] Consistent alignment and sizing? **Data** - [ ] Context provided (vs. target, vs. prior period)? - [ ] Appropriate precision for audience? - [ ] Exceptions highlighted, not hidden? **Perception** - [ ] Can grasp key message in <5 seconds? - [ ] Related items grouped by proximity? - [ ] No chartjunk, gradients, or decorative images? Few's most critical warnings: **Gauges and Meters** — Waste space, encode poorly, look "impressive" but communicate little. Replace with bullet graphs. **Pie Charts** — Humans judge angles poorly. Use bar charts for comparison, bullet graphs for target tracking. **Traffic Lights Alone** — Red/yellow/green without values loses critical context. Always show the number alongside status. **Bright Color Everywhere** — When everything screams for attention, nothing gets it. Default to gray, highlight exceptions. **Scrolling Dashboards** — If it scrolls, it's not a dashboard. Prioritize ruthlessly. **Decoration** — Gradients, 3D effects, background images, corporate logos in data area. Every non-data pixel costs attention. This skill works with: - **d3-expert** — Implementation patterns for bullet graphs, sparklines - **stakeholder-adaptation** — Adjust dashboard for executive vs. analyst audience - **charts-visualization-review** — Complementary principles (data-ink ratio, lie factor) - **behavioral-health-viz-patterns** — Domain-specific dashboard patterns

ralph-loop

Scan your own AI components