Dashboard design

Dashboard Design

Property	Value
Domain	Data Analytics
Category	Dashboard Architecture & Layout
Components	SKILL.md + dashboard-design.instructions.md + dashboard.prompt.md
Depends	data-visualization (chart content), data-analysis (KPIs and segments)

Staleness Watch: See EXTERNAL-API-REGISTRY.md for source URLs and recheck cadence

Overview

Design dashboards that tell stories through layout. A dashboard is not a collection of charts -- it's a guided reading experience where position, size, and interaction design control what the viewer understands and in what order.

The cardinal rule: reading order = story order. KPIs answer "what?", charts answer "why?", tables answer "how much?", and drill-downs answer "show me more."

Module 1: Layout Patterns

Pattern	Best For	Reading Flow	Use When
Inverted Pyramid	Executive dashboards	KPIs → hero chart → supporting → detail	Audience scans from top, drills as needed
Z-Pattern	Marketing / consumer	Top-left → top-right → bottom-left → end	Two primary metrics side-by-side
F-Pattern	Data-heavy analytics	Left column scanned top-down, right detail	Dense detail dashboards for analysts
Hub-and-Spoke	Exploration dashboards	Central question, surrounding perspectives	When multiple equally important views exist
Narrative Scroll	Data stories / reports	Top-to-bottom linear reading	Self-contained HTML stories with guided arc

Reference Pattern: Inverted Pyramid (from population.html)

┌────────────────────────────────────────────────────┐
│ HEADER: Title + subtitle + badge                   │
├────────────────────────────────────────────────────┤
│ KPI ROW: 3-6 cards (auto-fit grid)                 │
├────────────────────────────────────────────────────┤
│ CONTROLS: Search + filter buttons                  │
├────────────────────────────────────────────────────┤
│ HERO CHART: Full-width, clickable                  │
├───────────────────────┬────────────────────────────┤
│ SUPPORTING CHART 1    │ SUPPORTING CHART 2         │
├───────────────────────┴────────────────────────────┤
│ DATA TABLE: Sortable, ranked, inline indicators    │
├────────────────────────────────────────────────────┤
│ FOOTER: Data source + attribution + last updated   │
└────────────────────────────────────────────────────┘

         ┌── DRILL-DOWN MODAL ──┐
         │ Detail stats grid     │
         │ Horizontal bar        │
         │ Detail table          │
         └───────────────────────┘

Module 2: KPI Card Design

Each KPI card follows the metric + delta + context pattern:

Element	Purpose	Example
Primary value	The headline number	"334M"
Label	What it measures	"Total Population"
Delta	Change indicator (arrow + %)	"↑ 2.1% YoY"
Sparkline	Trend miniature (optional)	Tiny 30-day line
Subtitle	Time range or source	"2024 Census estimate"
Icon/Emoji	Visual anchor	"👥"
Accent color	Category signal	Brand purple

KPI Selection Rules

Rule	Why
3-6 KPIs maximum	More overwhelms; fewer under-informs
Each answers a different "what?"	Redundant KPIs waste prime real estate
Include at least one delta	Static numbers lack context
Lead with the most important	Top-left KPI gets the most attention
Color thresholds if applicable	Green/amber/red for status-driven metrics

Module 3: Filter Architecture

Filter Type	Placement	Behavior	Example
Global filter	Top, below KPIs	Affects all charts and table	Region buttons, date range
Chart filter	Inline on chart	Affects only parent chart	Sort toggle, limit slider
Cross-filter	Click on chart	Highlights related data elsewhere	Click bar → table highlights
Drill-through	Click on item	Opens detail modal or new view	Click state → city breakdown
Search	Top controls	Text filter across primary column	Search box for state names

Filter Design Rules

1. Global filters visible at all times (sticky if page scrolls)
2. Active filter state clearly indicated (colored buttons, badges)
3. "All" or "Reset" always available as first option
4. Filter changes animate smoothly (transition: 300ms)
5. Filter state reflected in chart titles or subtitles

Module 4: Narrative Flow

The dashboard tells a story through panel arrangement:

Layer	Component	Answers	Time to Absorb
1	KPI cards	"What's the story?"	2 seconds
2	Hero chart	"Show me the shape"	10 seconds
3	Supporting charts	"What else?"	15 seconds
4	Data table	"Give me the details"	As needed
5	Drill-down modal	"Tell me more about X"	On demand

Progressive Disclosure

• Above the fold: KPIs + hero chart (must convey the main story without scrolling)
• Below the fold: Supporting views + table (deeper evidence)
• On demand: Drill-down modals, tooltips, expandable rows

Module 5: Self-Contained HTML Pattern

Every dashboard outputs as a single HTML file:

Requirement	Implementation
Zero build step	Open in browser, no npm/webpack/build
Embedded data	JavaScript arrays/objects, not external API calls
Single CDN dependency	Chart.js 4.x via jsdelivr
Responsive	CSS grid with auto-fit, minmax, media queries
Themeable	CSS custom properties for colors, spacing, fonts
Accessible	Semantic HTML, ARIA labels, alt text on charts
Print-friendly	@media print styles for clean output

CSS Custom Properties Template

:root {
  --bg-primary: #0f0f23;
  --bg-card: #1a1a2e;
  --text-primary: #e8e8f0;
  --text-secondary: #888;
  --accent-1: #6366f1;
  --accent-2: #3b82f6;
  --accent-3: #10b981;
  --accent-4: #f59e0b;
  --accent-5: #ec4899;
  --border-radius: 12px;
  --shadow: 0 2px 8px rgba(0,0,0,0.3);
  --transition: 0.3s ease;
}

Module 6: Platform Targets

Platform	Output Format	Key Differences
HTML (primary)	Self-contained	Chart.js + Canvas, embedded data
Power BI	Report spec	DAX measures, M queries, visual config
Streamlit	Python app	Plotly charts, st.columns, st.metric
React	Component tree	Recharts/Nivo, state management, props

Default to HTML unless the user specifies otherwise.

Module 7: Chart Grid Pattern (from visuals.html)

For dashboards that showcase multiple chart types or let users explore:

Component	Purpose	Implementation
Category filter	Show charts by category	Button row with colored states
Card grid	Visual index of available charts	CSS grid, auto-fit, minmax(280px)
Mini preview	Quick visual (130px height canvas)	Chart.js render at small scale
Metadata badges	"Best for", "Data type"	Tags below chart preview
Modal detail	Full chart + metadata + examples	Fixed overlay, arrow key nav

Module 8: Interaction Patterns

Pattern	Trigger	Effect
Hover tooltip	Mouse over bar	Show exact value + label
Click drill-down	Click chart item	Open modal with detail view
Filter select	Click button	Filter all charts to selection
Sort toggle	Click column head	Re-sort table ascending/descending
Search highlight	Type in search	Fade non-matching, highlight match
Responsive collapse	Screen < 768px	Stack columns, hide secondary charts

Module 9: 5-Visual Rule (VT BIT 5424)

Executive dashboards follow the 5-Visual Rule:

1. 1 KPI row (3-6 metric cards)
2. 1 hero chart (the main story)
3. 2 supporting charts (different lenses on same data)
4. 1 data table (detail on demand)

Maximum 5 visual elements above the fold. More than 5 creates cognitive overload for executive audiences. Analysts get more through drill-down, not more panels.

Anti-Patterns

Anti-Pattern	Problem	Fix
Chart soup	Too many charts with no narrative connection	Apply 5-Visual Rule, add story arc
Dashboard as database	Every field shown "just in case"	Show only what answers the KPI questions
Inconsistent colors	Same category = different color across charts	Use semantic color map (region → color)
No drill-down escape	Modal/detail has no clear close/back	Always show close button + ESC handler
Tiny charts	Cramming 10 charts into one screen	Fewer, larger charts with drill paths
Static title	"Revenue Dashboard"	"Revenue grew 34% -- but growth is slowing"