data-visualization

name	data-visualization
description	Data visualization best practices covering chart selection, color theory, layout design, and accessibility. Use when choosing the right chart type for data, designing dashboards, creating infographics, improving existing visualizations, or building data-driven presentations. Triggers on tasks involving chart type selection, color palette design, dashboard layout, data storytelling, matplotlib/plotly/D3.js visualization code, or visual data communication.
metadata	{"version":"1.0.0"}

Data Visualization Best Practices

A comprehensive guide for creating effective, accurate, and beautiful data visualizations — covering chart selection, color theory, layout design, accessibility, and common pitfalls.

When to Apply

• Choosing the right chart type for your data
• Designing dashboards or data reports
• Creating visualizations with matplotlib, plotly, D3.js, ECharts, etc.
• Improving existing visualizations for clarity or aesthetics
• Building data-driven presentations
• Designing accessible visualizations for diverse audiences

---

1. Chart Selection Guide

1.1 Choosing the Right Chart

What You Want to Show	Best Chart Type	Alternatives
Comparison (categories)	Bar chart (horizontal)	Grouped bar, Bullet chart
Comparison (over time)	Line chart	Area chart, Slope chart
Composition (static)	Stacked bar chart	Pie chart (≤5 items), Treemap
Composition (over time)	Stacked area chart	100% stacked bar
Distribution	Histogram	Box plot, Violin plot
Relationship (2 variables)	Scatter plot	Bubble chart, Heatmap
Relationship (3+ variables)	Scatter matrix	Parallel coordinates
Ranking	Horizontal bar chart	Lollipop chart, Ordered table
Deviation	Diverging bar chart	Waterfall chart
Flow / Process	Sankey diagram	Alluvial diagram
Geography	Choropleth map	Bubble map, Cartogram
Single metric	Big number + trend	Gauge, Bullet chart
Part-to-whole	Donut chart	Waffle chart, Treemap
Correlation	Scatter plot with trend line	Heatmap matrix

1.2 When to Use (and Avoid) Common Charts

Bar Chart — The workhorse

• Use for: Comparing discrete categories, showing rankings
• Avoid: Too many categories (>20), continuous data
• Tip: Always start Y-axis at 0

Pie / Donut Chart — Use sparingly

• Use for: Showing part-to-whole with ≤5 categories
• Avoid: >5 slices, similar-sized slices, precise comparison
• Tip: Donut > Pie (center space for label), always include percentages

Line Chart — Time series default

• Use for: Showing trends over continuous time periods
• Avoid: Too many lines (>7), categorical X-axis
• Tip: Highlight the key line, gray out others

Scatter Plot — Relationship explorer

• Use for: Showing correlation between two continuous variables
• Avoid: Very large datasets (>10k points without aggregation)
• Tip: Add trend line, color by category, size by third variable

Heatmap — Density and patterns

• Use for: Showing concentration across two dimensions
• Avoid: Too many categories, non-sequential color scales
• Tip: Use perceptually uniform color maps (viridis, plasma)

---

2. Color Theory for Data Visualization

2.1 Color Palette Types

Palette Type	Use Case	Examples
Sequential	Ordered data (low to high)	Viridis, Blues, Oranges
Diverging	Data with meaningful midpoint	RdBu, PiYG, BrBG
Qualitative	Categorical data (no order)	Set2, Paired, Tab10
Highlight	Draw attention to one category	Gray base + one accent color

2.2 Color Rules

1. Use color meaningfully — Every color should encode information
2. Limit palette — Maximum 7-10 distinct colors for categories
3. Respect color blindness — ~8% of men are colorblind
4. Use sequential palettes for ordered data — Never use rainbow
5. One accent color — Gray for context, one color for focus
6. Consistent mapping — Same category = same color across all charts

2.3 Color-Blind Safe Palettes

Safe sequential:   viridis, plasma, inferno, magma, cividis
Safe diverging:    RdBu (red-blue), PRGn (purple-green), BrBG (brown-green)
Safe qualitative:  IBM Design, Okabe-Ito, Wong palette

# Wong palette (8 colors, colorblind-safe)
#   Black:   #000000
#   Orange:  #E69F00
#   Sky Blue:#56B4E9
#   Green:   #009E73
#   Yellow:  #F0E442
#   Blue:    #0072B2
#   Vermillion:#D55E00
#   Purple:  #CC79A7

2.4 The Highlight Technique

# Good: Highlight one category, gray out others
colors = ['#E69F00' if cat == 'Target' else '#CCCCCC' for cat in categories]

# Bad: Using many saturated colors for equal emphasis
colors = ['#E69F00', '#56B4E9', '#009E73', '#F0E442', '#0072B2', '#D55E00']

---

3. Typography & Labeling

3.1 Label Hierarchy

Level	Element	Font Size	Weight
1	Chart title	16-20px	Bold
2	Axis labels	12-14px	Medium
3	Tick labels	10-12px	Regular
4	Annotations	10-11px	Regular / Italic
5	Source / notes	8-9px	Light / Italic

3.2 Labeling Best Practices

• Always label axes — Include units (e.g., "Revenue ($M)")
• Direct labeling > Legends — Place labels near data points when possible
• Rotate labels only as last resort — Prefer horizontal text
• Avoid chart junk — Remove unnecessary gridlines, borders, backgrounds
• Use annotations to highlight key insights, not just data points

3.3 Number Formatting

Good:                    Bad:
$1.2M                    $1200000
45.2%                    0.452
1,234                    1234
2024-01                  Jan, 2024 (inconsistent)
3.2x                     320%

---

4. Dashboard Design

4.1 Dashboard Layout Principles

┌─────────────────────────────────────────────────┐
│  Title + Filters + Date Range                    │
├──────────┬──────────┬──────────┬────────────────┤
│  KPI #1  │  KPI #2  │  KPI #3  │  KPI #4        │
├──────────┴──────────┴──────────┴────────────────┤
│                                                   │
│           Primary Chart (Largest)                 │
│                                                   │
├──────────────────────┬──────────────────────────┤
│   Secondary Chart    │   Secondary Chart          │
├──────────────────────┴──────────────────────────┤
│  Data Table (Detail)                             │
└─────────────────────────────────────────────────┘

4.2 Dashboard Rules

1. Top-left to bottom-right — Most important info first (F-pattern)
2. KPIs at top — Big numbers with sparklines or trend indicators
3. One primary chart — Takes up the most space, answers the main question
4. Consistent filters — Apply to all charts on the dashboard
5. Progressive disclosure — Summary → Detail (drill-down)
6. Max 5-7 charts per dashboard — More = cognitive overload
7. Responsive design — Adapt to different screen sizes

4.3 KPI Card Design

┌──────────────────────┐
│  Total Revenue        │  ← Label (12px, gray)
│                       │
│  $2,458,930           │  ← Value (32px, bold, dark)
│  ▲ 12.3% vs last mo. │  ← Change (12px, green/red)
│  ━━━━━━━━━━━━━━━━━   │  ← Sparkline (optional)
└──────────────────────┘

---

5. Accessibility

5.1 Accessibility Checklist

• Color is not the only encoding — Use patterns, shapes, or labels alongside color
• Color-blind safe palette — Test with CVD simulation tools
• Sufficient contrast — Minimum 4.5:1 ratio for text (WCAG AA)
• Text alternatives — Provide data tables alongside charts
• Keyboard navigable — Interactive charts support keyboard input
• Screen reader compatible — ARIA labels for chart elements
• Avoid red/green only — Most common color blindness

5.2 Accessible Pattern Examples

Bad: Red = bad, Green = good (only color encoding)

🟥 Failed  🟩 Passed  🟩 Passed  🟥 Failed

Good: Color + icon + text

✗ Failed  ✓ Passed  ✓ Passed  ✗ Failed
(red)     (green)   (green)   (red)

---

6. Common Visualization Mistakes

6.1 The Hall of Shame

Mistake	Problem	Fix
Truncated Y-axis	Exaggerates differences	Start at 0 (for bar charts)
Dual Y-axis	Misleading correlation	Use two separate charts or normalize
3D charts	Distorts perception	Always use 2D
Rainbow color scale	Not perceptually uniform	Use viridis or sequential palette
Too many categories	Unreadable	Group small categories as "Other"
Missing units	Ambiguous values	Always include units in axis labels
Pie chart with many slices	Hard to compare	Use bar chart instead
No title	Unclear purpose	Every chart needs a descriptive title
Decorative chart junk	Distracts from data	Remove unnecessary elements
Inconsistent scales	Misleading comparison	Use same scale across comparable charts

6.2 The Truncated Y-Axis Problem

Bad (Y starts at 90):          Good (Y starts at 0):
  │                            │
  │  ██                        │  ██
  │  ██ ██                     │  ██ ██
  │  ██ ██ ██                  │  ██ ██ ██
  └────────────                └────────────
  90  95  100                   0   50  100

  Looks like 3x difference!    Shows true ~5% difference

6.3 When Dual Y-Axis is OK (Rarely)

Only acceptable when:

• Both axes measure the same unit
• There's a clear, documented reason
• Alternative (two charts) is significantly worse
• The relationship is genuinely meaningful, not spurious

---

7. Visualization Code Patterns

7.1 Matplotlib Best Practices (Python)

import matplotlib.pyplot as plt
import matplotlib.ticker as mticker

plt.rcParams.update({
    'font.family': 'sans-serif',
    'font.size': 11,
    'axes.spines.top': False,
    'axes.spines.right': False,
    'figure.dpi': 150,
})

fig, ax = plt.subplots(figsize=(10, 6))
bars = ax.barh(categories, values, color='#0072B2', height=0.6)
ax.set_xlabel('Revenue ($M)', fontsize=12)
ax.set_title('Quarterly Revenue by Product', fontsize=16, fontweight='bold', pad=20)
ax.xaxis.set_major_formatter(mticker.FuncFormatter(lambda x, _: f'${x/1e6:.1f}M'))

for bar, val in zip(bars, values):
    ax.text(bar.get_width() + 0.5, bar.get_y() + bar.get_height()/2,
            f'${val/1e6:.1f}M', va='center', fontsize=10)

ax.grid(axis='x', alpha=0.3, linestyle='--')
ax.set_xlim(0, max(values) * 1.15)
plt.tight_layout()
plt.savefig('chart.png', bbox_inches='tight', dpi=150)

7.2 Plotly Best Practices (Python)

import plotly.express as px

fig = px.bar(
    df, x='category', y='value', color='segment',
    color_discrete_sequence=['#0072B2', '#E69F00', '#009E73'],
    title='Revenue by Category and Segment',
    labels={'value': 'Revenue ($M)', 'category': 'Product Category'},
)

fig.update_layout(
    font=dict(family='sans-serif', size=12),
    title_font=dict(size=18),
    plot_bgcolor='white',
    legend=dict(orientation='h', yanchor='bottom', y=1.02),
    margin=dict(t=80, b=60, l=60, r=40),
)

fig.update_xaxes(showgrid=False)
fig.update_yaxes(gridcolor='#E5E5E5', gridwidth=0.5)
fig.show()

---

8. Data Storytelling Framework

8.1 Structure

1. Context — What is the question? Why does it matter?
2. Data — What does the data show? (Visual evidence)
3. Insight — What is the key finding? (Annotation)
4. Action — What should we do about it? (Recommendation)

8.2 Annotation Best Practices

• Use callout boxes for key insights, not just data labels
• Highlight the "so what" — why this data point matters
• Use before/after comparisons for impact
• Add trend lines and reference lines for context
• Include a narrative title that states the finding

Bad title: "Sales by Month" Good title: "Sales peaked in March (+35%) driven by new product launch"

---

9. Quick Reference Checklist

• Chart type matches the data relationship being shown
• Y-axis starts at 0 (for bar charts)
• Color palette is color-blind safe
• All axes are labeled with units
• Chart has a descriptive title
• No 3D effects or unnecessary decorations
• Data source is cited
• Interactive elements have keyboard support
• Numbers are formatted consistently
• Legend is necessary (or replaced with direct labels)
• Font sizes follow the hierarchy (title > labels > ticks)
• Color encodes information, not decoration