Data Visualization Expert

Data Visualization Expert

Overview

Create publication-quality, accessible, and reader-friendly data visualizations that communicate insights effectively. This skill provides comprehensive guidance on visualization design principles, chart selection, color theory, dashboard design, and best practices for professional data communication.

When to Use This Skill

Invoke this skill when working on tasks that involve:

• Chart Creation: Creating any type of chart or graph
• Chart Selection: Choosing the right visualization for your data
• Color Design: Selecting color palettes and ensuring accessibility
• Dashboard Design: Creating executive dashboards or operational monitors
• Visual Design: Improving readability and aesthetic appeal
• Accessibility: Ensuring visualizations work for colorblind viewers
• Presentation Design: Preparing visualizations for slides or reports
• Data Communication: Making insights clear and actionable

Typical User Requests:

• "Create a bar chart comparing sales by region"
• "What's the best way to visualize this time series data?"
• "Make this chart more readable and professional"
• "Design a dashboard showing our key metrics"
• "Choose colors that work for colorblind people"
• "How should I present this data in a presentation?"
• "Create visualizations that tell a story"

Prerequisites

Before using this skill, ensure:

• Python 3.8+ with the following packages installed:
• matplotlib>=3.5.0
• seaborn>=0.12.0
• pandas>=1.3.0
• numpy>=1.20.0
• Japanese Font Support (optional, for Japanese text):
• macOS: Built-in (Hiragino Sans)
• Windows: Built-in (Yu Gothic, Meiryo)
• Linux: Install fonts-noto-cjk or fonts-takao
• Input Data: CSV files with headers for chart creation scripts

Install dependencies:

pip install matplotlib seaborn pandas numpy

Core Visualization Workflow

Follow this systematic approach when creating visualizations:

1. Understand the Goal
   ↓
2. Analyze the Data
   ↓
3. Select Chart Type
   ↓
4. Choose Color Palette
   ↓
5. Apply Design Principles
   ↓
6. Add Context & Annotations
   ↓
7. Ensure Accessibility
   ↓
8. Review & Refine

Workflow 1: Choosing the Right Chart Type

Step 1: Identify Your Communication Goal

Ask yourself:

• What insight am I trying to convey?
• What question does this visualization answer?
• What action should viewers take?

Common Goals:

• Compare values across categories → Bar Chart, Lollipop Chart
• Show distribution → Histogram, Box Plot, Violin Plot
• Reveal relationship → Scatter Plot, Line Chart
• Display composition → Stacked Bar, Tree Map, Pie Chart (sparingly)
• Track change over time → Line Chart, Area Chart
• Show geographic patterns → Choropleth Map, Symbol Map
• Display hierarchical data → Tree Map, Sunburst, Tree Diagram

Step 2: Consider Your Data Characteristics

Data Type:

• Categorical: Bar chart, pie chart (2-5 categories only)
• Continuous: Line chart, scatter plot, histogram
• Time series: Line chart, area chart, sparkline
• Geographic: Map visualizations
• Hierarchical: Tree map, sunburst
• Multivariate: Scatter plot matrix, parallel coordinates, small multiples

Data Volume:

• Few data points (100): Density plots, heatmaps, sampling, or interactive charts

Step 3: Consult the Chart Selection Guide

Reference: references/chart_selection_guide.md

Quick Selection:

COMPARISON → Bar Chart (vertical/horizontal)
  ├─ Few categories → Vertical bar
  ├─ Many categories or long names → Horizontal bar
  ├─ Multiple groups → Grouped bar
  └─ Parts + Total → Stacked bar

DISTRIBUTION → Histogram + KDE
  ├─ Quick summary → Box Plot
  ├─ Full shape → Violin Plot
  └─ Compare groups → Side-by-side box plots

RELATIONSHIP → Scatter Plot
  ├─ Continuous time → Line Chart
  ├─ Correlation → Heatmap
  └─ 3+ variables → Bubble chart or color/size encoding

COMPOSITION → Stacked Bar Chart
  ├─ Simple (2-3 parts) → Pie chart (acceptable)
  ├─ Over time → Stacked area
  ├─ Hierarchy → Tree map
  └─ Sequential changes → Waterfall

TIME SERIES → Line Chart
  ├─ Volume/cumulative → Area chart
  ├─ Multiple metrics → Line chart (max 5 lines)
  └─ Compact trend → Sparkline

Step 4: Avoid Common Mistakes

❌ Don't Use:

• Pie charts with >5 segments (use bar chart instead)
• 3D charts (distorts perception)
• Dual Y-axes (can mislead)
• Too many colors (limit to 5-8 for categorical data)
• Inconsistent time intervals

✅ Do Use:

• Simple, clear chart types
• Appropriate scales (start Y-axis at zero for bar charts)
• Direct labels when possible (instead of legends)
• Consistent formatting

Example: Choosing for Sales Analysis

Scenario: Compare sales across 5 product categories

Decision Process:

1. Goal: Compare values across categories
2. Data: 5 categorical values (products) with one numeric value (sales)
3. Chart Type: Bar Chart (vertical or horizontal)
4. Enhancements: Highlight top performer, add target line
5. Result: Clear comparison with immediate insight

Implementation:

python scripts/create_visualization.py \
  --type bar \
  --input sales_data.csv \
  --x product_category \
  --y sales_amount \
  --title "Q4 Product Sales Comparison" \
  --sort \
  --output product_sales.png

Workflow 2: Applying Color Best Practices

Step 1: Understand Color Purpose

Color Should:

• Encode information (categories, magnitude)
• Create visual hierarchy
• Guide attention to insights
• Maintain consistency across visualizations
• Work for all viewers (including colorblind)

Color Should NOT:

• Be purely decorative
• Use red/green for critical information (colorblind issue)
• Overwhelm with too many hues
• Lack sufficient contrast

Step 2: Select Appropriate Palette Type

Reference: references/visualization_principles.md → "Color Theory" Resource: assets/color_palettes.json

Palette Types:

1. Qualitative (Categorical Data):

• Use when: Showing distinct categories with no order
• Limit: 5-8 colors maximum
• Example: Product lines, regions, departments
• Recommended: Okabe-Ito (colorblind-safe), Tableau10

# Okabe-Ito palette (colorblind-safe)
colors = ['#E69F00', '#56B4E9', '#009E73', '#F0E442',
          '#0072B2', '#D55E00', '#CC79A7']

2. Sequential (Ordered Data):

• Use when: Showing low-to-high progression
• Example: Population density, temperature, sales volume
• Recommended: Single hue (light to dark blue, green, etc.)

3. Diverging (Data with Midpoint):

• Use when: Data has meaningful center (zero, average)
• Example: Positive/negative, above/below average, correlation (-1 to +1)
• Recommended: Red-Blue, Brown-Teal (colorblind-safe)

4. Semantic (Meaning-Based):

• Use when: Colors have conventional meanings
• Green = positive/success, Red = negative/danger, Yellow = warning
• Caution: Cultural differences exist

Step 3: Ensure Accessibility

Colorblind Considerations:

• 8% of males, 0.5% of females have color vision deficiency
• Most common: Red-green colorblindness

Best Practices:

1. Use colorblind-safe palettes (Okabe-Ito, Viridis)
2. Don't rely on color alone - add patterns, labels, or shapes
3. Test with simulator (Color Oracle, Coblis)
4. Ensure contrast - 4.5:1 for text, 3:1 for graphics (WCAG 2.1)

Safe Color Combinations:

• Blue + Orange (excellent contrast)
• Blue + Yellow
• Purple + Green
• Brown + Teal

Avoid:

• Red + Green (most problematic)
• Blue + Purple (for some types)
• Light green + Yellow

Step 4: Apply Color Strategically

Emphasis Pattern:

• Main data: Primary color (e.g., blue)
• Highlight: Accent color (e.g., red/orange)
• Background data: Gray (de-emphasize)

Example:

# All bars in gray except the maximum
colors = ['#CCCCCC'] * len(data)
colors[max_index] = '#E63946'  # Highlight in red

Example: Colorblind-Safe Dashboard

Scenario: Create dashboard viewable by all users

Solution:

1. Use Okabe-Ito palette for categories
2. Use Blue-Orange for comparisons
3. Add patterns to bars (diagonal lines, dots)
4. Include value labels directly on charts
5. Test with Color Oracle simulator

# Implementation
from visualization_templates import executive_summary_template

kpis = [
    {'title': 'Revenue', 'value': 1234567, 'change_pct': 12.5},
    {'title': 'Customers', 'value': 5432, 'change_pct': 5.2}
]

fig = executive_summary_template(kpis, title='Q4 Dashboard')
fig.savefig('accessible_dashboard.png', dpi=300)

Workflow 3: Designing Professional Dashboards

Step 1: Define Dashboard Purpose and Audience

Dashboard Types:

1. Strategic (Executive):

• Audience: C-level executives
• Content: High-level KPIs, trends, targets
• Update: Weekly/monthly
• Focus: Strategic decision-making

2. Operational:

• Audience: Managers, operations team
• Content: Real-time metrics, status indicators
• Update: Real-time to hourly
• Focus: Day-to-day operations

3. Analytical:

• Audience: Analysts, data teams
• Content: Detailed data, drill-down, filters
• Update: Daily or on-demand
• Focus: Deep exploration

Step 2: Apply Information Hierarchy

Reference: references/dashboard_design.md

F-Pattern Layout (Western Reading):

┌─────────────────────────────┐
│ 1. Most Important (Top-Left)│
│    ↓                        │
│ 2. Secondary (Middle)       │
│    ↓                        │
│ 3. Details (Bottom)         │
└─────────────────────────────┘

Layout Guidelines:

• Top Row: 3-5 KPI cards with sparklines
• Middle: Primary visualization (largest chart)
• Bottom: Supporting charts and tables
• Whitespace: 20-30% of dashboard should be empty space

Step 3: Design KPI Cards

Components:

• Metric Name: Clear, concise (10-12pt)
• Value: Large, bold (28-36pt)
• Change Indicator: Arrow + percentage with color
• Sparkline: Optional mini-trend (last 7-30 periods)
• Comparison: "vs last month/year"

Color Coding:

↑ Green: Positive change
↓ Red: Negative change
→ Gray: No significant change

Example:

from visualization_templates import kpi_card

fig = kpi_card(
    value=1234567,
    title='Monthly Revenue',
    change_pct=12.5,
    sparkline_data=[100, 105, 102, 108, 115, 120]
)

Step 4: Maintain Consistency

Style Guide Elements:

• Font: Single family (Arial, Helvetica, Roboto)
• Colors: Consistent palette across all charts
• Spacing: Regular margins (16px, 24px, 32px increments)
• Grid: 12-column system for alignment
• Chart styles: Same border, gridline, and label format

Example: Executive Dashboard

Scenario: Create monthly executive summary

Requirements:

• 4 key metrics (Revenue, Customers, Retention, NPS)
• 2 trend charts (Revenue over time, Regional breakdown)
• Suitable for presentation
• Print-friendly

Implementation:

from visualization_templates import executive_summary_template

kpis = [
    {
        'title': 'Revenue',
        'value': 1234567,
        'change_pct': 12.5,
        'sparkline': [100, 105, 102, 108, 115, 120]
    },
    {
        'title': 'Customers',
        'value': 5432,
        'change_pct': 5.2,
        'sparkline': [90, 92, 95, 97, 98, 100]
    },
    {
        'title': 'Retention',
        'value': 89.5,
        'change_pct': 1.2,
        'sparkline': [85, 86, 87, 88, 89, 89.5]
    },
    {
        'title': 'NPS',
        'value': 72,
        'change_pct': 8.0,
        'sparkline': [65, 67, 68, 70, 71, 72]
    }
]

fig = executive_summary_template(
    kpis=kpis,
    title='Q4 2024 Executive Summary',
    subtitle='October - December 2024',
    figsize=(16, 10)
)

fig.savefig('executive_summary.pdf', dpi=300, bbox_inches='tight')

Workflow 4: Creating Story-Driven Visualizations

Step 1: Define Your Narrative

Story Structure:

1. Context: Establish baseline or background
2. Insight: Reveal the key finding
3. Action: What should viewer do with this information?

Example:

• Context: "Historical sales have been flat for 3 years"
• Insight: "New product line drove 45% growth in Q4"
• Action: "Invest more resources in new product development"

Step 2: Use Visual Emphasis

Techniques:

1. Color Contrast:

• Key element: Bright color (red, orange)
• Others: Gray or muted colors

# Highlight one bar
colors = ['#CCCCCC'] * 5
colors[3] = '#E63946'  # Highlight 4th bar

2. Annotations:

• Add arrows, text boxes, or callouts
• Keep text concise (5-8 words)
• Position near relevant data

ax.annotate('45% increase',
            xy=(date, value),
            xytext=(10, 10),
            textcoords='offset points',
            fontsize=12,
            fontweight='bold',
            color='#E63946',
            arrowprops=dict(arrowstyle='->', color='#E63946'))

3. Size:

• Make important elements larger
• Reduce size of supporting elements

4. Position:

• Place critical insight in top-left (F-pattern)

Step 3: Add Context

Essential Elements:

• Title: Clear, descriptive, includes key insight
• Axes Labels: Include units (¥, %, etc.)
• Legends: When needed, but prefer direct labeling
• Source: Data source and date
• Notes: Important caveats or methodology

Good Title Examples:

❌ "Sales Over Time"
✅ "Sales Jumped 45% After New Product Launch in Q4"

❌ "Customer Satisfaction"
✅ "Customer Satisfaction Reaches All-Time High of 8.9/10"

Step 4: Progressive Disclosure

For Complex Stories:

1. Start with summary/conclusion
2. Show high-level view
3. Allow drill-down to details
4. Provide data table for reference

Dashboard Pattern:

Top: Summary statement + key number
Middle: Primary chart showing main insight
Bottom: Supporting details and breakdowns

Example: Product Launch Analysis

Scenario: Show impact of new product launch on revenue

Story:

• Context: Revenue flat for 12 months
• Event: New product launched in September
• Impact: 45% revenue increase in Q4
• Action: Scale up production and marketing

Visualization Approach:

1. Line chart showing 18-month revenue trend
2. Vertical line marking launch date
3. Annotation highlighting 45% increase
4. Color: Gray before launch, Blue after launch
5. Title: "New Product Drives 45% Revenue Growth in Q4"

import matplotlib.pyplot as plt
import pandas as pd

fig, ax = plt.subplots(figsize=(12, 6))

# Pre-launch (gray)
ax.plot(dates[:9], revenue[:9], color='#CCCCCC', linewidth=2, label='Before Launch')

# Post-launch (blue)
ax.plot(dates[8:], revenue[8:], color='#2E86AB', linewidth=3, label='After Launch', marker='o')

# Launch date line
ax.axvline(launch_date, color='#666666', linestyle='--', linewidth=2, alpha=0.7)
ax.text(launch_date, ax.get_ylim()[1], ' Product Launch', ha='left', va='top')

# Highlight annotation
ax.annotate('↑ 45% increase',
            xy=(dates[-1], revenue[-1]),
            xytext=(-50, 20),
            textcoords='offset points',
            fontsize=14,
            fontweight='bold',
            color='#E63946',
            arrowprops=dict(arrowstyle='->', lw=2, color='#E63946'))

ax.set_title('New Product Drives 45% Revenue Growth in Q4', fontweight='bold', fontsize=16)
ax.set_ylabel('Revenue (¥ Million)', fontweight='bold')
ax.legend()
ax.grid(alpha=0.3)
plt.tight_layout()

Workflow 5: Ensuring Accessibility and Readability

Step 1: Test for Colorblindness

Tools:

• Color Oracle (free desktop app)
• Coblis - Color Blindness Simulator (online)
• Chrome DevTools Accessibility features

Process:

1. Create visualization
2. Run through colorblind simulator
3. Check if information is still clear
4. If not, add patterns, labels, or adjust colors

Step 2: Check Contrast Ratios

WCAG 2.1 Standards:

• Normal text: 4.5:1 minimum
• Large t

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: takusaotome
• Source: takusaotome/claude-skills-library
• License: MIT
• Homepage: https://takusaotome.github.io/claude-skills-library/

Install and usage instructions live in the source repository linked above.