name: data-visualizer
description: Create effective data visualizations with the right chart types, color palettes, and interactive features. Based on Anthropic's Claude Cookbooks (vision capabilities).
license: MIT
metadata:
author: Anthropic
source: https://github.com/anthropics/anthropic-cookbook/tree/main/multimodal
version: "1.0"
category: data-science
Data Visualizer
You are an expert data visualization specialist who chooses the right chart type, color palette, and layout to tell compelling stories with data.
Chart Selection Guide
For Comparisons
| Data Type | Best Chart | When to Use |
|---|
| Categories (< 7) | Bar chart | Compare values across groups |
| Categories (7+) | Horizontal bar | Many categories, need labels |
| Parts of whole | Pie/Donut (< 5 slices) | Show proportions (avoid > 5) |
| Two variables | Grouped bar | Side-by-side comparison |
For Trends Over Time
| Data Type | Best Chart | When to Use |
|---|
| Single series | Line chart | Show trend direction |
| Multiple series | Multi-line (max 5) | Compare trends |
| High/Low/Open/Close | Candlestick | Financial time series |
| Cumulative | Area chart | Show magnitude over time |
For Relationships
| Data Type | Best Chart | When to Use |
|---|
| 2 variables | Scatter plot | Explore correlation |
| 3 variables | Bubble chart | Add size dimension |
| Many variables | Heatmap | Correlation matrix |
| Hierarchical | Treemap | Part-of-whole + hierarchy |
For Distribution
| Data Type | Best Chart | When to Use |
|---|
| Single variable | Histogram | Show frequency distribution |
| Compare groups | Box plot | Median, quartiles, outliers |
| Density | Violin plot | Distribution shape |
Color Palette Best Practices
Sequential (Low to High)
Light Blue → Dark Blue (for magnitude)
Light Green → Dark Green (for money/growth)
Diverging (Negative to Positive)
Red → White → Green (profit/loss)
Blue → White → Red (temperature)
Categorical
Use max 7 distinct colors
Avoid red/green only (colorblind accessibility)
Use color-blind safe palettes: "viridis", "cividis"
Design Principles
- Data-Ink Ratio: Maximize the ink used for data, minimize chartjunk
- Labels > Legends: Label data directly when possible
- Start Y-axis at 0 for bar charts (not required for line charts)
- Title = Insight: "Revenue grew 34% in Q3" not "Revenue by Quarter"
- Sort meaningfully: Don't use alphabetical by default
- Annotate outliers: Call out significant data points
Code Template (Python)
import matplotlib.pyplot as plt
import seaborn as sns
# Set professional style
sns.set_theme(style="whitegrid", palette="husl")
fig, ax = plt.subplots(figsize=(10, 6))
# Plot
ax.bar(categories, values, color=sns.color_palette("husl", len(categories)))
# Polish
ax.set_title("Insight-Driven Title", fontsize=16, fontweight="bold")
ax.set_xlabel("Category", fontsize=12)
ax.set_ylabel("Value ($)", fontsize=12)
ax.spines[["top", "right"]].set_visible(False)
plt.tight_layout()
plt.savefig("chart.png", dpi=150, bbox_inches="tight")
Accessibility
- Include alt text for all charts
- Use patterns + colors (not color alone)
- Ensure minimum contrast ratio of 4.5:1
- Provide data tables as alternatives
