tufte-chart-design

name	tufte-chart-design
description	Apply Edward Tufte's visualization principles to produce data-honest, high data-ink ratio charts with minimal chartjunk
triggers	["make a chart","create a visualization","design a dashboard","improve this chart","what chart should I use","apply Tufte principles","visualize this data","build a graph"]

Tufte Chart Design Skill

Skill by ara.so — Claude Code Skills collection.

This skill teaches you to create charts following Edward Tufte's principles from his three foundational books: The Visual Display of Quantitative Information, Envisioning Information, and Visual Explanations. It turns "make me a chart" into a Tufte-compliant visualization with maximum data-ink ratio and zero chartjunk.

What This Skill Does

• Activates automatically when you're asked to create, improve, or critique any chart or visualization
• Applies ten core principles distilled from Tufte's work
• Chooses the right chart type based on data shape and analytical goal
• Removes chartjunk — 3D effects, unnecessary gridlines, redundant legends, pie charts, dual axes
• Produces clean output as inline SVG, React (Recharts), or D3-in-React
• Enforces graphical integrity — lie factor between 0.95 and 1.05, bars always start at zero

The Ten Principles

1. Show the data — Above all else. Every decision asks "does this help see the data?"
2. Maximize data-ink ratio — Ink that changes when data changes is data-ink. Everything else is overhead.
3. Erase non-data-ink — Remove frame boxes, gridlines, tick marks at minor units.
4. Erase redundant data-ink — One encoding per quantity. Don't show "37" as a bar + number + gridline + tick.
5. Graphical integrity — Bars start at zero. Never use area for 1D quantities. Lie factor = (size of effect shown) / (size of effect in data).
6. Small multiples — N small charts on shared scales beats one chart with N overlapping lines.
7. Layering and separation — Data on top, labels next, scaffolding faintest. Avoid 1+1=3 visual noise.
8. Micro/macro readings — One shape from across the room, every point legible up close.
9. Smallest effective difference — Gray for context, one accent for focus. Subtle but clear.
10. Word-data integration — Sparklines in sentences. Numbers next to visuals. Tables are charts.

When This Skill Activates

Auto-triggers on:

• "Make a chart of...", "visualize...", "graph...", "plot..."
• "Build a dashboard", "design a KPI tile"
• "Improve this chart", "critique this visual"
• "What chart should I use for...?"
• /tufte command or mentioning Tufte by name
• Any code editing session involving chart/visualization code

Does NOT trigger for:

• Logos, illustrations, decorative graphics
• Flowcharts, architecture diagrams, wireframes
• Pure typography or layout without quantitative data

Chart Selection Guide

Data Shape	Goal	Use	NOT
One time series	Show trend	Sparkline or line chart	Area chart, bar chart
Multiple time series	Compare trends	Small multiples (separate charts)	Multi-line spaghetti
Categories (≤10)	Compare values	Sorted dot plot or bar chart	Pie, donut, 3D bars
Categories (>10)	Show distribution	Strip plot or histogram	Rainbow color scales
Two time points	Show change	Slopegraph	Clustered bars
Part-to-whole	Show composition	Sorted table with percentages	Pie, donut, stacked bars
Process stages	Show flow	Horizontal bars + percentages	3D funnel
Correlation	Show relationship	Scatterplot	Dual-axis line chart
Cohorts over time	Show retention	Sequential heatmap + sparklines	Rainbow heatmap
Geographic	Show regional pattern	Choropleth (single-hue sequential)	Rainbow choropleth

Kill List — Never Use

❌ 3D effects — distort area perception, add no information
❌ Pie charts — humans can't compare angles; use sorted table instead
❌ Donut charts — same problem as pie, worse
❌ Dual-axis charts — viewer can't tell if correlation is real or axis-manipulation
❌ Rainbow color scales — perceptually non-uniform; use sequential single-hue
❌ Gradient fills — decorative ink, not data-ink
❌ Drop shadows, glows — chartjunk
❌ Chart frames/boxes — erase, data should float
❌ Heavy gridlines — if you need them, print values instead
❌ Redundant legends — direct-label instead

HTML/SVG Output Pattern

When asked for inline SVG or given no preference, use this pattern:

<!DOCTYPE html>
<html>
<head>
  <meta charset="UTF-8">
  <style>
    body {
      font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif;
      font-size: 11px;
      color: #1a1a1a;
      background: white;
      padding: 40px;
    }
    .label { font-size: 10px; fill: #666; }
    .value { font-size: 13px; fill: #1a1a1a; font-weight: 500; }
    .data-point { fill: #1a1a1a; }
    .data-line { stroke: #1a1a1a; stroke-width: 1.5; fill: none; }
    .reference-line { stroke: #e0e0e0; stroke-width: 1; }
    .accent { fill: #d73027; }
  </style>
</>

  
    
  

Sparkline Example

<div style="font-family: sans-serif; padding: 20px;">
  <div style="font-size: 11px; color: #666; margin-bottom: 4px;">Monthly Active Users</div>
  <div style="display: flex; align-items: baseline; gap: 12px;">
    <span style="font-size: 24px; font-weight: 500; color: #1a1a1a;">247K</span>
    <svg width="120" height="24" style="display: block;">
      <polyline 
        points="0,18 10,15 20,16 30,14 40,12 50,13 60,10 70,9 80,11 90,8 100,6 110,4 120,3"
        stroke="#1a1a1a" 
        stroke-width="1.5" 
        fill="none"
      />
      <text x="0" y="24" font-size="9" fill="#999">Jan</text>
      <text x="110" y="24" font-size="9" = =>Dec
    
  

Dot Plot Example

<svg width="400" height="180" viewBox="0 0 400 180">
  <!-- Sorted by value, descending -->
  <circle cx="320" cy="20" r="4" fill="#1a1a1a"/>
  <text x="330" y="24" font-size="11" fill="#1a1a1a">Product A</text>
  <text x="310" y="24" font-size="11" fill="#666" text-anchor="end">$247K</text>
  
  <circle cx="280" cy="45" r="4" fill="#1a1a1a"/>
  <text x="290" y="49" font-size="11" =>Product B
  $213K
  
  
  Product C
  $145K
  
  
  
  $0
  $250K

React/Recharts Output Pattern

When asked for React or Recharts:

import { LineChart, Line, ResponsiveContainer, XAxis, YAxis } from 'recharts';

const data = [
  { month: 'Jan', value: 180 },
  { month: 'Feb', value: 195 },
  { month: 'Mar', value: 205 },
  { month: 'Apr', value: 198 },
  { month: 'May', value: 215 },
  { month: 'Jun', value: 247 },
];

export default function RevenueChart() {
  return (
    <div style={{ fontFamily: 'system-ui, sans-serif' }}>
      <div style={{ fontSize: 11, color: '#666', marginBottom: 4 }}>
        Monthly Revenue (thousands)
      </>
      
        
          
          
          
        
      
      
        $247K
      
    
  );
}

Recharts Theme Config

Apply these props to strip chartjunk:

// Universal settings
const tufteTheme = {
  xAxis: {
    axisLine: false,
    tickLine: false,
    tick: { fontSize: 10, fill: '#999' },
  },
  yAxis: {
    axisLine: false,
    tickLine: false,
    tick: { fontSize: 10, fill: '#999' },
    // Often better to hide Y-axis and direct-label instead
  },
  grid: {
    strokeDasharray: '3 3',
    stroke: '#f0f0f0', // very faint if you must show
  },
  // NO tooltip by default — print values on chart
  // NO legend — direct label instead
};

D3-in-React Pattern (for Custom Charts)

For slopegraphs, custom dot plots, small multiples:

import { useEffect, useRef } from 'react';
import * as d3 from 'd3';

export default function Slopegraph({ data }) {
  const svgRef = useRef();
  
  useEffect(() => {
    const svg = d3.select(svgRef.current);
    const width = 400;
    const height = 300;
    const margin = { top: 20, right: 80, bottom: 20, left: 80 };
    
    // Clear previous render
    svg.selectAll('*').remove();
    
    const y = d3.scaleLinear()
      .domain([0, d3.max(data, d => Math.max(d.q2, d.q3))])
      .range([height - margin.bottom, margin.top]);
    
    // Draw lines
    svg.selectAll()
      .(data)
      .()
      .(, )
      .(, margin.)
      .(, width - margin.)
      .(,  (d.))
      .(,  (d.))
      .(,  d. < d. ?  : )
      .(, );
    
    
    svg.()
      .(data)
      .()
      .(, )
      .(, margin. - )
      .(,  (d.))
      .(, )
      .(, )
      .(, )
      .(, )
      .( );
    
    
    svg.()
      .(data)
      .()
      .(, )
      .(, width - margin. + )
      .(,  (d.))
      .(, )
      .(, )
      .(, )
      .(, )
      .( d.);
    
    
    svg.()
      .(, margin.)
      .(, )
      .(, )
      .(, )
      .(, )
      .();
    
    svg.()
      .(, width - margin.)
      .(, )
      .(, )
      .(, )
      .(, )
      .();
      
  }, [data]);
  
   ;
}

Small Multiples Pattern

When comparing multiple series, create separate charts on a shared scale instead of overlaying:

const regions = ['North', 'South', 'East', 'West'];
const sharedYDomain = [0, 300]; // consistent scale across all

return (
  <div style={{ display: 'grid', gridTemplateColumns: 'repeat(2, 1fr)', gap: 20 }}>
    {regions.map(region => (
      <div key={region}>
        <div style={{ fontSize: 11, color: '#666', marginBottom: 4 }}>
          {region}
        </div>
        <ResponsiveContainer width="100%" height={100}>
          <LineChart data={data.filter(d => d.region === region)}>
            <YAxis domain={sharedYDomain} hide />
            <XAxis dataKey="month" tick=   }} />
            
          
        
      
    ))}
  
);

Dashboard KPI Tile Pattern

Replace giant number tiles with table + sparkline + target delta:

const kpis = [
  { metric: 'Revenue', value: 247, trend: [180,195,205,198,215,247], target: 250, unit: 'K' },
  { metric: 'Users', value: 12.4, trend: [10.2,10.8,11.1,11.5,12.0,12.4], target: 12, unit: 'K' },
  { metric: 'Churn', value: 3.2, trend: [2.1,2.4,2.7,2.9,3.1,3.2], target: 2.5, unit: '%' },
];

return (
  <table style={{ fontFamily: 'sans-serif', fontSize: 11, borderCollapse: 'collapse' }}>
    <thead>
      
        Metric
        Current
        Trend
        vs Target
      
    
    
      {kpis.map(kpi => {
        const delta = ((kpi.value - kpi.target) / kpi.target * 100).toFixed(0);
        const isBad = (kpi.metric === 'Churn' ? delta > 0 : delta < 0);
        return (
          
            {kpi.metric}
            
              {kpi.value}{kpi.unit}
            
            
              
            
            
              {delta > 0 ? '+' : ''}{delta}%
            
          
        );
      })}
    
  
);

Pre-Publish Checklist

Before declaring a chart done, verify:

• Data-ink ratio > 0.7 — most ink changes when data changes
• No chartjunk — no 3D, gradients, shadows, decorative elements
• Lie factor 0.95–1.05 — visual effect matches data effect
• Bars start at zero — if showing magnitude
• Direct labels — no legend unless unavoidable
• Sorted by value — categories ordered by size, not alphabet
• One accent max — gray for context, one color for focus
• Readable at two scales — shape visible from distance, values legible up close
• No dual axis — if tempted, make small multiples instead
• Numbers printed — if precision matters, show the value
• Appropriate chart type — matches data shape + analytical goal
• All ink necessary — tried removing every element; only kept what's essential

Common Overrides

If a user explicitly requests something on the kill list:

"Make a pie chart — the CFO wants a pie"

Comply, but add a comment noting the Tufte alternative:

<!-- Pie chart as requested. Tufte alternative: sorted table with percentages. -->
<svg>...</svg>

User stays in control; you provide the alternative for reference.

Color Rules

• Default: Grayscale. Black for data, mid-gray (#666) for labels, light-gray (#e0e0e0) for reference lines.
• One accent: Use red (#d73027) or blue (#4575b4) to highlight one series or flag an outlier.
• Sequential scales: Single-hue (light to dark). Never rainbow.
• Diverging scales: Two hues meeting at neutral (red-white-blue). Equal steps.
• Categorical: Only if categories have no natural order. Max 5 colors. Consider direct-label + grayscale instead.

Troubleshooting

Issue	Fix
"Chart feels cramped"	Increase padding, reduce font size, or use small multiples
"Can't read labels"	Direct-label at data points instead of axis; rotate if needed
"Too much white space"	Tufte prefers it — resist urge to fill; let data breathe
"Legend overlaps data"	Remove legend; direct-label each line/bar at its endpoint
"Gridlines requested"	Print values on chart instead; use faint reference line only at key thresholds
"Dual-axis requested"	Make two separate charts with shared X-axis; stack vertically

Configuration

No installation needed — this skill is already active in your environment.

Override strictness per request:

• Default: Apply Tufte principles unless user explicitly asks otherwise
• User can say "ignore Tufte for this one" or "CFO wants X" and you comply (with a note of the alternative)

Stack preference (auto-detected from user request):

• "inline SVG" or no preference → self-contained HTML/SVG
• "in React" or "with Recharts" → Recharts component
• "with D3" or "custom layout" → D3-in-React

Reference

This skill distills:

• The Visual Display of Quantitative Information (1983, 2nd ed. 2001) — data honesty, data-ink ratio, lie factor, small multiples
• Envisioning Information (1990) — layering, micro/macro, escaping flatland
• Visual Explanations (1997) — smallest effective difference, narrative, visual confections

Core Tufte quote (VDQI p.105):

Above all else show the data. Maximize the data-ink ratio. Erase non-data-ink. Erase redundant data-ink. Revise and edit.

---

When in doubt: Show the data. Remove everything else. Let the numbers speak.