What Makes a Good Graph
A good graph does more than display data - it tells a story, reveals insights, and drives action. After analyzing millions of visualizations, here are the principles that separate effective graphs from chart junk.
1. Answer a Specific Question
Every graph should answer a clear question. Before creating a visualization, ask yourself: "What decision will this graph help someone make?" If you can't articulate the question, the graph probably isn't needed.
Good: "How have monthly sales changed over the past year?"
Bad: "Here's some data about sales."
2. Choose the Right Chart Type
Different data relationships call for different visualizations:
- Comparison: Bar charts for categorical data, grouped bars for multiple series
- Trend over time: Line charts show continuity and direction
- Part-to-whole: Pie charts (sparingly), stacked bars, or treemaps
- Correlation: Scatter plots reveal relationships between variables
- Distribution: Histograms and box plots show data spread
- Geographic: Maps when location matters
3. Maximize Data-Ink Ratio
Edward Tufte's principle remains essential: minimize non-data ink. Remove:
- Unnecessary gridlines
- Redundant labels
- 3D effects that distort perception
- Decorative elements that don't convey information
Every pixel should either represent data or help interpret it.
4. Make Comparisons Easy
The brain processes visual comparisons in a specific hierarchy:
- Position along a common scale (most accurate)
- Position on identical but non-aligned scales
- Length
- Angle/slope
- Area
- Color saturation (least accurate)
Use position when precision matters; reserve color for categorical distinctions.
5. Label Clearly and Directly
Labels should be:
- Immediate: Place labels near the data they describe
- Unambiguous: Include units and context
- Legible: Horizontal text, appropriate font size
- Complete: Title that states the insight, not just the topic
Good title: "Sales increased 23% after the marketing campaign"
Bad title: "Sales data"
6. Use Color Purposefully
Color should encode meaning, not decoration:
- Use consistent color schemes across related graphs
- Consider colorblind-friendly palettes
- Highlight the most important data point
- Use sequential colors for continuous data, diverging for values around a midpoint
7. Show Context and Scale
Data without context is meaningless:
- Include benchmarks or targets when relevant
- Show historical context for trends
- Start axes at zero unless you have a good reason not to
- If truncating, clearly indicate it
8. Make it Actionable
The best graphs lead to decisions. Include:
- Annotations for significant events or anomalies
- Reference lines for targets or thresholds
- Clear indication of what "good" looks like
How Helix Helps
Helix automatically applies these principles when generating visualizations from your data. Simply describe what you want to understand, and Helix will:
- Select the appropriate chart type for your data
- Apply clean, minimal styling
- Generate meaningful titles and labels
- Highlight key insights automatically
Ready to create effective visualizations? Try Helix now - just describe what you want to see.