Comparing Data
Comparison is the most fundamental task in data visualization. When comparing values, the human brain relies on length and position along a common scale as the most accurate pre-attentive attributes.
The Workhorse: Bar Charts
A bar chart encodes data via the length of rectangles. It is the most robust, unambiguous way to compare categorical data.
| Rule | Why |
|---|---|
| Always start Y-axis at 0 | Because bar charts encode via length, truncating the axis distorts the proportionality. A bar representing 100 should be twice as long as a bar representing 50. |
| Use horizontal bars for long labels | Prevents text rotation. Rotated text is notoriously difficult to read and strains the neck. |
| Sort the data | Unless the categories have a natural sequence (like age ranges), sort by value ascending or descending to facilitate rapid ranking. |
Dot Plots (Cleveland Dot Plots)
When you have a baseline that doesn't start at zero (e.g., comparing temperatures or index scores), or when you want to minimize ink, a dot plot encodes data using position along a common scale.
Common Mistakes
- Radar Charts (Spider Charts): Humans are terrible at comparing areas of irregular polygons. The arbitrary ordering of the axes drastically changes the area, misleading the reader. Just use a bar chart or small multiples.
- 3D Effects: Adding a third dimension to 2D data introduces occlusion (hiding data) and distortion (altering perceived size based on perspective).
Next Steps
If your data adds up to a whole, view our guide on Compositions. If you need to build a comparison chart, check our Aspect Ratio Calculator.