From Business Intelligence
Color should be used to differentiate and encode elements of a visualization such as labels and measures, and especially to highlight important data. It should be used sparingly and carefully.
It is best to use different shades of the same color rather than many different colors to distinguish graphic elements. Bright colors draw attention away from the data itself. More than one bright color is garish and can cause "vibrating" at the edges, making the graphic difficult to look at. Color may also present problems for color-blind people, who represent 10% of the population. This is especially true for hierarchical or multivariate data - a "rainbow" (ROYBGIV) does not afford an obvious hierarchy, while varying shades of the same color do.
The examples above demonstrate this. Although the one on the left uses traffic light colors to convey what the targets are, it is difficult to look at (especially in a dashboard with dozens of these on one screen) and not accommodating of color-blind people. The second example shows clearly that the same information is gained by using just shades of gray with different saturation values. Perceptual Edge
These examples show a redesign of a black-and-white marshalling chart. The original chart had heavy grid lines that made the white space in between "vibrate", and and had too many visual similarities between body, lights and typography to easily interpret at a glance. Muting of the grid, careful addition of a few strong colors, and elimination of redundant "motion" lines around the signals greatly improves clarity of this visualization.
 Relevant Visualizations
- Bar chart
- Box plot
- Bullet graph
- Design considerations
- Line graph
- Main Page
- Network diagram
- Template:Home menu
- Graphics and web design based on Tufte's principles, University of Washington, <http://www.washington.edu/computing/training/560/zz-tufte.html#Colors>. Retrieved on 6 August 2008
- Edward R. Tufte (2001), The Visual Display of Quantitative Information, Graphics Press