Color in data visualization has exactly one purpose: to encode information. When color is used for any other reason — to make the chart look "professional," to match brand guidelines, to add visual interest — it is noise. And noise obscures signal.
This is a strong claim. It is also true.
The Encoding Function
Jacques Bertin identified the visual variables available to cartographers and chart designers: position, size, shape, value (lightness), color (hue), orientation, and texture. Color is one of seven channels. It is powerful for categorical distinctions — this line is product A, that line is product B — and for sequential encoding of magnitude, from light to dark.
Christopher Healey's research on preattentive attributes confirmed that color (hue) is processed by the visual system before conscious attention engages. A red dot in a field of blue dots is found instantly. This makes color uniquely effective for highlighting — drawing the reader's eye to a specific element.
But preattentive processing is a limited resource. When everything is colored, nothing stands out. A chart with seven bright colors for seven categories does not use preattentive processing — it overwhelms it. The reader must engage conscious attention to decode every element, which defeats the purpose.
The Grey-Plus-One Rule
The most effective use of color in most charts follows a simple rule: make everything grey except the one thing you want the reader to notice. That one thing gets color.
The reader's eye goes immediately to the Nordic bar. The context — other regions — is visible but subordinate. This is what color should do: create hierarchy. One element speaks; the rest listen.
When all bars are different colors, the chart is shouting five things at once. The reader hears none of them clearly.
Sequential Color
For encoding magnitude — a heatmap, a choropleth map, a gradient fill — sequential color palettes are essential. These move from light to dark in a single hue, or across a perceptually uniform gradient like viridis.
The critical requirement is perceptual uniformity: equal steps in data should correspond to equal perceived steps in color. The rainbow palette — red through yellow through green through blue — fails this test catastrophically. Yellow appears brighter than every other color, creating a false peak in the middle of the scale. Despite this, the rainbow palette remains one of the most commonly used palettes in scientific visualization.
Cynthia Brewer's ColorBrewer palettes, designed with perceptual science in mind, solve this problem for most use cases. They are available in nearly every visualization tool and should be the starting point for any color choice.
Accessibility
Roughly 8% of men and 0.5% of women have some form of color vision deficiency. A chart that relies on the distinction between red and green — the most common choice — is invisible to this population. This is not an edge case. In a meeting of twenty people, one or two likely cannot read the chart.
The solution is redundant encoding: use color and another visual variable — shape, pattern, or direct labeling — to distinguish categories. This is good design for everyone, not just those with color vision deficiency. Redundant encoding makes charts readable in black-and-white printouts, in low-contrast projector environments, and on poor-quality screens.
Brand Color
The most common misuse of color in corporate visualization is brand compliance. The marketing department mandates that all charts use the company's brand palette — a palette designed for logos and advertisements, not for data encoding. Brand palettes are often too saturated, insufficiently distinct in lightness, and limited to three or four colors that do not form a usable sequential or diverging scale.
Data visualization color should be determined by perceptual science, not by the brand guidelines document. If these conflict — and they almost always do — perception must win.