Color in Data Visualization: Rules That Protect Meaning

Color in data visualization serves a fundamentally different purpose than color in brand design. In branding, color communicates personality. In data visualization, color communicates quantitative meaning — and when it does so imprecisely, it introduces systematic error into how readers interpret data. **Perceptual Uniformity** The most critical property a data visualization color scale can have is perceptual uniformity: equal steps in data value should appear as equal steps in visual difference. Naive rainbow color scales (red → orange → yellow → green → blue → violet) fail this test dramatically — yellow appears lighter and more visually similar to adjacent colors than the jumps between red-orange or blue-violet, creating the false impression that the data has valleys and peaks in the wrong places. Perceptually uniform scales (like Viridis, Inferno, or carefully constructed chroma-lightness ramps) eliminate this distortion. **Sequential vs. Diverging Scales** Sequential scales — running from light to dark within a single hue family — work for data that has a meaningful low end and high end with no neutral midpoint. Examples: population density, temperature above zero, revenue. Diverging scales — running from one hue through a neutral midpoint to a different hue — work for data with a meaningful zero or neutral point. Examples: temperature above/below freezing, survey agreement scales, profit/loss. Using a sequential scale on diverging data (or vice versa) systematically misleads readers about the structure of the data. **Categorical Color in Visualization** For categorical data (group A, B, C…), color needs to be simultaneously distinguishable, not implying order, and accessible to color-blind readers. The practical constraint is that human color discrimination saturates at around 8-10 clearly distinguishable categories — beyond that, charts need supplementary encoding (pattern, shape, label). Within those 8-10, hue spacing should maximize perceptual distance, which usually means distributing around the hue wheel rather than using adjacent hues. Desaturated categorical palettes also work because they reduce competition for attention between data colors and background/axis elements. **Accessibility in Charts** Approximately 8% of men and 0.5% of women experience some form of color vision deficiency. The most common type (deuteranopia/protanopia) affects red-green discrimination — making the most common data visualization convention (red = bad, green = good) partially or fully inaccessible. Accessible chart design combines color with a second encoding: position, shape, or pattern. When color-only encoding is unavoidable, use blue-orange or blue-yellow contrasts instead of red-green pairs.