Color in Data Visualization: Sequential, Diverging, and Categorical Palettes

Data visualization color is a domain where aesthetic preference frequently overrides perceptual accuracy, often invisibly. A palette that looks beautiful to the designer can actively mislead the viewer if it applies the wrong color structure to the underlying data type. Understanding the taxonomy of data visualization palettes — sequential, diverging, categorical, and highlight — is the foundation of rigorous data design. Sequential palettes encode magnitude: the higher the value, the darker, more saturated, or more shifted in hue the color. The critical requirement is that the perceptual distance between color steps should be proportional to the data distance. A palette that steps from light yellow to dark blue in equal visual increments correctly encodes the fact that the data steps from low to high in equal increments. The problem with many default sequential palettes (including the notorious rainbow palette, which is perceptually non-monotonic) is that the color steps are not perceptually equal — some steps look like big jumps while others look small, misrepresenting the data even when the scale is perfectly linear. Diverging palettes encode deviation from a meaningful midpoint. A diverging palette for temperature anomaly data should have a neutral color at zero (the historical average) and increasingly warm and cool colors as the temperature deviates in each direction. The neutral midpoint must be visually neutral — a grey or light desaturated hue — not a visually salient hue that would make the zero position look significant. The two arms of a diverging palette should be perceptually balanced: equal perceived saturation and equal perceived lightness change from midpoint to endpoints. Categorical palettes encode nominal categories — groups with no inherent rank or order. The requirement here is maximum distinctiveness: each category should be visually separable from all others, including when printed in black and white, viewed under different lighting conditions, or seen by colorblind viewers. Standard categorical palettes limit themselves to 8-12 colors because beyond that, reliable distinctiveness is difficult to maintain. For colorblind accessibility, relying solely on hue is insufficient — luminance and saturation must also differ between categories. Highlight palettes use desaturated base colors for background data and a single saturated color to draw attention to the focus. This is the visual equivalent of whispering context while shouting the main point. It is the correct palette type when the story is about one specific data element relative to everything else, rather than a comparison among multiple elements.