Color Encoding in Data Visualization: The Rules That Prevent Misreading

Color in data visualization is not decorative — it is encoding. Every color choice maps a visual attribute to a data attribute, and that mapping either works or it does not. The failure modes are systematic and well documented, which makes them preventable. The first failure is using a sequential (single-hue) palette for categorical data. Sequential palettes — blue to white, or light green to dark green — imply order. They work for showing more/less, low/high, early/late. Applied to categorical data (country, product type, department), they imply a ranking that does not exist. The viewer's eye reads the darkest category as 'more' of something and the lightest as 'less', even when the categories have no meaningful order. The fix: use qualitative palettes with distinct hues and similar values for unordered categorical data. The second failure is rainbow color schemes. The rainbow palette — red, orange, yellow, green, blue, violet — seems intuitive because it is familiar, but it fails as a data encoding tool for several reasons. Hue is perceived non-linearly across the rainbow: the transition from yellow-green to green reads as smaller than the transition from blue to violet even when the underlying data intervals are identical. Rainbow palettes also perform poorly for color-vision-deficient viewers (about 8% of men) because large hue spans become indistinguishable. Better sequential palettes use lightness variation as the primary encoding dimension, with hue as secondary. The third failure is encoding data magnitude with saturation alone (without value). High saturation reads as emphatic and attention-grabbing, but saturation alone does not clearly communicate quantitative magnitude. A medium-saturation red and a high-saturation red read as 'present' and 'very present' rather than as specific values on a scale. Value — lightness — is the most reliably perceived quantitative dimension in color. Well-designed sequential palettes run from light (low data value) to dark (high data value) with consistent value steps and minimal hue change. Color-vision deficiency affects data communication in predictable ways. Red-green combinations — the most common palette for showing positive/negative values — are completely invisible to about 5% of male viewers. Standard alternatives: blue-orange (works for protanopia and deuteranopia), blue-red (works for most CVD types except rare tritanopia), or using both hue and texture/pattern encoding simultaneously.