Color in Data Visualization: The Rules That Make Charts Readable and Honest

Data visualization color is one of the few domains in design where there are objectively correct and incorrect choices. A pie chart colored in palette designed to look beautiful will almost certainly communicate less accurately than one designed around perceptual principles. The gap between aesthetically appealing color and perceptually accurate color is widest in visualization, which makes it one of the most important areas to understand the science before making decisions. Visualization color palettes fall into three functional types that correspond to three types of data: sequential (ordered data with a meaningful direction — temperature, quantity, intensity), diverging (data with a meaningful midpoint, such as profit/loss or positive/negative sentiment), and qualitative (categorical data with no inherent order — country, product category, team). Each requires a different color architecture. Sequential data uses a single hue that varies in lightness — from light (low values) to dark (high values) — because human vision perceives lightness variation as quantity with high reliability. Diverging data uses two hues moving away from a neutral midpoint. Qualitative data uses maximally distinct hues at similar lightness so no category appears more important than others. The most common data visualization color error is using a rainbow (spectral) palette for sequential data. Rainbow palettes violate two perceptual principles simultaneously: they have uneven lightness variation (the yellow band is perceptually much lighter than the green and blue bands, creating a false emphasis), and they include multiple hues that create categorical perceptual boundaries at arbitrary data values. When a rainbow palette shows a boundary between green and blue regions in a map, viewers interpret that boundary as meaningful even if it simply corresponds to a hue transition in the palette. Spectral palettes create artifacts that are not in the data. Colorblind accessibility in visualization is non-negotiable given that color is the primary encoding channel. The practical approach is to use palettes validated against deuteranopia and protanopia simulation for all charts where color carries critical information. For qualitative palettes, this means avoiding red-green combinations as the primary distinction between categories. The IBM Data Visualization palette, the ColorBrewer schemes, and Vega's built-in palettes are all designed with colorblind safety as a constraint — using any of them as a starting point is significantly more reliable than designing from scratch. For critical applications, verify with actual simulation tools before publishing.