Three Jobs Color Does in Data Visualization — Decide on Color Last and the Chart Gets Stronger
Where do you start when you're picking colors for a chart? Most people start from "looks." But color has concrete jobs to do before aesthetics. Distinguish, emphasize, show order. Understand these three roles and palette choice turns from feel into design.
At the final step of making a chart, when you think "now to pick colors," do you ever just grab whatever you find pleasing from a palette? I had that habit for a long time. But while color is treated as the "final flourish," charts stay weak as carriers of information.
In data visualization, color does specific work. Not decoration, not mood. Color carries information. Good charts have form and color delivering meaning together. Don't bolt color on at the end — derive it from the structure of the message you want to send. That's the theme of this article.
Job 1: Distinguish (categorical)
Color's most basic job is telling readers "this and that are different." Palettes for distinguishing multiple series are called categorical palettes.
What matters most for categorical palettes is that the hues are sufficiently far apart. Place near-hues like blue and green, or red and orange, side by side and readers can't separate them at a glance. The reason world-default palettes (Tableau 10, ColorBrewer Qualitative, D3 category10) all share a similar 6–10 color makeup of blue / orange / green / red / purple isn't accidental — it's the result of careful "hue separation" design.
One thing to watch in categorical palettes: don't smuggle in meaning. Coloring "Revenue" red and "Cost" green creates a clash with cultural connotations (red = danger / negative; green = OK / positive) and momentarily confuses the reader. Color carries cultural baggage, so even categorical color choice deserves care.
Job 2: Emphasize (focus)
A chart that treats every element the same emphasizes nothing. The job of focus color is using color to say "look here."
Concretely, make the one (at most two) elements you want to highlight stand out, and dim everything else to gray or muted tones. The contrast between highlight and dim determines the chart's "main character."
What you emphasize is case-by-case: the maximum value, the latest value, your own brand, the items hitting target. The important thing is that the emphasis is narrowed to one. Highlighting five of ten bars is no longer emphasis — it's "half." Truly making just one or two stand out is what effective focus looks like.
A chart with nine muted colors and one strong one communicates far more than a chart with ten distinct colors.
Job 3: Show order (sequential / diverging)
The third job uses color itself to encode "the order of magnitude." Two flavors: sequential and diverging palettes.
Sequential palettes use a gradient from light to dark to encode value magnitude — for instance, "darker blue cells = more PV" in a heatmap. Hue stays fixed; only lightness or saturation varies. This creates the intuitive correspondence "larger value = darker color."
Diverging palettes radiate two different hues out from a "center reference value." For example, when visualizing "actual vs. budget ±20%," set zero to white, plus to blue, minus to red. Readers receive a double layer of information at once: "the further from white, the larger the deviation" and "hue tells you the direction (plus or minus)."
Where do you actually use ordered palettes? Heatmaps, choropleth maps (geographic data), contour plots — not bar or line charts. In a regular bar or line chart, the bar length or line height already encodes magnitude; using color for it too is just redundant.
Three problems caused by "colors I just liked"
Choose by feel and a few classic problems show up.
Problem 1: too many colors. Using everything in a colorful palette. The more colors there are, the less information each one carries. Categorical palettes beyond 3–5 colors usually reduce information rather than add it.
Problem 2: the element worth highlighting gets buried. Drawing every category in the same hue's gradient and ending up unable to convey "which one matters." When you make a chart, decide first "what do I want the reader to see?" — then have the courage to visually downgrade everything else.
Problem 3: forgetting accessibility for color vision. Difficulty distinguishing red and green affects roughly 1 in 12 men. The traditional "positive = green, negative = red" pairing doesn't work for them. As a known alternative, blue (positive) + orange (negative) is identifiable across nearly all color vision types.
A design process where color comes "last"
So when should color be decided? My recommendation: at the final step.
The order goes like this. First, put the message you want to send into words. "The Q4 campaign worked." "We're #3 in the industry." Write the conclusion as one sentence. Second, choose the chart type that conveys it best. Third, design axes and labels — data range, ticks, item names. Fourth, only now, decide on color, consciously asking which of the three jobs color is doing.
Decide color last in this process and color stops being decoration — it becomes a tool. "Highlight our company → only us in blue, others in gray." "Show direction of change → plus blue, minus red." Color derives automatically from the message. That's design.
"When in doubt, go grayscale" is the strongest rule
Finally, a rule of thumb I've used for years. When unsure about color, draw everything in gray first. If that conveys the message, you don't need color. Where the message fails to come through, add color only there.
Starting from grayscale is the strongest preventative against color overuse. Each addition forces you to ask "what job is this color doing?" Colors that can't answer that get cut. This alone makes charts dramatically stronger.
Color isn't a matter of sensibility — it's a matter of logic. Decide color last; ask which of the three jobs it's doing; when unsure, start from gray. With these three habits alone, the quality of your reports will rise sharply from tomorrow.