Choosing Color Palettes for Charts and Data Visualization

Color in a chart is not decoration. It is meaning. Done well, it lets a reader separate series, follow a trend, or spot an outlier before they read a single label. Done badly, it invents patterns that do not exist and hides the ones that do.

The good news is that chart color is not a question of taste. It is a question of type. Match the palette to the kind of data you are showing and most of the hard problems solve themselves.

The three palette types

Every serious data-visualization system, from academic style guides to modern charting libraries, converges on three categories. Choosing the right one is the single most important decision.

Categorical palettes distinguish groups that have no order — regions, product lines, teams. The colors should be as different from one another as possible, with roughly similar visual weight, so no category looks more important than another. Humans can reliably tell apart about six to eight colors; beyond that, a legend becomes a reading exercise. If you have more categories, consider grouping, small multiples, or a different chart type.

Sequential palettes encode a single ordered variable that goes from low to high — revenue, density, temperature. They are built by varying lightness within one or two related hues, so the eye reads "more" as either darker or more saturated. The critical rule: sequential scales must change in one direction only. A scale that gets darker, then lighter, then darker again will make readers misread the data.

Diverging palettes encode a value with a meaningful midpoint — change from last quarter, deviation from average, sentiment from negative to positive. Two hues run outward from a neutral middle. The center should be visually quiet (often a light gray) so that the extremes carry the emphasis, and the two directions should be balanced in weight.

Matching palette to data

The mapping is straightforward once you know the question each chart answers:

• Asking "which group is this?" → categorical.
• Asking "how much, where more is further along one scale?" → sequential.
• Asking "which way has it moved from a baseline?" → diverging.

Most chart-color mistakes come from using the wrong intent: a rainbow of eight hues to show a single continuous value, or a subtle two-tone scheme to distinguish eight unrelated categories. Both make the reader work harder than necessary.

Accessibility is not optional

Roughly one in twelve men has some form of color vision deficiency, so charts that depend on color alone will fail a meaningful share of any audience. A few habits make visualization accessible without sacrificing clarity:

• Give every color a backup channel. Distinguish series with direct labels, line style, or marker shape in addition to color. This is the same principle behind our guide to accessible palettes for color blindness.
• Test lightness, not just hue. Colors that look distinct to you may collapse to the same gray for a reader with deuteranopia. If the series are still distinguishable after you remove all color, the chart works.
• Keep text contrast high. Axis labels, values, and legends are text and must meet WCAG contrast thresholds, even though the plot marks themselves are exempt.
• Avoid red-green pairs as the only distinction. They are the most common confusion. Blue and orange, or blue and red, separate far more reliably.

Practical rules that hold up

• Limit the palette. Fewer colors, used consistently, beat many. Around six categorical colors is a sensible ceiling before grouping.
• Reserve your strongest color. If one series is the story — the current year, the highlighted region — give it the most saturated color and mute the rest. Emphasis, not equality, is often what you want.
• Be consistent across charts. The same category should be the same color in every view; otherwise readers rebuild their mental legend each time.
• Do not invent a scale. Use perceptually even palettes designed for data rather than sampling colors from a brand deck. Brand colors can be used for accents and framing, not for encoding continuous values.
• Let the background recede. Light, low-saturation surfaces keep the data as the subject. Gridlines should be quiet.

Build your own

You do not need a design degree to make good chart colors. A working recipe:

1. Decide the palette type from the data's intent.
2. For categorical, pick six hues spaced around the wheel and pull them to similar lightness.
3. For sequential, pick one hue and generate a lightness ramp — the same process as a monochromatic palette.
4. For diverging, pick two hues that already sit opposite each other and join them with a subtle neutral.
5. Run a color-blindness check and add a second distinguishing channel to any series that survives on color alone.

Frequently asked questions

How many colors should a chart have?

Aim for six or fewer categorical series. Beyond that, readers stop distinguishing and start consulting the legend, which defeats the purpose of color.

Should I use my brand colors in charts?

Yes for framing, titles, and highlights; be cautious using them to encode data, because brand palettes are chosen for identity, not for perceptual separation or accessibility.

What is the most common chart-color mistake?

Using a rainbow or multi-hue ramp for a single continuous value. It forces readers to memorize an arbitrary color order instead of reading darkness as magnitude.

How do I check accessibility?

Desaturate the chart and check that series remain distinguishable by lightness and shape, and verify the whole thing passes a color-blindness simulation. Our image color picker can help you sample and compare the exact values you are using.

Try palettes on real data

The most reliable way to judge a chart palette is to apply it and look. Generate candidate sets in the palette generator, check contrast in the playground, and read the WCAG contrast guide to keep labels legible while the data does the talking.