Cheat sheet
Palettes by industry and by mood, each shown in a sample dashboard layout. Everything is copy-and-paste ready — and there's a glossary at the end if any of the vocabulary is new.
Click any swatch to copy its hex · click Copy to take the whole rampDoes position already separate these?
Bar charts, columns, dot plots — the axis has already done the work. Adding a hue per category spends color on information the reader can already see.
Gray everything, one accent hue on the series you're talking about.
Six hues encode what the axis already says. Color is now unavailable for anything meaningful.
Position does the separating. Color is saved to say one thing: look here.
Do the categories have an order?
Tiers, satisfaction bands, age brackets, deal stages, sizes. These look categorical but aren't.
Sequential. A qualitative palette here destroys the order.
Genuinely unordered and tracked across several charts?
The real case for categorical: EMEA must be the same blue on every chart in the dashboard so the eye can follow it.
Now a qualitative palette earns its place. Cap it at 5–6.
Gray is an instruction, not an absence. It actively says don't look here — it isn't what's left when you run out of colors. Default every series to gray, then pay one hue to the series you're actually discussing. Reach for a full palette only when a question above forces it.
Ordered continuous data. One hue, light→dark.
A meaningful zero or benchmark. Two hues, neutral middle.
Only after the gate above. Unordered, cross-chart identity, 5–6 max.
Here's the idea underneath everything else: color is a way of directing attention, not just a way of labelling things. These three follow from that, and they're the ones that tend to surface only after a few projects.
Viewers read the whole screen. A hue spent in one chart makes it less distinctive everywhere else. Blue used for a KPI, a category and a map ramp means nothing.
Do this: Decide your hues once, for the whole dashboard, before you build any single chart — this hue is always region, this one is always alert. In Power BI or Tableau that means setting a theme up front rather than picking colors visual by visual.
Red means something is wrong because it's rare. Three reds on a healthy dashboard and users learn to scan past it.
Do this: On a good day, a well-built dashboard should look almost entirely neutral. If someone calls it boring when nothing is broken, that's the system working — and it's worth saying so when you hand it over.
The darkest, most saturated color reads as most important — whatever the data says. Alphabetical legends hand that weight to whoever starts with A.
Do this: Assign color by importance or by data order. Try not to leave it to whatever order the tool happened to use.
Mood is not a hue lookup. These five dials do the work, roughly in order of impact — hue comes last.
Saturation
The biggest mood lever by far. Low reads expensive, calm, serious. High reads consumer, urgent, young.
Lightness spread
Wide spread reads analytical and legible. Narrow spread reads soft, tonal, atmospheric — and charts badly.
Neutral temperature
Warm grays read human and handmade. Cool grays read corporate and technical. Almost nobody touches this dial; it's strong.
Hue count
One reads disciplined. Four or more reads friendly — or chaotic, depending on whether lightness is matched.
Hue itself
Last, and mostly a matter of picking a convention your audience already holds. It carries less of the mood than you might expect — the dials above are doing most of the work.
Proof: every ramp below is the same blue. Only saturation and neutral temperature change — and they land in completely different registers.
“Luxury = brown” is really “luxury = low saturation” wearing a hue costume — brown is desaturated dark orange. Deep green or charcoal-and-bone would read just as premium.
Premium = subtraction, not addition.
HowSaturation 15–35% (HSL). Near-black such as #1a1a1a rather than pure #000. One accent hue, neutrals do the rest. Hairline gridlines, wide margins. No gradients or shadows — heavy ink reads cheap.
Competent and unexcitable. Safe, rarely memorable.
HowMid-to-deep blue anchor, saturation 40–60% (HSL). Cool grays for all secondary elements. Warm hues only as a diverging counterpoint.
High contrast, unexpected pairings. Fatigues fast on dense data.
HowSaturation 70–90% (HSL) on one or two elements only — energy comes from contrast against neutrals. Pair warm and cool at matched lightness.
Near-monochrome. Unhurried. Risky for dense data.
HowOne hue family, vary lightness only. Saturation under 30% (HSL). Off-white rather than pure white. Failure mode is insufficient contrast — check the contrast between your lightest and darkest steps (WCAG AA wants 4.5:1 for normal text).
Earthy and handmade rather than corporate.
HowShift every hue toward yellow/orange, including neutrals (warm grays). Saturation 35–55% (HSL). Terracotta against sage for diverging.
Deliberately uncomfortable. Reads as confident, but it won’t land with every audience — check the room first.
HowBreak one convention, not three. Full-strength hue against near-black or off-white, no midtones. A second bold hue cancels the first.
Multi-hue and friendly. Breaks when you need order.
How4–6 hues at matched lightness — that's what makes multi-hue read friendly not noisy. Build ramps from one hue kept outside the categorical set.
Engineering and observability register. Dark, because the tooling this audience lives in is dark.
HowCanvas #12151a–#1c2028, never pure black. Gridlines at 6–10% white opacity, axes 35–45%, labels around 70%. Saturation reads ~15–20% hotter on dark — pull every hue back. One exception: academic papers, print and LaTeX still assume a light background.
Viridis (the standard perceptually uniform ramp, built into Python, R and most tools) is designed for light backgrounds — its dark purple end, #440154, all but disappears on a near-black canvas. Lift the ramp or swap it.
Conservative, high-contrast. Red/green is entrenched — make it accessible, don't fight it.
In Greater China, red = gains, green = losses. Opposite of Western convention.
Moderate saturation, cool hues. This audience scans under stress.
Reserve saturated red for real clinical alerts only, or staff learn to ignore it.
Highest brand-color pressure. Use the brand hue as accent, not as the whole scale.
One brand hue rarely yields 5–8 distinct steps. Ask for their gray scale too.
Higher saturation tolerance. Keep alert color separate from brand color.
Keep seasonal campaign palettes out of permanent reporting.
Often a legal WCAG obligation. Colorblind-safe by default, not retrofit.
Red/blue two-category reads as partisan in US-facing work, whatever the data.
Green is overused to the point of greenwashing fatigue. Deploy it selectively.
For emissions, state green = less in the legend. Viewers assume green = more.
Serves donors and funders at once. Restrained data palette, warmth from imagery.
Funder packs get printed. Test grayscale and projection.
Premium comes from subtraction: fewer colors, lower saturation, more space.
Metallics have no flat hex. A muted ochre reads richer than literal gold.
Warm brand hue, but the data (load factor, delays) is operational.
Coral and red are near-identical when small. Pick a different alert hue.
Dark-first, so the palette inverts: brightest chip = high end.
Dark-tuned palettes wash out when exported to white PDF. Build both.
Technical, graphite-heavy. Color used for signal, not decoration.
Silver/graphite brand palettes are all mid-gray. Push lightness apart hard.
Warm and saturated by nature — which collides with alert colors.
Yellow blows out against white. Anchor the pale end with warm off-white.
Students read these as verdicts on themselves. Blue–orange beats red–green.
Minors and parents see these. Red 'failing' fills land harder than intended.
Deck-first. Optimize for projection and grayscale print, not screen.
Assume a board member prints it in B&W. Test grayscale separation.
Brand hue chosen for billboard recall, not chart legibility.
Magenta and red look nearly identical to people with protanopia (a common form of red-green color blindness). Use amber for alerts instead.
Red/green even more entrenched than finance — but desaturate on dark.
Flashing red/green cells are hostile to colorblind and motion-sensitive users.
The one category where saturated-on-dark is native, not a mistake.
Purple + blue is a near-invisible collision. Separate by lightness.
Choropleths — maps shaded by value — lean on color as the whole encoding, with no position cue to fall back on.
The basemap is part of your palette. Mute roads, water and terrain hard.
Entire brand palette lives in a narrow light band. Hardest to chart.
No dark values = no usable ramp. Negotiate one deep anchor tone early.
Plain-language versions of the terms used above. No shame in needing these — most of them are field jargon rather than anything fundamental.
Two quick checks before you deliver, both about a minute: paste a screenshot into a colorblindness simulator such as Coblis, and view the same screenshot in grayscale to see whether your categories still separate.
Where the ideas in this sheet come from. Nothing here is original research — it's a working synthesis, and these are the sources worth going to directly.
Brewer, C. A., & Harrower, M. (2002–). ColorBrewer 2.0: Color advice for cartography. The Pennsylvania State University.
Source of the sequential / diverging / qualitative taxonomy used in section 2.
https://colorbrewer2.orgHarrower, M., & Brewer, C. A. (2003). ColorBrewer.org: An online tool for selecting colour schemes for maps. The Cartographic Journal, 40(1), 27–37.
The peer-reviewed paper behind the tool.
van der Walt, S., & Smith, N. (2015). matplotlib colormaps.
Design write-up for viridis, magma, inferno and plasma, including the perceptual-delta plots that show why jet fails.
https://bids.github.io/colormap/Crameri, F. (2018–). Scientific colour maps.
Perceptually uniform, CVD-safe ramps with ports for most tools.
https://www.fabiocrameri.ch/colourmaps/Moreland, K. Color map advice for scientific visualization.
Per-map guidance plus downloadable colour tables.
https://www.kennethmoreland.com/color-advice/World Wide Web Consortium. (2025). Web Content Accessibility Guidelines (WCAG) 2.1 [W3C Recommendation].
Level AA requires a contrast ratio of at least 4.5:1 for normal text and 3:1 for large text.
https://www.w3.org/TR/WCAG21/Crameri, F., Shephard, G. E., & Heron, P. J. (2020). The misuse of colour in science communication. Nature Communications, 11, 5444.
The clearest modern case against rainbow and red–green maps, with practical alternatives.
https://doi.org/10.1038/s41467-020-19160-7Borland, D., & Taylor, R. M. (2007). Rainbow color map (still) considered harmful. IEEE Computer Graphics and Applications, 27(2), 14–17.
The paper that made anti-rainbow the default position.
https://doi.org/10.1109/MCG.2007.323435Ware, C., Stone, M., Szafir, D. A., & Rhyne, T.-M. (2023). Rainbow colormaps are not all bad. IEEE Computer Graphics and Applications, 43(3), 88–93.
The rebuttal. Rainbow maps test well for reading a value off a key; the critique holds for judging shape and gradient.
https://doi.org/10.1109/MCG.2023.3246111Gołębiowska, I., & Çöltekin, A. (2022). What's wrong with the rainbow? An interdisciplinary review of empirical evidence for and against the rainbow color scheme in visualizations. ISPRS Journal of Photogrammetry and Remote Sensing, 194, 195–208.
Review of evidence on both sides; concludes the answer is task-dependent.
Olson, J. M., & Brewer, C. A. (1997). An evaluation of color selections to accommodate map users with color-vision impairments. Annals of the Association of American Geographers, 87(1), 103–134.
Early empirical grounding for the CVD-safe palettes still in use.
https://doi.org/10.1111/0004-5608.00043Rogowitz, B. E., & Treinish, L. A. (1998). Data visualization: The end of the rainbow. IEEE Spectrum, 35(12), 52–59.
One of the earliest statements of the perceptual argument.
https://doi.org/10.1109/6.736450Knaflic, C. N. (2015). Storytelling with data: A data visualization guide for business professionals. Wiley.
The clearest treatment of the gray-plus-one-accent technique behind sections 1 and 3.
Ware, C. (2021). Information visualization: Perception for design (4th ed.). Morgan Kaufmann.
Standard reference for why lightness, not hue, carries perceptual separation.
Tufte, E. R. (1983). The visual display of quantitative information. Graphics Press.
Origin of the restraint argument and the data-ink ratio.
Stone, M. (2003). A field guide to digital color. A K Peters.
Colour theory written for screens rather than pigment.
Cairo, A. (2016). The truthful art: Data, charts, and maps for communication. New Riders.
On the ethics of encoding choices, including unintentional misleading.
Few, S. (2013). Information dashboard design: Displaying data for at-a-glance monitoring (2nd ed.). Analytics Press.
Dashboard-specific; source of much of the alert-scarcity thinking in section 3.
Brewer, C. A. (2015). Designing better maps: A guide for GIS users (2nd ed.). Esri Press.
Book-length version of the ColorBrewer reasoning.
Accessed September 2026. This sheet is a working synthesis, not original research — where a claim matters to a decision, go to the source rather than citing the sheet.
Hex values are approximations inspired by public brand identities, not official values — pull real codes from brand guidelines before client delivery. Brand names are references only; no trademarked artwork is reproduced.
