Colour is one of the fastest ways to direct attention in a chart. Before a reader has studied a label or compared two values, colour has already suggested what belongs together, what is important and what may require action.
That speed makes colour useful, but also easy to misuse. A poor palette can exaggerate a difference, hide a pattern, imply an order that does not exist or make the chart unreadable for part of its audience. Colour is therefore not a finishing touch. It is part of the chart's meaning.
Begin with the message
Do not start by asking which colours look attractive. Start by asking what the reader needs colour to do.
- Direct attention: Which result, exception or business unit matters most?
- Show magnitude: Do values move from low to high?
- Show direction: Are results above and below a target, or positive and negative?
- Separate groups: Are several categories different but not ordered?
- Signal a state: Does a colour mean warning, failure, success or unavailable?
Each question requires a different kind of palette. If the palette does not match the question, the chart can communicate the wrong relationship even when every value is correct.
How people perceive colour
We do not see colours as isolated measurements. We see them in relation to their surroundings.
A grey mark can appear lighter on a dark background and darker on a light background. A small yellow point can disappear on white even when the same yellow is obvious in a large block. Two colours that look distinct in a palette may become difficult to separate when used as thin lines.
Three characteristics matter most:
- Hue is the colour family: blue, red, green or yellow.
- Lightness is how light or dark the colour appears.
- Intensity is how vivid or muted it appears.
For ordered data, lightness is usually the strongest cue. Readers naturally interpret a steady movement from light to dark as a movement from less to more. Hue is particularly useful for separating categories. Intensity attracts attention, which makes a vivid colour effective for emphasis but disruptive when everything is vivid.
This leads to a simple principle: the strongest colour should usually belong to the most important information. When every series is bright, colour creates competition rather than clarity.
What is a colour space?
A colour space is a system for describing colours. It is similar to a coordinate system on a map: the same place can be described using different kinds of coordinates, and the choice of system affects how distances and directions are represented.
Screens normally create colour by mixing red, green and blue light. This system, called RGB, is excellent for telling a display what to show. It is less reliable for judging how different two colours will look to a person. Equal numerical steps in RGB do not always appear to be equal visual steps.
Other colour spaces were designed to describe colour in ways that more closely reflect human vision:
- HSL describes hue, saturation and lightness in terms that are relatively easy to discuss, but equal changes do not always look equal.
- CIELAB and LCH were designed so that numerical distance more closely resembles perceived difference.
- OKLab and OKLCH are newer approaches intended to behave more consistently for colour shown on modern screens.
The practical lesson is not that business readers need to calculate colour coordinates. It is that a palette can have the same starting and ending colours yet still contain very different colours between them.
For an ordered palette, those middle colours must progress evenly. If one step suddenly becomes much darker or more vivid, readers may see a boundary or jump that does not exist in the data.
The main palette families
A palette is a set of colours designed to work together for a particular analytical purpose. Most business charts need one of four families.
Highlight
A highlight palette uses one strong colour against neutral context. It is often the best choice for everyday business communication.
Use it when one result is the subject: the current period, the selected division, the exception requiring attention or the series discussed in the commentary. The neutral marks still provide evidence, but they do not compete with the message.
Sequential
A sequential palette represents values moving from low to high. It normally progresses from light to dark, sometimes within one hue and sometimes across related hues.
Use it for measures such as revenue density, utilisation, risk exposure or response time. The progression should be consistent: each step should look stronger than the one before it.
Use a smooth scale when small changes matter. Use a limited number of distinct bands when the business has meaningful thresholds such as low, medium and high. More bands do not automatically provide more insight; five clearly distinguishable levels are better than nine ambiguous ones.
Diverging
A diverging palette shows movement in two directions from a meaningful centre. Common examples include profit and loss, above and below target, or increase and decrease.
The midpoint must mean something. Zero, target and long-term average can all be legitimate centres. The middle should not be introduced merely because a two-sided palette looks balanced.
Both sides should have similar visual strength. A 10% shortfall and a 10% surplus should appear equally distant from the centre unless the business deliberately gives one direction greater importance.
Categorical
A categorical palette separates groups that have no natural order: regions, products, channels or customer segments.
The colours should be easy to distinguish and broadly similar in visual weight. One very dark or vivid colour will appear more important even if the categories are meant to be equal.
Keep the number of colour-coded categories small. Readers struggle to remember and match a long legend of similar colours. When categories multiply, direct labels, grouping or small multiples will usually communicate more clearly.
Status colours need restraint
Red, amber and green are familiar in business reporting, but familiarity does not make them sufficient.
Some readers cannot reliably distinguish red from green. Colour meanings also vary by context and culture. A status should therefore be reinforced with a label, icon, position or shape. Colour can make the state faster to notice, but it should not be the only way to understand it.
Reserve strong status colours for information that genuinely requires attention. If every satisfactory result is bright green, the dashboard becomes a field of decoration and green loses its ability to signal anything useful.
Accessibility is broader than colour blindness
Readers may have a colour-vision deficiency, reduced contrast sensitivity, low vision or age-related changes in perception. They may also be viewing the chart on a poor projector, in bright light, on a small phone or as a black-and-white printout.
A robust chart should still make sense when colour is weakened or removed.
Check whether:
- Adjacent colours remain distinguishable.
- Text and marks have enough contrast against the background.
- Thin lines and small points remain visible.
- Selected, muted and disabled states are still clear.
- Labels, position, shape or line style reinforce important distinctions.
- The chart remains understandable in greyscale.
Testing a palette for common colour-vision deficiencies is valuable, but no simulation can approve the whole chart. The final judgement must be made using the actual marks, labels, background and display size.
Brand palettes are not data palettes
Brand colours create recognition and personality. Data colours explain relationships. These purposes can support each other, but they are not interchangeable.
A brand colour can work well as the highlight in a chart. A collection of corporate colours does not automatically form a usable sequence from low to high, a balanced scale around zero or a set of equally visible categories.
The chart should still feel part of the organisation's visual identity, but analytical meaning comes first. Adapt the brand palette to the data relationship rather than forcing the data into every available brand colour.
Start with palettes that have been tested
Choosing a palette does not need to begin with a blank colour wheel. Established palette collections provide a much stronger starting point because they were designed and tested for analytical use.
ColorBrewer is the best-known example. It grew from the work of cartographer Cynthia Brewer and was developed with Mark Harrower at The Pennsylvania State University. It was created to help people choose colours for statistical maps, where poor colour choices can easily create false geographic patterns.
ColorBrewer organises palettes according to the relationship in the data:
- Sequential schemes for ordered values from low to high.
- Diverging schemes for values moving away from a meaningful midpoint.
- Qualitative schemes for distinct, unordered categories.
The number of colours is part of each scheme. A five-colour palette and a nine-colour palette are not merely shorter and longer versions of the same design; each set has been chosen so its steps remain useful together.
ColorBrewer also identifies schemes intended to remain usable for readers with common colour-vision deficiencies, in print and when photocopied. These filters do not remove the need to test the final chart, but they provide a more credible starting point than personal taste or a software default.
Although ColorBrewer began with maps, its central ideas apply to charts, tables and dashboards. A palette designed for large map regions may still need adjustment for thin lines or small points, where contrast is harder to see.
Test colour in the finished chart
Large swatches make almost every palette look convincing. The real test is whether the colours work in the form the reader will encounter.
Apply them to the actual bars, lines, points, regions and labels. Review the chart at its final size, on its final background and in its likely viewing conditions. Ask someone unfamiliar with the design to identify the important result and explain what the colours mean.
A practical process is:
- Define what the reader needs to notice or understand.
- Choose the palette family that matches that relationship.
- Use only as many colours or steps as the reader can distinguish.
- Begin with a tested palette where possible.
- Apply it to the real chart rather than judging swatches alone.
- Check contrast, colour-vision accessibility and non-colour cues.
- Test whether a reader can explain the colour meaning within seconds.
Good colour design often looks restrained. That is not a lack of creativity. It is evidence that colour is doing a specific analytical job and then getting out of the reader's way.
Further reading
- Cynthia Brewer and Mark Harrower, ColorBrewer, for tested sequential, diverging and qualitative schemes.
- Cynthia Brewer, Designing Better Maps: A Guide for GIS Users.
- Lisa Charlotte Muth, Your friendly guide to colours in data visualisation.
- Maureen Stone, A Field Guide to Digital Color.
- Colin Ware, Information Visualization: Perception for Design.