Picking good colours for your visualisations is one of the more difficult tasks in data visualisation (see the Colour: the basics and the Colour: use in data visualisation modules of the “Design principles for data visualisation” training for an extensive introduction to the topic). This module discusses some of the common pitfalls in the use of colours.

Go beyond the defaults

The first and easiest pitfall to fall into when it comes to colours, is to just use the default colours the tool you are using to make your visualisation.

The default colour palette in a version of Microsoft Excel. Source: Maarten Lambrechts, CC BY SA 4.0

There are several reasons to run the extra mile of picking your own custom colours:

• the tool you are using does not know what the message is you are trying to communicate with the visualisation. Chances are that some data points in the visualisation are more important than others, so it makes sense to use colours to highlight the more important data points
• your tool does not “understand” the data you are visualising. It might assign green to “bad” values and red to “good” values, or yellow to the “oranges” category and orange to the “bananas” category, for example
• colours are an opportunity to brand a visualisation and really make it your own. Default colour palettes have a very generic feel to them, and miss identity and personality
• when the default colour palettes in your tool are not thought through very well by its developers, the colours might be inaccessible to people, or can even be misleading (see further)

Accessible colours

A significant proportion of the human population is not able to distinguish between certain combination of colours (8 percent of European men cannot distinguish greens from reds well, for example).

When your visualisation relies heavily on colour to be well understood, you need to make sure the colours you use are accessible to people suffering from colour blindness.

A chart relying on colour to identify the lines (above) and the same chart as seen by people suffering from deuteranomaly (unable to perceive green, below). Source: Maarten Lambrechts, CC BY SA 4.0

The solution is to always test your visualisation to be colourblind proof, and to avoid some colour combinations that do not work for people suffering from certain forms of colour blindness.

Colour combinations to avoid are

• reds, greens and browns (see the charts above). This implies that using the traffic-light colour palette to signal “good” or “safe” data points with a green colour, and “bad” and “dangerous” data points with red, is not a good idea
• pink, turquoise and greys
• purples and blues

ColorBrewer includes many colourblind proof palettes (make sure to check the “colorblind safe” option). A good tool to check the palette you use in your visualisation is Viz Palette.