# How many different colours can I have in a data visualisation?

I've designed a (6x6) matrix of similarity values, where there are 10 different values. I want to assign color to the cells that the values are representing, where color signifies a certain value.

Is there a limit on how many colours I can use in a visualisation? I'm trying to find related papers or documents that argue for or against using more than 6 colours (6 is an example).

• Are you are mapping a continuous or ordinal range to a color scale? If continuous, do you need users to be able to specify absolutely the exact quantized value represented by a given color? – duhaime Sep 7 '18 at 16:07

This may not be the kind of answer you look for, but this is an answer from my experience.

It is hard to pinpoint the exact usable number exactly, but for example, common data visualization tool Tableau uses 10 to 20 distinct colors. More exactly, it has a base palette of ten visually well discernable colors and then ten lighter tones of the same colors, achieving sufficient readability.

There is even a blog describing their design process (though it might be more of a marketing piece).

The catch is, there are other limits. While those 20 colors are visually distinct, my feeble human brain cannot hold 20 concepts in short-term memory. So even if I can see that two values are different (differently colored), I may not be able to (quickly, without looking them up in a legend) come up with what those colors actually represent behind them.

The limit for that is for most humans somewhere between 5 to 10. I have no citation handy for this, it's just that common knowledge thing, that everybody repeats (the common number is often seven for this). Obviously, some people are really good at this (like people who are able to count cards) and some aren't.

The thing to take away from this is, that since you usually design things for "average" people, you need to try to be as conservative as the subject matter allows. And then you have colorblindness, which might be a good thing to look out for – the more kinds of colorblindness you wish to design for, the fewer colors you will be able to use.

Practical sidenote:

If ten or twenty colors is not enough, you can adopt other forms of distinctions, such as:

• Shape
• Outline color
• Texture (hatching, cross-hatching)
• Striped versions (e.g. white and blue and so on)

Giving you almost unlimited possibilities. Keep in mind that the human mind will not work as well with higher numbers of elements, colors and so while you can use these, it doesn't always mean you should.

• @user10230515 Thank you for accepting my answer, appreciated. I'm not sure what the customs of this stack are, but it might be better for you to postpone accepting until later (perhaps a day), so that others are not discouraged from attempting to answer. – J.E Sep 7 '18 at 11:33
• Excellent answer! I'd only add that the number of colors you permit yourself may depend on the intended message. If I'm expected to translate colors into values and absorb them, then fewer is better. If you're trying to present a range of data and don't know which will pique my interest, more colors will be ok. I can check the legend, work out that e.g. light blue represents the specific segment I'm interested in and can focus on that, regardless of how many other colors are shown. – Steve Rindsberg Sep 7 '18 at 15:33

Following the perfect answer of @J.E, my recommendation is not to get block with the colors, here some proposal, two cited in the answer and my addition is the volume.

Texture

Shape

Volume

• Nice viz! And this nicely illustrates why on should hesitate before using things like these, since they can get very cluttered, real fast. Especially with the shapes, my eyes just dance across it without getting much information in, as it it was some optical illusion. – J.E Sep 7 '18 at 11:30
• It is obvious I'm not a professional in this field, I'm just a graphic designer ✍🏻😎 I just wanted to show other options beyond the color that in my opinion is the most immediate. – Danielillo Sep 7 '18 at 11:37