All good charts use this colour rule

there!

I'm back from Paris where I gave a little talk about dataviz in front of real people. That was awesome, even though I didn't get time to see the Eiffel Tower.

If you're looking to level up your team's dataviz skills, let me know! I gave 10 talks in 2024 and would love to do more in 2025.

Anyway. As always, my goal this week is to share a quick dataviz tip you can add to your toolkit.

The hardest thing: Colors

Last week I started a new section for my Matplotlib Journey project, diving into one of the hardest parts of chart design: colors.

Colors remind me a lot of CSS—they're highly susceptible to the Dunning-Kruger effect:

At first, it feels simple: just pick what looks good to your eyes and you're good to go!

But then you encounter concepts like color models, color blindness, contrast, luminance, saturation...

Suddenly, you realize there's a lot more to it.

The Challenge of Background Colors

For instance, how do you choose the right background color for a chart?

While researching, I came across an excellent article by Lisa Charlotte Muth, who analyzed hundreds of charts from leading organizations.

She found out that most of these charts follow the same pattern:

Basically, charts with:

• Light-mode backgrounds: use high luminance and low saturation.
• Dark-mode backgrounds: never use pure black but slightly softened tones instead.

Why is this the case?

Medium-bright backgrounds are rare because they're tricky to pair with effective colors. For good contrast, the colors on top need to be either very dark or very bright, which isn't always practical.

Similarly, highly saturated backgrounds clash with most color choices. Your background should support the data, not compete with it.

Key Takeaway

• Designing with colors is hard, and teaching it to programmers is even harder.
• For your next chart, use a background with high luminance and low saturation.

And if you're into learning more, I can't recommend Lisa Charlotte Muth's work enough. It's the best and believe me I read a lot about data visualization!

Have a great evening and let's make some great charts this week!
Yan