Ten rules, in short: better figures

In 2014, Rougier, Droettboom and Bourne published "Ten simple rules for better figures". Here are the ideas that matter most.

Know your message and your audience

Before plotting, decide what one thing the figure should show. A figure for a paper (read closely) can be denser than one for a talk (seen for seconds).

Don't trust the defaults

Plotting software defaults are designed to work for everything, so they're perfect for nothing. Adjust font sizes (readable when printed), remove unnecessary gridlines and borders, and label axes with units.

Use colour carefully

• Use colour to mean something, not to decorate.
• About 1 in 12 men and 1 in 200 women of northern European ancestry have some colour-vision deficiency, most often red–green. Avoid red–green pairs as the only difference.
• Avoid the rainbow ("jet") colour map: its uneven brightness creates false boundaries in the data. Crameri and colleagues (2020) showed how it distorts and excludes readers. Use perceptually uniform maps such as viridis or cividis.

Show the data

Where possible, show individual data points (or their distribution), not just bars of averages. Bars hide spread, outliers and sample size.

Avoid chartjunk

3D effects, shadows and heavy decoration make figures harder to read. If an element doesn't carry information, remove it.

Check it in black and white, and small

Print it in greyscale and shrink it to column width. If it still makes its point, it works.

RememberOne message per figure, readable labels with units, meaningful colour that's colour-blind safe, and the data itself shown.