There's value in trying your [dataviz] worst

In the past I’ve always asked students to create their best possible graphs in ggplot2 to practice creating clear, engaging data visualizations. Recently, I’ve realized value in adding a few early exercises that encourage students to make their worst.

Why is it good to make ggplot2 graphs so (intentionally) bad?

Here are four ways that creating a purposefully disgusting graph promotes learning and exploration in ggplot2:

  1. Making the worst graph encourages creativity while eliminating the stress of producing a perfect final visualization (unless, of course, it’s perfectly hideous)
  2. Making the worst graph requires extensive customization and code
  3. Making the worst graph means critically thinking about and identifying contributors to “bad” graphs
  4. Making the worst graph is surprisingly fun

I was recently inspired to try my absolute dataviz worst by a #tidytuesday prompt using data from Sarah Leo’s efforts to improve imperfect graphs in The Economist.

Instead of trying to recreate or further improve on their graphs, I decided to use my powers and energy for a much different purpose. A darker and more anarchistic purpose…

Behold, my worst!

The undiscerning eye might think “these graphs are abominations and there was clearly no effort put into them.”


It’s a long dive from ggplot2 defaults into the depths of truly terrible DataViz, and it takes creativity and effort to keep sinking.

Here are just a few things I learned or re-learned during my nosedive:

  • Arrange multiple graphs with cowplot by Claus O. Wilke
  • Add figure labels and subtext
  • Expand figure margins to let axis labels breathe
  • Update fonts and colors in all graph components
  • Convert to polar coordinates
  • Customizing date formatting on axes

(click here for my complete worst dataviz code)

Through the process of trying my worst I learned some new customization skills, created useful reference code for future ggplot2 efforts, and had fun – it felt oddly liberating to be deliberately and creatively awful.

From now on, I’ll be adding a few early ggplot2 activities requiring students to do their absolute dataviz worst. There’s plenty of room at the bottom, and we can learn a lot and have fun on our way down.

Data visualization resources (in the other direction):

Allison Horst

My teaching interests are data science, statistics, and science communication.