I'm not so sure about Tufte

Kandinsky

A couple years ago, I was fortunate enough to be able to catch Edward Tufte, the father of the current wave of thinking about data visualization, at a seminar he was giving in Philadelphia.

I, like many people at the seminar, were really excited by the examples of visual presentation based on Tufte’s books, as well as his declaration that PowerPoint is evil.

But, looking back on the seminar, I can’t say that I came away with anything practical to take back to work. I was, at the time, specifically looking for very concrete guidelines for how to un-uglify Excel graphs for presentation to executives. Today, I’m very interested in how to make R output look good.

Tufte does have some general bullets: + no clutter (aka “chartjunk”) + clear labels + the data-ink ratio and + sparklines

but I was looking for something more specific, with examples in spreadsheets or code. The closest I’ve come so far is the stuff on Juice Analytics’ website. I’ve also come up with a couple principles on my own, garnered from personal experience and picked up from colleagues.

A couple of rules for clear visual presentation:

  • If you’re making a PowerPoint, make it easy to email to other people and make the context easily discernible from the slides. If you do good work, it will get around the organization.
  • Include sources for all of your images in each image and slide, including when you pulled the data.
  • Don’t use red and green. Use neutral tones like blue, beige, and light green together. Use ColourLovers if you need hints.
  • Make it easy to reproduce (either have some link to your code or make it easy to look under the hood.)
  • NEVER use Excel default charts. Take out the extra grid lines. Make the font larger, more legible, and make it the same across all your presentations.
Written on May 31, 2015