This too much discussed paper “found that people’s accuracy in describing the embellished charts was no worse than for plain charts, and that their recall after a two-to-three week gap was significantly better. In addition, participants preferred the embellished charts“.
OK, let’s take a deep breath. Apparently, all things being equal, you should use a junk-ridden chart because people actually remember it (and like it, too).
I don’t think so. Actually, making a memorable chart is a misplaced obsession in many cases.
Think of addiction, for a moment. When you get addicted to something, you just want more and more of it. And while a small dose was enough, now you need a larger dose to get the same results. And when you over stimulate your senses, getting back to square one is often hard or even impossible.
The same happens with the so-called “high impact communication“. That’s basically people yelling at you (or you yelling at someone else). IT MAY WORK ONCE OR TWICE, BUT WHEN EVERYONE STARTS YELLING AT EVERYONE IT JUST CREATES A STRESSFUL ENVIRONMENT. THEN YOU NEED TO UPGRADE TO AN EVEN HIGHER-IMPACT COMMUNICATION (YELL LOUDER) AND THE VICIOUS CIRCLE NEVER ENDS. You see? You can relax now.
If you start making junk-ridden charts just because the long-term recall is better then I’m afraid you are missing the point of what data visualization is about. In a corporate environment, where you have to use dozens of charts every single day, you just can’t make them all stand out (it’s like the absurd idea of exploding all the slices in a pie chart). People just need the god damn chart to learn something. They don’t want to remember it three weeks later.
You must be careful about what you want to emphasize, and the series (or even the data point) is the right level to do it. A simple model is to mute grid lines, use gray or pale colors for contextual data and a primary color to force a series to stand out.
Attention span is a finite resource, and you should use it frugally. An “embellished” chart has a higher impact but depletes that resource at a higher rate than a “Tufte-compliant” chart.
The Sad True About This Paper
Although the authors “are cautious about recommending that all charts be produced in this style” (how nice of them) some people will read the paper as a license to kill every decent chart on the planet. These findings may apply to infographics and charts in some magazines, but that’s where users in a corporate environment look for “inspiration” to avoid making boring charts.
If taken seriously, this paper would be a disservice for the data visualization community. Let’s laugh, then.