I had a realization while reading Stephanie Evergreen’s dissertation yesterday (yes, really, Stephanie, I read the whole thing).
When I put together what seems to me a high-quality data visualization, I often get feedback that it’s “too simple.” Stephanie’s dissertation reminded me that contemporary visual processing theory has found that we can only really hold around 4 bits of information in our working memory at any time.
What this means for data visualization is that attempting to pack in too much into a single graphic is actually counterproductive, as it will almost certainly overwhelm people. Most high-quality data visualization is, for this reason, intentionally simple, with a focus on one finding. See, for example, this recent example from a New York Times op-ed about the growth in inequality. The chart is extremely simple. And extremely powerful.
I’ve begun to think of each data visualization more like a sentence, rather than a paragraph. Overwhelm a sentence with too many clauses and the reader will be confused. Pare it down to one main point and you’ll communicate effectively. Similarly, with data visualization, if you have multiple points you want to make, don’t combine them into one. Instead, make several, one for each point.
For example, while working on a project about Outdoor School in Oregon, I debated about how best to demonstrate one of my main findings: that most camps are located in Northwestern and Central Oregon, and as a result schools in other parts of the state are less likely to participate.
I began by creating a map showing both the camps (in green and with their size proportional to the number of schools that did Outdoor School at each of them) and the schools (blue dots for those that did Outdoor School and red dots for those that did not).
The result was a bit of a mess. It’s hard to see all of the dots. Some of the green dots are so small that they are almost invisible next to the red and blue dots. And, worst of all, the message I want to convey is muddied by the overwhelming mass of information.
What I did instead was break this into multiple maps. I made one map for the the camps, highlighting the finding that most are in Northwestern and Central Oregon.
Then, I produced maps for the schools, one showing schools that participated and one showing schools that did not.
Each map is extremely clear. In the text, then, I draw the connections between the multiple maps, highlighting the main finding:
Very few Outdoor School programs take place in Southern and Eastern parts of Oregon. This presents a challenge for schools located in these areas. If they want to participate in Outdoor School, they often have to travel many hours to reach sites.
Had I attempted to do this in a single map, my readers would have been confused. By incorporating visual processing theory, and understanding that it is best to present data visualization as bite-sized chunks of information, I was able to present a clear finding.
So, yes, my data visualization may be simple. But that’s exactly how I want it to be.