Monday, October 15, 2012

Visualizing political shifts: Data and interaction design

The New York Times has just published a very nice visualization (or "Sankey diagram"; thanks for the reminder, Len DeGroot) about how state-by-state voting has evolved since 1952. It was designed by Shan Carter and Mike Bostock, two visual journalists I'd recommend you to follow closely. When I first saw the graphic —half an hour ago, and really quickly— I was enthralled. I wrote so on Twitter.

A few minutes later, Chrys Wu, someone whose opinions I deeply respect, replied with a sound objection: "It's very nice, but also a little awkward. Once I find the state I'm looking for, I lose it as I scroll down the page." That is a problem indeed, one that I hadn't noticed in the quick preliminary scan that I had made.

So I went back to the graphic and tried to use it, not just to enjoy it aesthetically. Chrys' note is spot-on: Why don't the designers let me click on one state to keep it active while I scroll? I would add something else: Why don't they let me highlight more than one state simultaneously? What about if I want to compare Florida, where I live, to North Carolina, where I used to live years ago? You can never predict everything your readers will try to extract from your visualizations, so the more reasonable options you give them to rank, compare, and correlate, the better.

Finally, why not adding a search box to this graphic —and to many other complex NYT's data graphics, for that matter? As much as I like Shan's and Mike's latest project (it's already one of my favorites this year), I believe that some tweaking based on very basic interaction design principles can greatly improve it.

(Update October 28th: Some of the issues discussed in this post seem to have been addressed in a recent update of the graphic)