If you know anything about the history of data visualization, Edward Tufte is a name you'll readily recognize. If he's your source of inspiration, then you ought to be in good hands.
A statistician and Yale University professor emeritus of political science, statistics, and computer science, Tufte is known as a supreme master of the art (and science) of delivering information visually. He's inspired many of today's data visualizers, among them Jonathan Schwabish, a US economist who says he considers Tufte the "godfather of modern data visualizations." For him, a one-day Tufte workshop turned into an eye-opening experience on how "we can use graphics and data visualization in a more strategic way and a way in which we can show our data in clearer and more innovative ways."
Since then, Schwabish has been putting what he's learned into his work, creating data visualizations in his role as a policy analyst. "I put those things into practice, and keep thinking about ways and better ways in which we can present our data," says Schwabish, who now runs his own workshops about visualizing and presenting data for people in public policy.
Jonathan Schwabish, US policy analyst
Schwabish joined us yesterday to present his data visualization dos and don'ts during our first live Facebook video chat, which you can view on demand here on AllAnalytics.com. I've included four takeaways below. Watch the video for the full rundown!
Know the differences between exploratory and explanatory data visualizations. Interactive data visualizations, those available through a web interface, for example, can be exploratory or explanatory in nature. Exploratory interactive visualizations encourage users to go into the data and play around with it, maybe even coming up with their own conclusions from it, Schwabish said. In explanatory visualizations, the data visualization "tells a story and the interactivity is leading you down that path."
Static visualizations are explanatory in purpose, too, since the user doesn't have the capability of working with the data or playing with the graphics. Static data visualizations include your basic bar charts and pie charts, as well as those towering infographics -- compilations of text, graphics, and images that have become so popular today. (View our own latest such data visualization, 3 Levels of Analytical Sophistication, and see one of Schwabish's below.)
Devote time to plotting out the infographic. Don't give short shrift to the amount of time needed to think about how to present and tell the story of your data, said Schwabish, adding that he spends about 70 to 80 percent of his time laying out his story for the larger infographics he creates. "I'm in an analog world here, actually sketching with pen and paper and colored pencils." Then you can move into the graphics software.
Understand what your audience wants. While you might gravitate toward wanting to create the fun and fancy type of data visualization, don't do it if it serves no purpose for your audience. For example, Schwabish mostly prepares static data visualizations because he's trying to provide members of Congress and their staffs the bottom-line, statistic, or headline piece of information. "At this point, I'm not sure my audience is really interested in an interactive infographic where they have to weave and explore and click -- that's not my audience." The standard static data visualization serves as much purpose as the interactive one for the right audience.
Deliver insight. Regardless of type, data visualizations ought to give users fresh insight. "If you can give your users insight that they may not have gotten from some other means, be it the written report, some other website, or some other source, that, I think, is a successful visualization."
I would agree. How about you?
All Analytics will be continuing our ongoing series of video chats on data visualization next week, when business intelligence consultants Tricia Aanderud and Ben Zenick join us for a conversation on how to create great data visualizations. You'll find us on Facebook next Wednesday, June 5, at 2:00 p.m. ET. I hope you can tune in!
I won't argue against imagination for interpretation but I think if you get too imaginative you lose the audience. There is a real skill/talent needed to make numbers look pretty and keep them from getting confusing.
@Beth, to make matters more complicated continuing with your writing analogy, I think there are times -- particularly longer research papers type projects -- when even someone who doesn't typically use outlines might need to use an outline to get their heads around a particularly complex topic. Same might go for visualizations.
I'll add imagination for interpretation, which supports the "artistic eye". See how a graph can be shown with respect to code can be tricky, but imagination has to be at work to get it down to an appreciable level.
@SaneIT, I would suspect it's a lot like the writing process. Some people swear by the outline; can't write cohesively without it. Others don't like the outlining process or the structure associated with it. I think forcing one way or the other on a person results in poorer quality than would result by letting the person do what works best for him or her.
I think if you've got the talent to sit down and organize everything by hand and draw it out chances are that you've got a better eye for aesthetics than I've got and that does go a long way in making visual representations that people will actually look at. I do think that it's a skill but it involves more than just interpreting data, if you have that artistic eye I'm sure it helps a lot.
It'd be interesting to compare data visualizations -- the "tell a story" sorts with graphics, text, and images, at least -- done each way. I would think depending on the complexity of the topic, how much information you're trying to convey, and the audience, storyboarding could really help.
""I'm in an analog world here, actually sketching with pen and paper and colored pencils." Then you can move into the graphics software."
This really surprised me. I figured most of the people building these were using visualization packages that generate the charts for them. I never imagined someone sitting there drawing out the charts by hand and putting colors and labels on that way. I know I just fire up the software pull in the data I want and do all the pretty changes inside the software.
I for one love the update Linked In creatd - it permits additional material to describe the profile. Given the importance of data visualization, it can give an opportunity to see how far a report with a visualization can go.