I’m always on the lookout for fresh, new examples of data visualization. It’s best to learn from other folks who are doing a good job.

So I ran across this and it struck me as a neat mashup of available data which was originally presented in a numbers format, but has been re-presented in a visual format to draw out (visually) some of the overall trends and patterns.

Phillip Kromer took newspaper endorsement data from the Editor and Publisher pages, and mapped the results. It’s an interactive map trying to understand the oft-heard accusation of left-leaning bias in the media. I’m not looking at this from a political slant, but interested in the data visualization and what I can learn from it as a designer.

My first impression was: “Cool!” But that quickly devolved into, “This is confusing. What’s going on?” It looks neat, but the context wasn’t clear to me until I paused to understand the legend. Even then I found myself reading the thing multiple times, checking the visual, and trying to understand what was being portrayed so I could start drawing conclusions.

One point of confusion, unfortunately, are the many layers of data on this visualization, but that their sum complexity rather than providing a keener insight into the problem, only muddies things up for the viewer who’s trying to sort through everything. Here’s what I mean:

• Layer 1: Red state and Blue state. Although it’s a bit simplistic of a model, it’s what many of us are used to. New York is a blue-leaning state, as is California. Texas, on the other hand, is a red-leaning state.
• Layer 2: Circles represent media outlets and their circulation size. Big circle means big circulation size, and small circle means small circulation size. No problem so far.
• Layer 3: Here’s where we get confusing, and if you’re not careful, it can be a doozy. Color of circle indicates a media outlet’s political endorsement: Red-filled circle = McCain, and blue-filled circle = Obama. No sweat.
• Layer 4: Exterior color of the circle (or lack of) indicates which candidate the media outlet endorsed in the 2004 election.

Now, I don’t know about you, but I think Layer 3 is fine. It makes sense. Even Layer 4 (conceptually) is okay… but the legend is absolutely confusing:

It’s confusing because Kerry (’04) is in Red and Bush (’04) is in Blue. Huh? Why do that? That breaks the mental category which which we’re familiar. We’re used to working that way, why upend it and introduce a legend that breaks our familiar categories? Sure, the legend makes sense in an of itself. But because of a viewer’s pre-cognitive categories, it can really leave a viewer in a lurch who’s trying to interpret this visualization.

That’s something we definitely want to avoid as we try to build a usable product. How do people think? What are they bringing to the table even before they reach this feature or task we’re building, and how do we account for that so the UI is consistent (or, even better, leverages) what they already know and are famliar with?