« Google UI changes? | Main | TVLand »

Visualizing Tufte

I spent the day yesterday in a ballroom listening to data visualization guru Edward Tufte. Given my increasingly hummingbird-like attention span, a full day of concentrated focus was as welcome as it was unusual. A good time, and well worth the money.

Perhaps the coolest things of the day were when he produced a copy of the first English translation of Euclid's Elements -- the book where he laid out the basics of geometry -- dating from 1582. And the other was his showing a first edition first printing of Galileo's 1610 observations of the moons of Jupiter and sunspots (and, oh by the way, the heliocentric model of the solar system).

But what I found most provocative...

But what I found most provocative in Tufte's talk was his rejection of the "rule of seven" in graphic design. You almost certainly know the rule: that people can't easily deal with series of things longer than seven: digits in a phone number, choices on a navigational menu, names of children (watch out, Kate Gosselin!). Tufte likes information density, says we deal with it every day, and insists that "we don't get dumber when we go to work." He derides focus groups, and his exemplar for successful density is NYTimes.com, with its 400-plus links.

Yet that flies in the face of any number of usability tests -- the one-on-one type; I agree that focus groups are useless -- that I've seen and run. Seven may be a limit on the low side, but having too many choices simply freaks people out and makes it impossible for them to see choices that are staring them in the face. What makes the Times's site work isn't so much the density as the intelligence of the clustering: This is always There, and That's always Down Over Here, and Everything makes Sense. It is that clustering coupled with a rigorously clean design (which is the other of Tufte's main threads) that makes the Times's site a success. The same thing is true of the National Gallery site that Tufte consulted on: many choices intelligently arranged.

And that's the other thing. Steal From The Best, Tufte says (along with many others). But he never quite got to the next step: test what you steal. The Times site has changed over the years. So has Apple's online store. The changes are because the Times and Apple never rest; they're always looking for a Better Way. The danger of stealing from the best is that "the best" keep changing. You'd better, too.

I was also struck by his saying repeatedly, "Do what it takes" to get your data to tell the story you need to tell. Data is good, and in most cases masses of data are better. If your visualization doesn't make your point, don't toss out data: rethink your graphic. Obviously, PowerPoint is going to be much too low-powered to handle anything interesting. But in example after example, the best visualizations become art. And art -- at least for me -- is hard. Visualization is not for sissies. It's not fast, it requires thought, there's no template and there's no standard tool. Sounds an awful lot like work. Boo hiss.

Not all of Tufte's examples, I think, were fair. He told one story of a colonel briefing (or trying to brief) the Joint Chiefs one Monday morning about Iraq. His data were a couple of weeks old and not particularly compelling, though he though he had done a fine job. But the brass asked him to compare his timeliness and thoroughness with that morning's coverage of the Washington Redskins in the WAPost. The post had play-by-play graphics, and statistics, and narrative -- an entire and exhaustive package generated, set in type, printed and distributed in the matter of a couple of hours. Why couldn't the colonel do something like that on a matter of national security?

My first reaction was that the colonel couldn't do it because the scope of the tasks being compared were not comparable. Football is a somewhat more limited conflict than war. The statistics are clearly proscribed -- not to mention provided by the team and the league. It's a wee tad harder covering a war than it is covering an NFL game.

But my second thought was, that's because I know how to cover an NFL game and not a war. It may well be entirely reasonable that the US Armed Forces has the data gathering ability and the established metrics to do this kind of analysis quickly and imaginatively. Upon further review, I rather hope they do. But I'm not sure that comparing it to an NFL graphics package is an especially valid point.

It was a long and interesting day, and I'm still processing it. More later.


This page contains a single entry from the blog posted on March 19, 2010 10:01 AM.

The previous post in this blog was Google UI changes?.

The next post in this blog is TVLand.

Many more can be found on the main index page or by looking through the archives.

Creative Commons License
This weblog is licensed under a Creative Commons License.