This is the first of a series on visual analyses (i.e charts) out in the wild. The commentary is based on how well the chart in question does its job of conveying information. Charts that offend will be cited by the Viz Police.

Today’s chart is from USA Today, on the subject of holiday greeting cards.

Bad Cop: Visual Analysis Conveys Almost No Data

This chart is unspeakably bad. It manages to create an entire graph from two points of data and a lot of chartjunk, sending the data-to-ink ratio to just about zero.

With only two data points, you’d think it would be hard to bastardize the data. But you’d be wrong. Here USA Today finds it relevant to compare two different measures: how many people actually sent cards last year, versus how many intend to send them this year. Somehow, in this tiny graph, USA Today managed to open up the thorny issue of human intentions versus actions. Maybe in the new year they’ll give us a chart of the number of people who lost weight in 2008 versus those who intend to lose weight in 2009. Somehow, I think the 2009 number will be larger.

Good Cop: Cute Santas

The way the two extraneous Santas are looking at each other is kind of funny and cute. The placement of the two data labels in the top of their hats is clever. That’s about all the good I can find in this chart.

Ellie Fields is Director of Product Marketing at Tableau Software. She did not send holiday cards last year, but does intend to send them this year.


Even more egregious from a visual standpoint is that it's impossible to visually make any meaningful comparison of the size of the cards due to the diagonal edges, dissimilar baseline, etc. It looks a bit like the 77% card is farther away, further compounding the perceptive problems. It's a mess.

A few more problems I have with this viz:

* The diagonal opening of the front of the card sets up a weird perspective. Comparing the right side of the left card and the left side of the right card exaggerates the difference between the two.

* The caption refers to the number of people, whereas the data shown refers to percentages. Besides being different units, it doesn't state what the percentage is of (shutterfly customers? total population?) -- and whatever that denominator is, the data shown doesn't account for changes in that number.

* Are those really Santas? I initially thought they were elves with Santa hats. Too much ambiguity in their iconic Christmas characters. But yeah, they're kinda cute.

On the plus side, they very cleverly kept the faces the same, but scaled the hats to fit the cards.

If you plan on using USA Today's graphics, then this will likely be a daily column.

I think we should have a way to provide you with graphics we would LIKE to see replicated in Tableau. When I see good examples, I immediately try to think of ways to get Tableau to present the data in the same fashion... whaddya think?

@Raif: Good catches. There are indeed additional problems in the use of perspective, units and holiday personas. USA Today provided us with a holiday cornucopia of flaws.

@Dan, great idea. If you want to nominate a chart for the Viz Police, feeel free to send it to efields at tableausoftware dot com, or just comment here with the URL. Replicating a graph using best practices, kind of an Extreme Viz Makeover, would be fun too. That's a little harder because we'd need a sample data set, but when one is available and we've got the time, I think we'd go for it. Or if you've got one you want to see on the blog, send it in

In particular, I like the concept of the scatterplots in the attached image. They make the most recent dot red, the last week or so black, and then all older points grey.

Can this easily be replicated? Can such functionality be a standard feature?

Dan, DSA Insights did just that- 11 awful charts from USA Today. You're right, it could be a daily feature.

Come on, it's USA Today, home of McNews and goofy graphics. You've set your expectations at a completely wrong level.

@Dan- yeah, I like that too. Ty Alevizos, one of our jedi masters, sent me a workbook where he's done the same by using the "Edit Colors" feature in Tableau. It's attached- write back here if you have any questions.

Yes, that's neat.

It'd be great to make applying this methodology a bit easier thru a wizard of some sort, so the average "business user" could easily apply that to uncover insights.