Playing with the Shape of Data

Are you interested in data visualization, presentation and graphical representations?

Me too!

I spent this morning playing around with various visual forms, to see how well they convey information. You might enjoy my mini-study, and like to share your own opinions on the matter!

Circles are Cool, But Circles are Bad

I was reading one of my favorite magazines: Technology Review, and stumbled upon this graph:

The circles are soothing. The orange is bright and happy. There will clearly be a lot of growth between last year and 2015. But do you get a sense for how much growth there will be?

Now look at this bar graph:

It feels quite a bit different than the two circles, doesn’t it? I was actually shocked by this and checked my math quite a few times!

It turns out that 277 is actually 4.62 times 60. To me, those circles do not immediately impart the feeling of “Four Times More”, but the bars do. Well OK, I don’t actually look at the bars and think “Four Times More”, but I feel like I truly understand the difference between the numbers when I look a the bars, but not when I look at the circles. Do you see it the same way?

Playful Study

I’ve read many articles about data visualization which  mention that human beings are not good at judging areas. While there has been quite a bit of rigorous study into the matter, I thought; “No better way to learn than to try it yourself!” So I fired-up my copy of Visio 2010 and created the visual bits you see above and below.

What follows is a playful study–my thoughts and ideas–about the problem of visually comparing numbers. Hopefully you’ll find it interesting and you’ll want to share your own opinions in the comments at the end of the post.

Unit Circles

Since circles are cool, I thought I’d experiment with them a bit more. Here’s another variation of the original graphic:

click to view larger image

You can clearly see that 2015 will yield more than four times as much as 2009. But to me, it almost looks like 5. At first glance–without thinking or analyzing–it looks like 4.8 or 4.9 to me.

Since the whole point of data visualization is to relay accurate info with little effort on the reader’s part, I don’t like graphics that force me to figure them out. I think the stacked circles would probably work best for comparisons of round numbers. The fractional circles don’t work well.

Bar Fatness

Back to bars. What about the thicknesses? Does the fatness of the bars also influence your perception?

click to view larger image

To me, the fatter bars seem to de-emphasize the difference, but I’m not really sure. Perhaps it is that bad-at-area-estimation human trait at work again? What do you think?

What about the labels? The left-most, skinny bars have vertical labels and vertical text. Does that read your eye astray?

Height vs. Length

What if we change the orientation of the bars as well as the thickness?

click to view larger image

If you’re reading this article, you’re probably used to reading left-to-right. Maybe horizontal bars are more comfortable to look at. But I feel like it is more natural to compare vertical heights than horizontal lengths. That might just be me.

I recall reading about perceptual differences between length and height. I don’t want to go and look for that information now, but if you want to share a link, please post it in the comments below. Right now, I find the horizontal bars comfortable to look at, but I feel like the vertical bars give a better sense of the numerical comparison. Am I being influenced by mis-perceptions and human weakness?

Again, I think the fat horizontal bars de-emphasize the difference in the quantities. The thin bars are clearer to me.

I also wonder if the spacing between the bars has an effect on perception?

Single Columns

The next variation puts both quantities in the same bar, so there’s no worry about fatness or spacing:

I think the only danger here is that it might not be clear that the “277 billion” bar also starts from the ground. I tried to make the black line at the bottom a bit thicker and more definite to give a feeling of Earth or Ground, but I’m not sure if it is clear. This is a problem with stack graphs that I have seen. Are the different colors all starting from zero, or are they piled on top of each other?

Overlay vs. Stacking

Here we can see the difference between both bars starting at zero, and when they are stacked on top of each other:

Another variation on the single column approach adds subtle “unit lines” to help in gauging multiples of one quantity over the other:

Again, we have the are they stacked or do they both start from zero problem.

Offsets and Unit Ticks

To alleviate that, we can offset the bars slightly, without making the graphic too visually distracting:

I like this version. It’s clean, easy to understand, and has some extra bits of info from the unit tick marks. You can easily compare the values, and people who want to be able to say; “Oh, 2015 will be 4.6-ish more than 2009!” get that extra bit of help, without too much distraction.

What do you think about all of this? Share your opinions and ideas in the comments section below. And if you have any interesting links to other articles and research on related topics, we’d love to have them!