Data Viz Done Right

September 7, 2009

Bubbles Bubbles Everywhere

The New York Times ran an article written by A.O. Scott back in November. The purpose is not to critique the article, but rather the, gasp, bubble chart used to rank media consumption hours.

I'm a big fan of Stephen Few and have learned a lot from Stephen and his books about effective visual design. Stephen point out that "Visual perception in humans has not evolved to support the comparison of 2-D areas, except as rough approximations that are far from accurate."

As soon as I saw this ranked bubble chart, I immediately began exploring other, more effective display mediums. Here are some examples.

I wanted to start by trying to find a way to use the bubble charts. The only method I could employ was to add color to the bubble charts, but I don't gain much at all.

Of course, the simplest way to rank data is through a simple bar chart. The first example is as intuitive as it gets; it's very easy to compare the relative size of the bars. The only purpose of this graph is to emphasize the rank.

I took this a step further. When reviewing Scott's bubble chart, I had the impression that he was emphasizing the percentage of time that we spend in each of the different medium. That led me to a bar chart that shows the contribution to the total. It's the same graphic as the ranking chart above, but this time I intentionally labelled the bars to emphasize the contribution of each activity.

I'll conclude with one of the least effective displays, the dreaded pie chart, but I think one of the pie charts is actually a bit effective. The first pie chart displays every category, which makes it impossible to compare the sizes and has way too much information.

I decided to group all but the top two categories into an "other" category to simplify the pie chart and I also ensured that they were ranked by contribution as you made your way around the pie.

Which display do you like best? Which display is most effective? My vote is for the bar chart displaying the contribution to the total.


  1. I prefer the bar chart that shows the hours of consumption. That gives a better sense of how much time is being spent overall.

  2. Another variation would be to use the bar chart with hours of consumption as the X-axis measure, but label each bar with the percent of total value.

    Peace and All Good!
    Michael W Cristiani

  3. Nice looking blog Andy. This is a good post, showing the basic problems with pie charts.

    This is probably the single most annoying aspect of poorly done data visualization. Nice illustration.

  4. Thanks for the comments! Michael, I had that exact same idea. I didn't put the percentages on because I thought it would confuse people since the axis in the raw hours. Personally, I agree that it would be helpful.