Subscribe to receive blog updates

Search form

search-btn

search-btn

Oct 12, 2015

Fancy Doesn't Mean Better!


This is a guest post by Janet Wesner.

The Economist’s daily chart is one of my frequent stops for data visualization ideas. Usually, their charts exhibit standard data visualization best practices, but once in a while they post something that makes me scratch my head. This post on how countries spend their money is a good example. I’ve already made a Slemma visualization with alternative charts based on the same dataset and critiqued the conclusions that were drawn in the post by The Economist. Now, I’d like to focus specifically on the issues with their visualization. Here’s their chart:



Things that are wrong with this chart:

  1. Chart Choice. Just because there’s a lot of hate on pie charts that hasn’t been directly voiced for bubble charts, the concept here is the same: they’re difficult to compare and hard to align. For some of the bubbles where the size difference is large-ish, I can tell which bubble is larger but not by how much. For the bubbles close in size, I just have no idea unless I use a ruler (and probably a magnifying glass)
  2. Specific Design Elements. The strong border for highest and lowest spend make the bubbles larger. This is especially a problem for the lowest spend bubbles because it makes what should be the smallest bubble look larger. Take the Education category for example – the EU-28 bubble looks smaller than the Russia bubble.
  3. Data Integrity. Based on the legend, the bubbles are orange if they represent a value that is above average for the category, and similarly blue means below average. So my first thought is that average means the simple average of the ten columns shown. There are a few problems with this:
    1. EU-28 is not a country.
    2. Because there are ten countries shown, I would expect 5 orange bubbles and 5 blue bubbles for each category row. Clearly, this is not the case.
    3. If the average does not mean simple average across, could the creator be using the EU-28 average as the average? The problem with this is:
      1. None of individual countries shown belong to the EU-28 and so without any additional commentary, that’s not what a user would typically assume as the average.
      2. The EU-28 column of bubbles vary between the orange and blue colors.



    An alternative?
    The simplest and most commonly suggested alternative would be to use bar charts, and for great reason: All the bars share the same baseline and your eyes would only need to compare the endpoints. Here’s my redesign using a simple stacked bar chart:
    With all the categories shown, the chart does look busy (I could have created separate bar charts for each category but the downside would be that the visualization would take up a lot more screen space). The filter toggle at the top helps by allowing users to show only a few selected categories. Also, showing the labels of actual percentages prevents users from needing to do too much visual math.

    Though I think this is an improvement over the bubble chart (and likely took way less time to create), the stacked bar chart does have problems. For example, the stacked bar chart has too many colors and doesn’t highlight which countries have higher/lower spend than average. 

    Since I’m recently having a fascination with small multiples charts (also known as panel charts, sparklines, trellis charts), I tried testing out this concept using funnel charts.
    What impresses me most about good small multiples charts is that they’re able to show so much information in a condensed space. To make it readable though, it’s important to have a consistent design across graphs, including having a single axis for all charts… which leads me to what I think is the major downfall of using funnel charts in small multiples: Categories for each country is sorted in descending order from largest to lowest spend so the category order is not the same for the ten funnel charts (highest spend in the US is Health, in Mexico/India/Russia is Food, and in the rest is Housing). 
    With the tooltips that show up on mouse over, users can see which category each segment is. But, that’s extra work for the user. If all ten small charts could be put in the same row and have the same category order (something like a violin chart), then I think that would greatly improve the visualization. Short of being able to create violin charts, a simple sorted table actually works quite well too:

    After all the testing, I guess this goes to show that just because you can use fancy chart designs, it doesn’t mean you should.

    No comments :

    Post a Comment

    Share a post