Radar: More Evil Than Pie?

It is by now common knowledge among the viz-savvy crowd that pie charts are best avoided. Many pixels have been used explaining why pies are bad, when occasionally pies are good, and my personal favorite from my colleague at Darkhorse: Salvaging the Pie. Far less attention, however, has been paid to pie’s evil cousin, the radar chart. Perhaps it’s not as ubiquitous, but it can be more nefarious.

One such example recently appeared in a Harvard Business Review (HBR) blog post: CEOs Get Paid Too Much, According to Pretty Much Everyone in the World

The visual design is appealing. The clarity is appalling. There are good things to be said about the choice of contrasting but complementing colors. The subtle axes are helpful without obscuring the data. The glaring question here is why a radar chart? Do countries somehow naturally follow each other alphabetically in a circle? (Only at the Olympic closing ceremonies!)

Comparison of data points is difficult
To compare one country to another, your eyes have to do a fair bit of work either by following the circle around, or counting grid squares. For example, how does the United States compare to Italy? You have three seconds. Ready, set, go!

Facts are lost in the mushy middle
What's Spain's value (or is that Sweden?). How about Hungary’s “Ideal” value? The inside blue values are almost impossible to read.

Over-emphasis of high numbers

The gap between countries gets larger as values increase. When connecting the points the area created by points higher on the scale is disproportionately larger than those created at the bottom.

To illustrate this point, compare the following charts.

The area of the second chart is four times larger than the first even though we've only doubled the values. The third chart is nine times larger than the first despite only tripling. The area grows exponentially as the values increase. The result is a disproportionate emphasis on the top few countries.

Sequence determines area
Another thing that happens in radar charts is the arrangement of the data points influences the area they create. The charts below plot the exact same data (three 9s and three 1s) but in a different order. The result is obviously problematic as neighboring points amplify each other.

Since the alphabetical sort order in the HBR chart is arbitrary, the shape created is equally arbitrary. The random “explosion” could have been ordered by value, creating this bizarre-looking seashell:

Disclaimer: the paper did not publish its raw data table. Transcribing the values here was an inaccurate exercise which involved lots fingers on the screen, squinting, and neck twisting. In the end, the values are probably not quite right.

A better way
So, what should this data really look like? Here is a quick take on a simpler representation of the same data:

The original chart obscures the facts. Perhaps the original chart's purpose was to provoke more than to inform. But the chart does so at the expense of the author's argument. We can't even see the points she wishes to highlight. In the new chart, interesting facts begin to emerge. It is now easy to compare any two countries and we can see that Spain’s “estimated” is actually twice as high as Sweden’s, whereas we would have easily confused them before. We see that although the “ideal” is always smaller than the “estimated”, there is little relationship between the two. For example, while South Korea and Australia are neck-and-neck for the lead in “estimated” ratio, Taiwan’s “ideal” is almost double that of any other nation.

In conclusion...
The radar chart should rarely be used. Maybe it should never be used? I can’t for the life of me think of a situation where its power of engagement compensates for its lack of clarity.

Do you ever use radar charts? Have you ever seen a radar chart used well? I would love to put a positive spin on this and showcase some successes. Let me know in the comments below.