Thx!

It’s very hard to accurately compare the wedge shapes in pies, because humans aren’t that good at determining angle sizes or angular area. If you were just looking at the pie chart above, could you tell what the difference between Lynx and Opera is easily and accurately? (That’s not quite fair, since the pie chart’s not actually labelled.) Most importantly, can you easily tell that Opera represents 6x Lynx? If you can’t tell simply by looking at the graph (not the data table), it’s not a useful visualization. Wikipedia’s got a terrific example of this problem. The examples of pie charts in Google’s Chart API impart literally no information at all other than “some things make up some other thing.”

Also, this particular dataset doesn’t need to show how much a given browser represents out of 100%. It’s probably obvious that if MSIE is about a quarter of the chart (and there’s no way to know this without manually doing the math–the chart doesn’t tell you if it’s 25%, 28%, or what) that it’s out of some total. This chart just says that it’s “out of 100%”, but not 100% of what. If you were comparing multiple datasets against each other, than the total might matter: for instance browser use by men vs. browser use by women. Even then, it’s probably unimportant.

A good rule of thumb is that any pie chart almost certainly would be better as a plain vertical bar graph. It’s easy to visually compare the relative sizes of various rectangles, plus a bar chart would give you a place to put labels or ticks which would clarify this data a lot. Finally: if you only have one column of numbers to compare, the original plain data table is the most effective presentation of them! Let the numbers speak for themselves.

I’d love to see this rewritten to show how to generate bar charts, perhaps with some datasets that really merited visualizations.

Thanks for the tutorial!