Why 3D pie charts are bad
fury.com
fury.com
The percentage of blue pixels in the chart is shown by the red section of the chart, and vice-versa (notice the labeling).
See also: http://xkcd.com/688/
This isn't a post about why pie charts are bad, it's a post about why lazy charters who waste our time are bad.
Finally, too many charts rely on colour rather than patterns or markings to distinguish choices. That usually makes them hard to read for colour-blind people, as light shades of green and yellow, or dark blue and purple, etc. end up being used.
[1] http://lilt.ilstu.edu/gmklass/pos138/datadisplay/badchart.ht...
Bar charts, especially horizontal, have skimmable labels. Pie charts do not.
For example: http://www.olivierlorrain.com/2010/02/21/data-visualization-...
In other words, I like the irony but this is not really a good example.
This is the author's point.
1. Totally dead-on, so it appears to be 2D (Good)
2. Parallel to the surface of the pie, so you can't see the pie and wind up using a better chart type instead (Better)
There are more blue pixels in the image than red. This is because of the thickness of the disk. Therefore in representing the proportion of blue pixels it's more than half the area of the circle. Therefore he has one slice which is more than half the area of the circle, and that half represents the blue pixels. He colors that red.
And vice versa.
The diagram is completely accurate, and very misleading. Adn deliberately so. It can be difficult to get pie charts to represent data accurately, and in a way that doesn't mislead the viewer.
On the other hand, correct use of pie charts can be very effective. It's just that you don't often see it.
THAT Pie chart was bad because of the legend being self representational, not the isometric projection.
So while you can clearly see the size of the elements, it requires thought and interpretation. Compare to a 2D flat chart, which is very clear although not so pretty. Shouldn't a visualisation require as little thought as possible?
I blame PowerPoint.
Not to mention with split bars you can more easily show the proportional change of things over another axis (usually time).
And the whole point of a pie chart is to use one thing -- the size of a slice -- to represent a single datum.
A 3d pie chart muddles both of those things for form-over-function reasons. How is this even contentious? Pie charts are notoriously bad anyway, especially if there's not a huge difference between the data points. This makes them notably worse.
So could you accurately estimate the two sizes? Research suggests you'd be better off with a bar chart than with a pie chart except in certain cases (25% vs 75%) and that isometric pie charts are worse than flat ones.
You'd want the blue part smelling like blueberries, and the red part smelling like strawberries.