Yesterday, I blogged about the new collaborative data visualization site called Many Eyes. I am only one among many who have contributed comments to the websphere in the last few days about Many Eyes. In scanning through some of the comments by others, I found an article by Shane Schick, who, although well intentioned, doesn’t seem to understand data visualization. Here’s an excerpt from Shane’s article titled “I’ll believe it when I see it”:
I’m more of a pie chart guy than a bar chart guy. With bar charts you have to read along the bottom to see what the categories are and then back to the bars themselves to determine how much larger one thing is than another. With a pie chart it’s all laid out for you, and it takes only a glance to assess the proportion. To me, pie carts are just easier on the eyes. That won’t necessarily be the case for Many Eyes.
Not being a data visualization expert, Shane’s comments about pie charts vs. bar charts are actually more accurate if you attach what he said about pie charts to bar charts, and vice versa. Unlike pie charts, which typically use a legend to label the slices, bars on a bar chart are labeled directly, just below the bars. You can read a bar’s label without moving your eyes, but with most pie charts your eyes must bounce back and forth again and again between the slices and the legend. Also, visual perception is notoriously bad at comparing 2-D areas, such as pie slices, whereas our greatest perceptual precision is achieved by comparing the lengths of objects that share a common baseline and orientation, such as the bars in a bar chart, or in comparing the 2-D position of objects along a common dimension, such as the ends of bars. Finally, most pie charts are actually harder, not easier on the eyes, because they consist of several colors (one per slice), which are usually (although not necessarily) bright, garish, and therefore jarring to the eyes. When a bar chart is used to display the same information that could also be displayed in a pie chart, all the bars share a single color (unless you unnecessarily make each bar a different color, which is gratuitous nonsense). A considerable body of research has confirmed the superiority of bar charts in supporting all but a few rarely used operations of graph interpretation that can also be performed with pies.
It’s important to mention that pie charts are used for only one type of quantitative comparison, whereas bar charts support many purposes that require the comparison of magnitudes (that is, the relative sizes of the values). Pie charts are used exclusively for part-to-whole comparisons. In fact, the one advantage that a pie chart has compared to bar charts is that when you look at one, you know you are seeing a comparison of the parts of some whole, such as regional sales (the slices), which add up to total sales (the whole pie). Because bar charts can be used to display many types of quantitative relationships, when you look at one that is used to display a part-to-whole relationship, this fact isn’t as obvious, despite a quantitative scale that’s expressed as percentages. All that this means, however, is that you must state this fact, such as in the title “Regional Breakdown of Total Sales.” In other words, even if Shane’s preference for pie charts were backed by the evidence of what actually works best, he would still need bar charts for a host of purposes that pie charts cannot support.
At one point in his article, Shane complains: “If you’re the lone pie chart person among a group of histogram people, how well are you going to be able to articulate a vision?” The truth is, when it comes to presenting data, you can’t be either a pie chart person or a histogram person. To articulate your vision effectively, you must know how to match the data and your message to the appropriate means of display. Pie charts, which exclusively display part-to-whole relationships, and histograms (a specialized version of a bar chart), which exclusively display frequency distributions, don’t align themselves with distinct teams that are trying to reach the same goal. To articulate your vision, you must speak the language—in this case the language of graphs.