Seth Godin invited a reaction last week when he ventured into unfamiliar territory by publishing a blog post entitled “The three laws of great graphs.”
- One story
- No bar charts
He was apparently engaging in a bit of hyperbole—a statement that is intentionally exaggerated for the purpose of making a point—when he listed as one of his rules: “No bar charts.” When Jesus made use of hyperbole to warn that great wealth can lead to corruption—”It is easier for a camel to pass through the eye of a needle than for a rich man to enter into the kingdom of heaven”—I suspect most listeners understood that he was using an exaggerated analogy to make his point. Godin’s spare but sweeping statement, “No bar charts,” however, runs the risk of being taken literally by many who struggle with data presentation, and have an undiscerning appetite for simplistic rules.
In a follow-up blog entry on the following day, Godin responded to the outcry of those he calls “data presentation purists” by explaining what he supposedly really meant: “bar charts are dramatically overrated.” I agree that bar graphs are often used when a different type of graph would do the job much better, but they are not overrated. In fact, for general presentation purposes, bar graphs along with line graphs are the two most valuable graphical tools for telling quantitative stories. Godin got it quite wrong when he wrote: “The correct use of a bar chart is to show how several items change over a period of time.” The story of change over time is actually told much more effectively with a line graph. Bar graphs work better than anything else for simple magnitude comparisons. In regards to this, Godin also veered from good advice when he argued that one should use “a simple pie chart when comparing two or three items at the same scale.” In fact, pie charts do one thing and one thing only—they compare parts of a whole (things that add up to 100% of something). They should never be used for other magnitude comparisons.
The example that Godin used to illustrate the superiority of pie charts actually tells a story that could be better told using a simple bar graph, even though it might indeed display parts of a whole (whether it does or not isn’t clear without context).
One of the first things that I noticed when I saw this chart is that it includes elements that aren’t needed to tell the story. Notice that the slices are directly labeled in the pie chart itself, which makes it entirely unnecessary to include a legend. Doing so works against Godin’s objective of keeping the story simple, because it draws people’s attention to items that add no meaning to the chart, thus wasting their time.
According to Godin, the message of the chart is, “Trolls are where we should focus our energy.” Apparently, however, the other creatures shown in Godin’s pie chart must also be important to the story, or else why not lump them together into a single slice labeled “All Other Creatures”? Based on this assumption, a bar graph would make it much easier to compare the values associated with the various creatures, and could be designed in a way that still clearly emphasized the greater importance of trolls to the story, as illustrated below:
There’s no need to address the other issues raised by Godin’s argument, because Jon Peltier has already done so with great dispatch in his blog. I would like to address Godin’s final rule, “Motion,” which he elaborates by saying, “You should animate your charts.” While there are times when it is useful to animate a chart in some manner, more often than not it adds nothing to and at times even obscures the story. Here are Godin’s specific instructions:
It’s simple: create two slides. The first one shows where the data used to be, the second one, on the same axes, shows where it is or where it’s going. Motion.
Establish the first slide. Make your point about your source and its validity. Then press the advance button. Boom.
Boom? While I appreciate Godin’s intention, there are usually better ways to show change in a presentation. In many instances, it is better to show change in a single line graph, which allows people to see the full course of change from one point in time to the last as a single, simple picture. When the story is better told using a different type of chart, such as a bar graph (or, God forbid, a pie chart), it almost always works better to begin by displaying a single graph that displays the state of things at one point in time and then cause a second graph that displays the state of things at the other point in time to appear beside or below the first. Doing it this way makes it possible for people to compare the two states, such as the past and today, which is almost impossible to do when they can only see one graph at a time.
I’m not trying to pick a fight with Seth Godin. He’s an excellent speaker with areas of expertise that differ from mine. When he gives advice about graphs, however, he wanders into my backyard, and I can’t help but point out the flaws when he gets it wrong. Perhaps I’m just one of those “data presentation purists” who believes that the rules ought to actually work, confirmed by thoughtful research.