You’ve been having so much fun with the last graph that I nominated for the Graph Hall of Shame, I couldn’t wait to announce the next nominee. This example comes from iDashboards. What you see below is an excerpt from a larger dashboard. Take a moment to examine it and make a list of its problems. Include anything that gets in the way of clear and efficient communication. Keep in mind that this is from a dashboard, and that one of its goals therefore is rapid communication.

Click to enlarge

Where to begin? Here’s my list of problems.

• The quantitative scale does not begin at zero. Not all quantitative scales must include zero, but when bars are used to encode the values, zero must be there. One of the ways that a bar encodes a value is by its height (or length, in the case of horizontal bars). If bars begin at some value other than zero, their heights do not accurately encode the values they represent. Also, without zero, comparing the heights of various bars does not accurately represent the differences in value between them. For example, it appears that the bar above the label “Maine” is about 150% the height of the bar to its right, but it is actually only about 107% the smaller bar’s value. I realize that the zigzag shape at the base of each bar is meant to indicate that the scale has been truncated, but what’s the point of using bars to graphically represent the data if you can’t compare them to one another? Without this ability, a table of the same data would work much better.
• You probably noticed in my point above that I didn’t refer by name to the bar next to the one labeled “Maine”. This is because it isn’t labeled. Only every other bar is labeled. You can get by with not labeling every month or year in a time-series graph, because we all know which month follows April or which year follows 2002, but we don’t all know which state follows Maine alphabetically, at least not without difficulty.
• Even by labeling only every other state, the state names could not be placed side by side without orienting them sideways. Unfortunately, this makes it difficult to know to which bar a particular label refers. For instance, does the “Maine” label actually refer to the bar above it or to the smaller bar to the right? This depends on whether the labels are centered below the bars or right justfied such that they end under the corresponding bar.
• If the purpose of this graph is to compare the average salaries of the various states to one another, ranking them by size (such as from largest to smallest) would have made this much easier to see. Alphabetical order is only useful for looking up individual items, which is something that tables support nicely, but graphs should be used to feature patterns in the data.
• On a dashboard, which is used to monitor what’s going on, do we really need to see the average salary for every single state? Probably not. It would probably be more appropriate to only show those states with the highest salaries or a combination of the top and bottom states (such as the top five and bottom five).

More problems could be added to the list, but I’ll leave it to you to point out the additional flaws.

Take care,