An elaborate data vessel
Chartjunk as marketing copy

Two metrics in-fighting

The Wall Street Journal shows the following chart which pits two metrics against each other:


The primary metric is the change in median yearly salary between the two periods of time. We presume it's primary because of its presence in the chart title, and the blue bars being more readable than the green bubbles. The secondary metric is the median yearly salary in the later period.

That, I believe, was the intended design. When I saw this chart, my eyes went to the numbers inside the green bubbles. Perhaps it's because I didn't read the chart title first, and the horizontal axis wasn't labelled so it wasn't obvious what the blue bars coded.

As with most bubble charts, the data labels exist to cover up the inadequacy of circular areas. The self-sufficiency test - removing the data labels - shows this well:


It's simply impossible to know what values should be in each bubble, or to perceive the relative sizes of those bubbles.


Reversing the order of the blue bars also helps:


The original order is one of the more annoying features in most visualization packages. Because internally, the categories are numbered 1, 2, 3, ..., and because the convention is to have values run higher as they run up the vertical axis, these packages would place the top-ranked item at the bottom of the chart.

Most people read top to bottom, which means that they read the least important item first, and the most important item last!

In most visualization packages, it takes only 1 click or 1 action to reverse the order of the items. Please do it!


For change over time, I like using a Bumps chart, otherwise called a slope graph:



Feed You can follow this conversation by subscribing to the comment feed for this post.


I agree with most of your points, but I'm not sure that a bumps chart is the best solution here. The strength of the original chart is that it provides a somewhat normalized view of salary change across industries--what I believe is the purpose of the chart. This insight is somewhat diluted in the bumps chart, unless you can find an intuitive way to compare slopes.


What if you stacked the bubbles from smallest at the bottom ($43k) to largest at the top ($74k), similar to the slope graph? Then it's apparent which bubble is the biggest as they will grow in size as you go up. The slope graph is nice though.

The comments to this entry are closed.