Charts and graphs! Charts and graphs!

Welcome, and thanks for stopping by! Subscribe to my RSS feed so you don't miss a single exciting post (or any of my regular ones). Thanks for visiting!

Beyond words

So I’m a copywriter by trade but, for me, it’s not about writing. It’s about communicating ideas. It’s about making a point. And, sometimes, that’s best done with pictures, not words. Frequently, those pictures and in the form of charts and graphs.

But you know what? A lot of charts and graphs stink.

An example of a pretty, but not useful, graph

Oh, sure, they’ll be pretty enough, but they won’t do their job, which is to convey information.

Consider this example (which I found at a site that purports to teach graphing methods). At first, it looks like a fairly straightforward graph, not unlike many of those in most news magazines these days. But it has a few problems.

So what are the problems? First off, instead of being a straight-on view, the graph is tilted and the viewer looks down at it slightly, making it harder to compare the bars visually to see how the numbers relate to each other. After some tweaking to take care of that, we get this:

Not perfect, but we can work with this and address the next problem: What actually represents the data? Specifically, do the data bars include the head of the match or not?

Since the artist thoughtfully included the data (more on that in a bit), I made new chart and overlaid them on our example. I scaled the overlay by fitting my last bar to the last tallest match. This gives:

As you can see, none of the other bars fit any of the matches, so it is not likely the data bars include the match heads. Let’s double check:

Well, the first bar now matches, but none of the others do, which leads us to the last problem: Are the bars even correct? After all, the two bars with a value of 0.8 should be the same height, and the 2.7-valued bar should be about twice as tall as the 1.4-valued bar, and they are not.

I believe this image tells us that the data bars do not include the match heads, and the middle bars’ lack of fit can be explained two ways:

1. The lack of fit results from the data being graphed in 3-D originally and the effects of perspective were not accounted for in straightening the graph out.
2. When graphing the data, the match sticks came out so short and small compared to the match heads that the artists lengthened each bar a bit to produce a better visual.

Honestly, I don’t know which it was, but it doesn’t really matter. In both cases we ended up with a chart that does not clearly represent the data it seeks to show.

So what should this graph look like?

Using the data provided, a correct graph would look like this:

This graph is clean and easy to read. There is no uncertainty about what’s part of a data bar and what isn’t. The two bars representing 0.8 are the same height, and the 2.7 bar is roughly twice the height of the 1.4 bar, as they should be.

Of course, the graph does have other problems, not the least of which is no definition of what “death rate” actually means is given. Deaths per hundreds of people? Thousands? And per week? Month? Year?

Make the graph worth doing

The last graph clearly and unambiguously communicates the (meaningless) data, without demanding too much work of the viewer. Of course, some of you may complain about the lack of data labels on the graph. To you I say, if the bars need to be labeled, what’s the point of making the graph?

Actually, in this particular case, there isn’t much point in making the graph. Ignoring for now that we don’t know what “death rate” means and so cannot really interpret the graph, there exists a superior solution for communicating this data. More on that later.