This blog is an attempt to promote better charts through the repeated application of various guidelines and best practices that can (and should) be employed when visualizing data. I did not set out to simply describe how to build effective, simple, attractive charts that can best summarize your data and point out its most salient features and patterns. There are plenty of materials out there for that. Rather, I thought that I would start from various examples that I stumbled upon - and point out some of the ways in which they could be made better.
If we start with a set of charts “caught in the wild”, and if we can determine the ways in which they fall short, then we should end up with a decent toolset of best practices that can guide us going forward. Crucially, each time I point out some deficiencies, I also make an attempt to come up with a better performing version myself. If I don’t think that I can assemble something that works better, then I refrain from criticizing it in the first place. [The opposite might work too. Perhaps I will also have some posts about charts that look quite nice to me - and are therefore hard to improve - in the future.]
I suspect that this method - start from an existing plot and point out everything that could be improved - is one of the better approaches that we have for learning how to build high quality visualizations that do a great job at communicating the most salient features of our data.
What do you mean, “Wrong”?
None of the original charts shown on this blog are wrong, as far as I can tell. My assumption is that the data has been processed and plotted correctly. Any errors made there are completely outside the scope of this blog. (Besides, I certainly don’t have access to the materials used to produce the original charts.)
Rather, “wrong” refers to various deficiencies that apply and become apparent during the process of analyzing the data and building charts to effectively communicate our findings. From that point of view, it is wrong to…
- Assume that others are aware of all the intricacies behind the data, and understand right away what you want to communicate.
- Prioritize form over function, by choosing certain chart formats simply because those are considered to look “better” (by someone who typically, in fact, doesn’t know better).
- Not make your charts self-explanatory and thus easier to read and understand. Chances are that many kinds of things will seem quite obvious to the analyst - while being far from obvious to the reader. Paying attention and adding concise yet fairly complete titles, subtitles, data labels, footnotes and so on, helps to ensure that our charts will become not just easy to read, but also hard to misunderstand.
- Obfuscate and make things more difficult for the readers than they need to be. Communication is hard, and repeating the same (or similar) information multiple times can help to carry our point across.
- Not be alert and not pay attention to all the ways in which charts could be misread and misinterpreted.
Finally, “wrong” also applies to my versions. It is possible that my attempts will occasionally (always? haha) lead to updates that are worse than the originals.
My versions of the original charts
Since I do not have access to the materials used to assemble the original charts, I did my best to locate a copy of the data. Sometimes that was fairly easy, sometimes it was not. If you notice some data differences between the Before and After charts, the explanation is that I could not find the original data, and ended up approximating the values from the charts. That is obviously not ideal and you do not want to do this for your projects; but for illustrating concepts related to communication and chart design, it works just fine.
Chart Utility vs. Graphical Design
Finally, I want to stress that this blog is focused primarily on the utility - rather than the Graphical Design - of charts. There are many dimensions that I largely ignore here: colors, fonts, size, orientation, arrangement on the page, transparency, etc. It’s unlikely that all of my versions look better than the originals (if any!). That was not one of my objectives. However, I believe that my versions do a better job at presenting the data in the context of a reader who wants to understand it to the extent that s/he can – as opposed to a chart that should just “look great” in a presentation, memo or article.