Dec 12
Book Review: Tufte’s Visual Display of Quantitative Information
in book review
3 Comments »

I’ve just finished reading Edward Tufte’s Visual Display of Quantitative Information (amazon link). I picked it up because I’ve seen good things mentioned about it online, and I was wondering, “Who is this guru of excellent design, and what does he have to say?”


I originally wet my appetite with commentary from sites like these:

The links above were intriguing, so I checked out the book. I must say that it is an interesting read, but it’s much more of a coffee table book than a serious reference manual. I’m actually glad that I didn’t buy it myself, because I would only ever read it once. In it, Tufte spends a lot of time showing good and bad, and ugly, graphics, and dissecting what makes them good and bad, and ugly. If you were to read the second link above, you would get a lot of what the meaty part of the text is - the rest is commentary about specific examples.

I did enjoy reading the book, but as I said, it’s more of a coffee table book. In other words, it was a pleasure to read, and the graphs, both good and bad, are very interesting. However, I can’t say that I learned that much. Okay, it’s not good to lie in your graphs. Fine, I think I can figure that out. Tufte’s examples of graphs in modern media that stretch the truth (be it on purpose or because the authors of those graphs don’t realize what they’re doing wrong) is fascinating.

I was disappointed in the book as a reference manual because some of the advice that Tufte gives is impractical. For example:

  • Don’t use graphs when a table suffices. He likes graphs when there is a lot of information, but not necessarily for smaller, say n=30, data points. Unfortunately, a lot of the time, I’m only dealing with a few things. I still like a nice little graph. Actually, more importantly, my supervisors like a little graph.
  • Don’t waste ink on stupid stuff. (I’m paraphrasing here). For example, the axes of a graph are not necessarily important - try erasing parts of the axis that are not important, i.e., the axes line that is less than your smallest value, and greater than your largest value.
  • This ends up producing a graph that looks like you’ve got a bad printer, i.e., you’re missing parts of your lines. (He also suggests that sometimes you can make do with a slightly offset line in places to emphasize things - I hate offset lines - looks like a cheap printer.)
  • Much of Tufte’s advice is useless for modern users. We’re using Excel, or the open office version, to create graphs. Excel can only do so much with your data. I don’t even think that I can delete part of my axes, even after hacking Excel. Granted, this limitation has nothing to do with Tufte, and everything to do with the tools that we’re stuck with, but it’s kind of like reading a book on how to write calligraphy, when the only tool you have is a typewriter.

    • That said, Tufte shows some amazing examples of wicked graphs. The most interesting ones are not purely quantitative in nature. For example, there’s a neat drawing of the life cycle of some kind of bug. Fascinating, but I wouldn’t have called it a “graph” per se. My most favourite one is a drawing showing the path that Napolean took on his way to Moscow, where a large tan line shows the size of his army, marching towards the right of the page. The graph also shows the return trip, and you can see how many soldiers have been lost, as the line gets smaller and smaller. The bottom axis of the graph is actually the temperature, showing that as it got colder, more and more soldiers died. I find this graph fascinating, especially because the dimensions involved include: time, direction, size of army, and temperature. In addition, it’s all placed on what looks like a map, showing major rivers, etc. etc. I would actually recommend taking the book out of the library, for this particular example alone.