I eagerly anticipated reading this book. I frequently design data visualizations for my job as a software engineer, and I have a deep love for effective graphs. I love to read about different strategies for representing information visually, and I know that Tufte's work in this area is very highly regarded.
I was completely astounded at how poorly argued this book is, how bizarre its recommendations can be, and how disdainful the author feels about any attempt to make graphs attractive. I know these are bold allegations against such a highly regarded work, so let me be specific.
Tufte argues in favor of graphic minimalism. He doesn't use the word "minimalism", but his principles include "erase non-data-ink, within reason" and "erase redundant data-ink, within reason." This seems reasonable on face -- who would argue in favor of redundancy? -- but he applies this in absurd ways. For example, the graph he uses to explain the idea of "redundant data-ink" is a bar chart with a single vertical bar on it, and a number on top of the bar. He writes:
"[this chart] unambiguously locates the altitude in six separate ways (any five of the six can be erased and the sixth will still indicate the height): as the (1) height of the left line, (2) height of the shading, (3) height of the right line, (4) position of the top horizontal line, (5) position (not content) of the number at the bar's top, and (6) the number itself. That is more ways than are needed."
I stopped for a second when I read this; surely Dr. Tufte is not arguing that a bar chart is inherently ambiguous because the bars are both outlined *and* filled, is he? But in case there was any question, he reinforces this concept a few pages later, when he takes a different bar chart and removes all of those "redundant" lines, and ends up with something truly unintelligible. Of this peculiar result he writes "The data graphical arithmetic looks like this--the original design equals the erased part plus the good part." I wish I could include the illustration in this review, because with words alone I simply cannot communicate how much worse Tufte's revision of this graphic is.
There are so many examples of this, but I will give just one more. At the beginning of Chapter 6, Tufte revisits the traditional box plot and again finds that it "can be mostly erased without loss of information." After offering several iterations of his minimalistic approach, he settles on a version which is just astoundingly bad. To represent the five data points (quartiles) Tufte draws a single line that is offset by a *miniscule* amount between the 25th and 75th percentiles, and has a *miniscule* break at the median. It is not hyperbole to say that when my eyes are 18 inches away from this graphic, the quartiles can barely be seen at all; it looks like he just drew a straight line. About this Tufte says "This design is the preferred form of the quartile plot. It uses the ink effectively and looks good."
These are examples of a larger trend throughout the book, which is to state general principles without much support, and then to judge graphs (and people's intelligence) by how well they adhere to those principles. Here is an example. In Chapter 3, Tufte argues that "relational" graphs -- graphs that show the relationship between two or more variables -- are more sophisticated than time-series or map-based graphs. I will include Tufte's entire analysis in support of this principle, because it will readily fit into this box:
"In order to make comparisons among a variety of newspapers, magazines, scientific journals, and books, I have compiled a rough measure of graphical sophistication--the share of a publication's graphics that are *relational*. Such a design links two or more variables but is not a time-series or a map. Relational graphics are essential to competent statistical analysis since they confront statements about cause and effect with evidence, showing how one variable affects another."
My first reaction (and I hope yours) to this was to note that relational graphs show how one variable is *correlated* with another, and cannot by themselves show cause and effect (we can thank statistics for an endless supply of "information" about what supposedly causes cancer). But besides that is just the overwhelming lack of support for the idea that we can judge the sophistication of a publication on what percentage of its graphs are relational. But that's exactly what Tufte proceeds to do; he trots out a table of publications from different countries and their "sophistication percentages", and uses it to achieve some conclusion that the Japanese are much smarter than anybody else, and the Americans stupider.
Another example of an unsupported principle: that more information is better. Throughout the book Tufte is consistently impressed when someone has discovered a way to cram more bits of information into the same graphic. For example, from page 20: "The most extensive data maps, such as the cancer atlas and the count of the galaxies, place millions of bits of information on a single page before our eyes. No other method for the display of statistical information is so powerful." This attitude inspires the reader to include as much information as they possibly can in their graphs. But Tufte never stops to ask the question: is there a point when more information just becomes noise? To quote Google documentation about their charts API: "Take care not to overestimate the number of data points required for a chart. For example, to show how popular chocolate ice cream was over the last ten years, aggregating search queries for each day would result in more than 3600 values. It would not make any sense to plot a graph at this granularity."
The major credit to Tufte's book is that he includes many examples of creatively designed graphs, many of them historical. He is particularly taken with a diagram of Napoleon's ill-fated attack on Moscow, which is undoubtedly a very engaging and effective graphic. But this makes Tufte's minimalistic recommendations all the more puzzling. He seems to completely miss that almost none of the historical work he admires follows the principles he spends the rest of the book advancing. Most of them use grid lines (which he hates; they are non-data-ink) and they invest effort into being attractive (which he sees as a dumbing down of graphs; he calls any visual flare "chartjunk.").
Tufte's principles totally ignore the primary purpose of graphs, which is to show a data set's *patterns* (or lack thereof) to humans. This is confounding, because many of the examples he cites do this brilliantly. His very first example, Anscombe's quartet (you can Google for it) is a fantastic example of how graphs show patterns even when basic statistical summaries do not. His Napoleon example shows the pattern of how the size of Napoleon's army was so severely diminished over time and space, and the points at which it suffered its greatest casualties. But Tufte seems to completely miss the point. Though his examples repeatedly show patterns, Tufte never talks about patterns at all. About the Napoleon example, Tufte writes "Minard's graphic tells a rich, coherent story with its multivariate data, for more enlightening than just a single number bouncing along over time. *Six* variables are plotted: the six of the army, its location on a two-dimensional surface, direction of the army's movement, and temperature on various dates during the retreat from Moscow." Tufte again is primarily impressed with the amount of data and the number of dimensions.
Principles like "remove non-data ink" and "forgo chartjunk" treat graphs as though they are a form of compression, and treat "ink" as a scarce resource. The truth is that the primary goal of a graph is to communicate data to a human, and humans respond to design and polish (if they did not, there would not be so many colors, icons, boxes, visual effects, etc. on the page you are viewing right now). Design can communicate structure. Visual weight can help draw the eye to the part of the graph that is most significant. Polish can make a graph visually appealing enough to look at in the first place. Tufte has no appreciation for these ideas: "Chartjunk does not achieve the goals of its propagators. The overwhelming fact of data graphics is that they stand or fall on their content, gracefully displayed. Graphics do not become attractive and interesting through the addition of ornamental hatching and false perspective to a few bars." This attitude puts Tufte in the company of usability expert Jakob Nielsen, who probably has good points to make, but when you visit his bland and text-heavy website [...] are you really inspired to spend time there reading?
This review is getting too long, so I can only just briefly state some more of my numerous problems with this book: he makes unsupported indictments against moire (patterns of lines or dots used to fill in regions), he spends almost no time talking about color, COLOR! (most of what he does say is negative -- he prefers grayscale), he rails against the idea of making graphs attractive or readily-understandable (he says that if the graph looks boring it's because you chose the wrong numbers), many of the graphs he cites are confusing or under-explained.
I don't know how to explain the high regard for this book. There are lots of beautiful graphs, to be sure, but most of them are not Tufte's and don't follow his principles. I am disappointed in what I expected to be a great book.
Loading your book clubs
There was a problem loading your book clubs. Please try again.
Not in a club? Learn more
Join or create book clubs
Choose books together
Track your books
Bring your club to Amazon Book Clubs, start a new book club and invite your friends to join, or find a club that’s right for you for free.
Download the free Kindle app and start reading Kindle books instantly on your smartphone, tablet, or computer - no Kindle device required. Learn more
Read instantly on your browser with Kindle for Web.
Using your mobile phone camera - scan the code below and download the Kindle app.
Flip to back Flip to front
Follow the Author
Something went wrong. Please try your request again later.
Discover the latest buzz-worthy books, from mysteries and romance to humor and nonfiction. Explore more
Frequently bought together
$27.20
Get it as soon as Thursday, Jul 27
In Stock.
Total price:
To see our price, add these items to your cart.
Choose items to buy together.
What do customers buy after viewing this item?
Page 1 of 1 Start overPage 1 of 1
About the author
Follow authors to get new release updates, plus improved recommendations.
Book is great (Amazon service and the physical book corner is damaged...not the author's fault)
Reviewed in the United States on August 1, 2020
As expected, this book (content) is great. Took one star off due to poor Amazon services: we cannot communicate with Amazon about if they would or could provide a replacement which is NOT damaged. the only choice is to take the risk and buy another book in order to compare the two same books to decide which one is in good condition. Since we pay full price for new hardcopy, we expect the book to be in new and good condition. As you can see the picture, it is not. Again it is not the author's fault. We don't want to return it just in case the book is sold out. We would like to exchange it but there is no way to communicate with Amazon (does Amazon care?) to see if there is a brand new book available that is not damaged. Sadly we are keeping this book
Reviewed in the United States on August 1, 2020
Images in this review
Top reviews from other countries
Walter
Imperdivel para quem tem de apresentar dados
Reviewed in Brazil 🇧🇷 on January 9, 2018
chillipadi
Excellent delivery time & communication, but faulty product (see pics)
Reviewed in Australia 🇦🇺 on November 30, 2020
chillipadi
Reviewed in Australia 🇦🇺 on November 30, 2020
I was left quite disappointed when I opened the package as the spine of the book had come off. This is a purchase for work-related projects and I was really looking forward to it. I would recommend others to get the hardback copy instead to avoid such issues, but limited funds didn't allow me to purchase or repurchase.
The 1 star is for the faulty item, but I would give 5 stars for delivery & seller communication.
Images in this review
Ricardo Wh3
Nice charts but not so relevant for business
Reviewed in Spain 🇪🇸 on October 22, 2015
