Beautiful Evidence [Tapa dura]

I finished tufte last night... what a disaster, or perhaps sunk with high expectations.

I'm a huge fan of dr. tufte's very influential writing on information visualization - as far as I know he's done the best work in the field. But this book - while simply physically and visually stunning - is a real disappointment.

In this work I read about 20% insight, 40% recycled material and preaching to what is probably the choir (this includes an overly repetitious chapter-long discussion of minard's lovely march to moscow graphic & his previously available power point piece), and 40% filler & drek. I don't find his comments on art, writing styles, baseball, and the like to be terribly compelling, and are certainly done better in many other works - and indeed, his thoughts on these ended up as being pretty grating and condescending, if not just wrong.

And that the book ends with several pages of photos (a few of really poor quality, I might add) his own outdoor artwork (which are of passable quality, but what the *bleep* does this have to do with evidence as defined at the front of the book?) only throws salt on the wounds.

This thing is maddeningly inconsistent. I wish I could simply dismiss the work, but it's full of beauty and joy as well as the bad. Sparklines are fun, but could be improved on. Words + images combined inline, some great stuff there. But while some of the really lovely things, like the translations of galileo, are wonderful and exciting to any science-loving person, they really are pretty pointless to the conversation at hand. He has gone straight down since his first major book - a 5+ star effort, the 2nd, 4.5-5 stars, 3rd, 3 stars, and this is about a 2 star one (2.5+ if you haven't read the others.)

If he'd stop believing his sycophants and stop taking himself so seriously in his quest to convince the reader that he's a high priest on a moral crusade it'd be wonderful. He really does try to convince the reader that this topic is of high moral concern - not just sometimes, but in general. I don't buy it.

And you shouldn't buy this if you haven't read his other works (although if you haven't I'll admit you'll probably like this, you just don't know any better ;-)). Read the staggeringly good "The Visual Display of Quantitative Information" or the wonderful "Envisioning Information". And if you must read this, soak up the good points, and try not to grind your teeth with the rest.

Out of all of the great ideas that are in this book, I am going to concentrate on the ones that relate to "what can be done with high-resolution display devices," such as 1200 dpi printers. An increasing amount of contemporary design is done for low-resolution displays, such as television and computer monitors. If we get a 1200 dpi version of one of these designs, as is easily possible with an inexpensive laser printer, we are not getting much benefit from that increased resolution. A lot of the ideas in Beautiful Evidence can be used today with Web scripts that generate PDF files to be printed. The rest of the ideas will be waiting for designers 20 years from now when computer monitors finally catch up to paper.

dea 1: Sparklines (there are examples on the author's Web site). Tufte points out that nothing stops the modern printer from including small graphs right in-line with text or tables and that these graphs make comparisons much easier. Baseball fans will enjoy Tufte's depiction of a baseball season, first for one team and then for all teams. Tufte argues convincingly that showing history in a "sparkline" reduces "recency bias, the persistent and widespread over-weighting of recent events in making decisions."

Idea 2: Forcing people to write English sentences instead of PowerPoint bullets results in a lot more clarity, especially with respect to causality.

Idea 3: If you're running a business, figure out how to pack a huge amount of information, including sparklines, onto a single 11×17″ sheet of paper and print it out on a laserprinter, then give it to decision makers. With that one sheet of paper, they will have as much information as 15 computer screenfuls or 300 PowerPoint slides.

A thought-provoking book that will reward repeat scrutiny.

This is the fourth of Edward Tufte's books on the graphical display of information, and one might fear that he might be stretching the point too far and running out of ideas. One would be wrong, however, because this is a wonderful book, and is possibly the best of the four. It is a must-have, must-read, must-understand, must-apply sort of book. No one who is seriously interested in preparing illustrations for conveying information can afford to be unfamiliar with Tufte's ideas.

Inevitably there is some overlap with the earlier books, but this is deliberate policy, not carelessness. As Tufte makes clear, it is better to repeat information than to expect readers to hunt for it somewhere else. Many potentially useful books have been rendered much more difficult to use than they ought to be, at worst by gathering together the artwork in one place, far away from the text that it relates to, or, slightly less bad, by failing to ensure that it appears on the same double-page spread as its accompanying text. Tufte doesn't even believe in referring to tables and figures by numbers, because he considers that any illustration can just be introduced with "here" or "in this example", etc., if it is properly placed. This is what he practises himself, but the technical demands of commercial publishers will make it difficult advice to follow, unfortunately. However, with modern computer-based publishing it ought to become easy in the future if enough pressure is put on publishers. If Galileo could integrate all of his diagrams into his text, why can we not do that now, with far more technical aids at our disposal than were available to him?

The main new idea that appears in Beautiful Evidence is the description of sparklines: small, data-intense, word-like graphics -- word-like in the sense that a sparkline can appear right in the middle of a sentence, but can contain the equivalent of hundreds of numbers. Sparklines are ideal for conveying time series, such as a series of blood-glucose measurements for a diabetes patient. With suitable shading they can indicately instantly whether the measurements fall within the normal ranges.

Tufte's short pamphlet about the presentation software PowerPoint, previously available as a separate publication, now appears as a chapter in Beautiful Evidence. His main points are that PowerPoint slides are typically so low in information-content that they insult the audiences they are directed towards, and that bulleted lists of slogans are just a pretence at supplying real arguments.

Charles Joseph Minard's map of Napoleon's invasion of Russia already played a prominent role in the first book in the series, The Visual Display of Quantitative Information, and it reappears here, with a whole chapter devoted to analysing it. This is space well used, because to emulate Minard it is essential to go beyond a casual appreciation of his work as excellent; it demands a careful analysis of what it is that makes it excellent.