The Visual Display of Quantitative Information by Edward R. Tufte
Posted on January 31st, 2006
Making charts and graphs, more formally known as "data graphics," is hard. Tools like Excel can make it possible, but often at the cost of aesthetic appeal. As with web page design, you can do things in such a way that you look like everyone else and can be spotted a mile away as such (Look-- a WordPress blog! Look-- a TypePad template!), or you can try to stake a claim for appealing design. If you're going to do the latter with your data graphics, you'd want to read Visual Display to better understand all the potential ways there are to screw things up.
The book is filled with graphics that illustrate Tufte's discussion of the earliest graphs on through to their modern equivalents, both the ones that suffer heavy distortion from bad design choices to the ones that can be simple and effective once they stop trying so hard.
Pretty pictures are nice and all, but the real cinchers of the book are the practical guidelines. In particular, Tufte offers 5 core rules:
Five principles in the theory of data graphics produce substantial changes in graphical design. The principles apply to many graphics and yield a series of design options through cycles of graphical revision and editing.
Above all else show the data
Maximize the data-ink ratio.
Erase non-data-ink.
Erase redundant data-ink.
Revise and edit.
Pretty common sense. I found the suggestions on graph shape much more provocative:
Graphics should tend toward the horizontal, greater in length than height... Several lines of reasoning favor horizontal over vertical displays.
First, analogy to the horizon. Our eye is naturally practiced in detecting deviations from the horizon, and graphic design should take advantage of this fact. Horizontally stretched time-series are more accessible to the eye... The analogy to the horizon also suggests that a shaded, high contrast display might occasionally be better than the floating snake. The shading should be calm, without moiré effects...
Second, ease of labeling. It is easier to write and to rad words that read from left to right on a horizontally stretched plotting field...
Third, emphasis on causal influence... a longer horizontal helps to elaborate the workings of the causal variable in more detail
Naturally, these guidelines are much better experienced on the printed page than the screen.