Notes on Edward Tufte's Visual Display of Quantitative Information
The Visual Display of Quantitative Information is one of those classics I've been meaning to read for years. I was worried that I had waited too long and my expectations were too high. But the book is shockingly brilliant and left me brimming.
Edward Tufte's argument in short is that data graphs should convey large sets of information efficiently and accurately. The design, structure, and every pixel used should all be in service of this goal.
Often they are not:
Shapes and shading distort the proportional differences in data
Extra components make it harder to read the core data
Design obscures data
Design variation doesn't reflect data variation
Unfortunately the standard tools that have made it so easy to produce graphs (Excel, etc) embed many of these bad habits.
—
The book is a love letter of sorts to the incredible power of graphs. They are vastly more efficient than text and tables for conveying large sets of information. And the best enable the eye to pick out complex patterns that would be laborious and confusing to convey otherwise.
Two of the many great graphics in the book are a graphical depiction of the advance and retreat of Napoleon's army in Russia, and a map of the count and location of galaxies in the universe.
One fascinating nugget is that while maps have been around a long time in human history, graphs are rather recent. The first data maps came about in the 17th century and the time series chart was invented more or less as it is today by Scottish economist William Playfair. Since then we've added some visual tricks and come up with a few new designs (box plot), but Playfair covered the majority of what I come into contact with.
The book is a love letter of sorts to the incredible power of graphs. They are vastly more efficient than text and tables for conveying large sets of information. And the best enable the eye to pick out complex patterns that would be laborious and confusing to convey otherwise.
Two of the many great graphics in the book are a graphical depiction of the advance and retreat of Napoleon's army in Russia, and a map of the count and location of galaxies in the universe.
While maps have been around a long time in human history, graphs are rather recent. The first data maps came about in the 17th century and the time series chart was invented more or less as it is today by Scottish economist William Playfair. Since then we've added some visual tricks and come up with a few new designs (box plot), but Playfair defined a surprising amount of what we still use today .
Tufte sets out some standards out for graphical excellence and universal principles:
Reduce the Lie factor = size of effect shown in graph / size of effect in data
Increase the Data ink ratio: data-ink / total ink
Spend less ink/pixels on the axes, gridlines, labels. Ink should be focused on the data
Reduce redundancy
Graph range should always be the actual data max-min range
Reduce non-data ink. A devastating example: Take a typical bar chart. Erase the box, shift to white grid from black. There's a lot less ink and a clearer data story.
Now I've noticing a lot of extra pixels and extraneous labeling in the user dashboard I'm working on. It’s been a good shock to see the problems in some of my design habits. But I will say Highcharts comes out looking pretty good. The time series defaults have a pretty good data-ink ratio.
What I loved about this book more than the ideas was Tufte's realization that a great argument about design must itself be beautiful. He insisted on self-publishing, taking out a second mortgage so he could include the legions of graphs to make his case properly. "Otherwise, why do it?"