Published: 2001 (original 1983)
Publisher: Graphics Press
ISBN: 0961392142

Edward Tufte lays out a very comprehensive study of how to effectively display information in a visual format.  Charts and graphs are a huge part of today's corporate communication medium and yet they are often malformed and fail to convey information properly.  This is a great book for laying down some principles that can apply across technologies and medium and help to formulate high quality information communication.  Due to the nature of this book, my notes here will be a very brief summary of the principles in the book.  I highly recommend reading this book if you are charged with visually displaying information!

Graphical Excellence

Graphical excellence consists of:

  • Well designed presentation of interesting data. 
    • Substance
    • Statistics
    • Design
  • Complex ideas communicated with clarity, precision and efficiency
  • Give the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space
  • Nearly always multivariate displays
  • Tell the truth about the data
Graphical Integrity
  • Physical dimensions should be proportional to the numbers
  • Clear, detailed labels should explain
  • Show data variations and not design variations
  • Deflated & standard monetary measurements should be used
  • # of dimensions depicted should not be more than the data
  • Graphics should not quote data out of context
Data Graphics
  • Above all else show the data
  • Maximize the data-ink ratio
  • Erase "no-data" ink
  • Erase redundant data ink
  • Revise and edit incessantly
Chart Junk
  • Moire vibrations (hatching) is ugly
  • The grid should be muted (or non-existant)
  • The duck (architectural term for a duck shaped building)
General Principles
  • Small, non-comparative, highly labeled data sets usually belong in a table
  • Relational graphs are more sophisticated than time series or maps
Bad ideas?

As a general rule, Tufte's ideas are applicable to data design today.  Some things stand out as bad ideas and some of the data is old but this is a great primer book on the topic.  One bad idea he had was the idea of using range frames.  Rather than extending the edges of charts from zero to some arbitrary number, Tufte argued for limited the frame to just the range of the data.  This leaves the zero (lower left) corner of the chart unanchored.  Some eye tracking studies I have read indicate the eye uses such anchors in order to expedite searching so while removing them might convey some extra little bit of information, they reduce the overall effectiveness of the chart. 

Another thought he had was moving the boxes from the classic box plot.  The problem I see with this is we are left using empty space to indicate the range.  If the range of the box plot is the important piece of data, empty space is not the best way to convey that.  Empty space can be used to effectively accentuate the important data but shouldn't be used to indicate the important data. 

Conclusion

This book is a must-read for anyone interested in HCI or data design.  The principles are effective and useful in a great number of areas.  Designers and programmers alike should read this book for a better understanding of the best way to communicate data to the user.