Notes – Edward Tufte’s “Presenting Data and Information” (@EdwardTufte, #display, #data, #visualization, #design, #statistics)

Edward Tufte is a Yale-connected academic who conducts several private seminars around the country each year promoting his view of visual design for the display of quantitative information and statistics. He has published multiple books through his own publishing mark such as The Visual Display of Quantitative Information, Beautiful Evidence, Envisioning Information and Visual Explanations. His personal website which contains articles, research papers, examples of his work and principles and other information is at EdwardTufte.com. A friend who is a fan has blogged some notes about the man and his seminar courses.

If I were to summarize Tufte’s philosophy of information design into a single sentence, which is certainly a crude way to approach the nuanced and thoughtful lifework of a person, I’d say this– beautiful design means creating the highest density information display the resolution of your medium will allow. This stands in stark contrast to the reigning paradigm of “less is more” and sacrificing much of the available real estate of an information display for white/empty space, navigational or UI elements and “inference assists” (my term) such as arrows, boxes and non-data lines which are supposed to draw the viewer’s attention to what’s important or where to focus their eyes.

In Tufte’s own words, he summarized the philosophy with the pithy, “The information is the interface; maximize content reasoning time; minimize design decoding time.” [Note: on my hand-written notes written in a darkened room early in the morning at the start of his seminar, I think I mistakenly wrote “maximize design decoding time” but meant to write “minimize”.] Even more pithy, and in Tufte’s own words:

The purpose of information display is to assist thinking about its content.

I attended his seminar in San Diego, CA in February. Below I am posting my notes which may or may not be useful to a person unfamiliar with his work or the content of the seminar. Tufte, who introduces himself as “ET”, likes to circulate amongst his audience before, during and after the session and introduce himself– clearly he enjoys what he does and appreciates the people who have taken an interest in his work which is a good professional example for others to follow.

  • The information is the interface.
  • Idea: maximize content reasoning time; minimize [maximize? see above] design decoding time
    • design decode effort/time is wasteful as the pattern for design is often not repeated in the future
  • Graphics are only useful when there is a lot of data, not a little bit of data
  • Design should encourage scanning, scrolling and choosing
  • Increases in resolution allows for spatial adjacency [note: this is the idea of putting lots of information side-by-side versus having to change mediums, windows, displays, repeatedly to compare and contrast blocks of information]
  • Digital display screen resolution is finally approaching “P.A.P.E.R. technology” (paper) resolution
  • Simple, clear conventional design with rich, complex data is preferable to complicated design devoid of content; many designers invest too much effort in display relative to content quality
  • NYT, WSJ are highly trafficked websites high in information density (many links, many pieces of data and text) which demonstrate this approach is desirable for design of corporate sites
  • Names have reputations, put your name on your work
  • Reasoning on a flat surface means all viewers can go at their own pace; a slideshow makes most people wait; think “documents” not “decks”
  • Listing sources for data provides credibility and reasons to believe
  • Look at sources, start points, end points, rates of change, to examine whether a chart establishes a relationship between evidence and conclusion
  • Annotations help explain all data by providing specific information about one data point captured in the graphic
  • Look at “excellence in the wild” to contrast your own efforts against the pros
    • Use Word, not PowerPoint
    • Be web-based
  • Order data tables by performance, not by alphabet; performance often tells a story
  • ESPN.com demonstrates that even complex data can be appreciated by lesser intellects (!)
  • Dashboards are idiotic and no way to operate a business or institution
  • How to Make A Presentation, some tips:
    • show up early (head off problems, ensure equipment works, room not double-booked, etc.)
    • talk to people
    • give them a document for discussion; don’t give it in advance of the meeting, no homework
    • begin meetings with study hall, people can read faster than you talk
    • the document addresses the principles of individualism and personalization as people can take what information from it they deem important
    • PPT disappears as you go higher up an org chart, the top execs have no time for the “long and winding road” (Steve Ballmer anecdote); submit ideas to discuss as written documents
    • provide intellectual leadership about content, stop discussing production methodology
    • finish early, your audience will thank you
    • Remember “Problem, Relevance, Solution”, three necessary components of any good presentation
  • How does Jeff Bezos run a meeting? Read the Forbes article or watch this Charlie Rose interview:
  • Applied presentation tip– provide notes/documents of medical concerns for a doctor to read during your doctor visit; this is what they’re trained to do and they’ll pay more attention to the information if you give them something to read
  • Check out The Public Library of Science and its templates for ideas on content rich documents
  • You can copy the source code from EdwardTufte.com and use the CSS to apply style ideas to your own blog or website
  • Real reading entails looting and hacking the valuable materials useful for later efforts, liberating them from the text; always read with an awareness for context (what came before this, what comes after, why did the author write it?); this echoes the idea of “making the work your own” of Mortimer Adler
  • Refer to “Beautiful Evidence”, pg. 78-79, using diagram trees appropriately (annotated linking lines)
    • links need to convey causality and action
    • replace generic lines with words and numbers– annotate!
  • Turn fundamental principles of analytical thinking into design decisions
  • The purpose of information display is to assist thinking about its content
  • Don’t pre-specify a data display method, use whatever method the job requires
  • Look at Google Maps and ask IT why you can’t achieve similar design capabilities; their maps are rich, colorful, multi-dimensional, varied fonts and orientation of information, etc.
  • Refer to “Visual Explanations”, pg. 90-91
  • Refer to “Beautiful Evidence”, pgs. 82-83, 114-115; exploring words, numbers and images together
  • Today’s computer interfaces separate and segregate information based on the method of production
  • Statistical graphics can be anywhere a number or letter can be
  • Statistical graphics can have the same resolution as topography
  • Refer to “Beautiful Evidence”, pg. 46-47, “sparklines” method for creating text-sized data graphics, embedded within text (inspired by Galileo’s revelation of Saturn)
  • “Nature” magazine has some of the best data-driven graphical displays, good place to look for examples of the possible
  • Why aren’t all data displays excellent? Tufte suggests there is a profit-driven bias and the dominance of Microsoft combined with the lack of scientific rigor of many data designers results in a failure of the “public spirit” principle; color me skeptical about profit and “public spirit” being at odds!
  • Excel, Google Analytics can both produce sparklines
  • Refer to “Beautiful Evidence”, pg. 58, for the famed Swiss mountain maps, or see this video (YouTube):
  • The human eye-brain optic system operates at 20mb/s in 16-bit color, digital displays don’t come close to this much data and resolution
  • Content and credibility are the keys to presenting and spectatorship
    • have the sources been credible in the past?
    • demonstrate your understanding of detail and mastery of verbs, not nouns (not who is who, but who does what to whom?)
    • threats to credibility: lying, cherry-picking (evidence vs. evidence selection), over-polished, hidden or absent sources (“proprietary”, “legal liability”, “violate federal law”, etc.)
  • Know your content, not your audience; maintain respect for your audience
    • “know your audience” leads to pandering
    • use presentations as a teaching moment to inform people of your content
  • Scan lots of material and drill down where you see discrepancies for superior economization on large volumes of data to achieve relevance
  • Investigate how data was measured; go out, walk around, see the process producing the data
    • people can not keep their own score; the metric is gamed as soon as it becomes important
    • eg, Google words are gamed by SEO, so use Google Images to search
  • Refer to “Beautiful Evidence”, pg. 32-33 for “small multiples” concept; use the need to learn a repetitious format to get people to focus on the content
  • Universality and “forever ideas”; Galileo was the supreme data designer; why should the “best thing ever” have occurred recently versus long ago?
  • Personal curiosity– why are US internet pipelines significantly slower than other developed nations?
  • Spatial adjacency versus temporal stacking (hi-res vs. low-res)
  • Different modes of display are not competitors, they are co-operators in communicating information; no one display is optimal