Presenting Data and Information
David Stack's notes from a seminar by Edward R. Tufte
The presenter is a guru on presenting data and information in clear, accurate and understandable formats in all kinds of media. He has cult followings among graphic artists, computists, statisticians and others.
The presenter has self-published three books that have each taken seven years to write.
The presenter can be contacted and publications ordered from:
He prefers to reserve his email address for spam.
Or order his books from amazon.com.
Here is an article about him and his work from SALON magazine.
The following web sites were recommended by the presenter:
The Two Essential Problems in the Display of Information
Escape from Flatland
The first "escape from flatland" was the English translation of Euclid's Elements of Geometry that was published over 400 years ago. The letter code of geometry (remember triangle ABC from high school) is an impediment to learning. It takes 42 roundt rips from the text to the diagrams to prove the Pythagorean Theorem. The Chinese did it with a single diagram and the one word proof, "Behold!".
In 1570 a pop-up book of three-dimensional geometry was published. The paper can still be lifted today because it is acid free and pH neutral. It has pyramid ABCF. This is known as the brute force method of escaping flatland, build a model. Page 16 of the author's Envisioning Information has such a three dimensional pyramid in it.
Another way to escape flatland is to be real smart about design. Look at the graph of Napoleon's March to Moscow during the war of 1812 by Charles Joseph Minard. The tan line from left to right that is overlaid on a map is the path of the march, and the width of the line shows the number of remaining soldiers. The black line returning from right to left shows the retreating troops. A half a million soldiers were withered to 10,000 survivors over the campaign and retreat.
The Grand Principles of Information Display
The First Grand Principle: Enforce Wise Visual Comparisons., i.e., force answers to the question "Compared with What?"
The Second Grand Principle: Show Causality. We are looking at information to understand mechanisms. Policy reasoning is about examining causality. Napoleon was defeated by the winter, not the opposing army, as shown by the temperature scale o n the bottom of Minard's graph,
The Third Grand Principle: The World We Seek to Understand is Multivariate, as Our Displays Should Be. The Minard graph has six dimensions: size of the army, the two dimensional route of the march, the direction of the march, the temperatures and the dates.
The Fourth Grand Principle: Completely Integrate Words, Numbers and Images. Don't let the accidents of the modes of production break up the text, images and data. Just because the artists, technical writers and database people work in differen t buildings doesn't mean reports should be disjoint with text, graphs and images in different boxes or on different pages.
The Fifth (most important) Grand Principle: Most of What Happens in Design Depends upon the Quality, Relevance and Integrity of the Content. Minard's graphic was made as an anti-war poster. To improve a presentation, get better content. If yo ur numbers are boring you have the wrong numbers. Design won't help, it is too late.
Page 18 of Envisioning Information shows a book by Galileo published in 1613 which reported the discovery of sunspots and the rings of Saturn for the first time. He wrote in Italian, not Latin, because he wanted to reach a wider audience than the scie ntific elite. His tone of writing is wide eyed, straight-forward, undiplomatic, sardonic and sounds a lot like the modern voice of Richard Feynman. The report of the discovery of sunspots has a simple drawing of the sun on each page to show daily obser vations. From these observations he learned that the sun was rotating as the spots moved across the page and changed apparent shape at the edges due to foreshortening. It is easy to make comparisons between the left hand and right hand pages because the y are within the eye span.
The Sixth Grand Principle: Information for Comparison Should be Put Side by Side., i.e., within the eye span, not stacked in time on subsequent pages, which is known as 'one damn thing after another', and also known as the computer interface. T he computer interface is a low-resolution display device compared to paper, so we have a relentless sequentiality. The most common user question after a sequence of computer operations is "Where am I?" The lesson: get the biggest monitor of the highest resolution that you possibly can.
One of Tufte's students scanned Galileo's images and animated them so the sun of 1612 could be seen to rotate. At a couple points in the annimation the images skip forward because there was missing data due to clouds, or Galileo taking a day off.
A Jesuit rival of Galileo republished the sunspot data (see p17 of Envisioning Information). He used the single most effective tool of information design, the small multiple, which puts all 38 images within the eye span.
The Seventh Grand Principle: Use Small Multiples. They are high resolution and easy on the viewer, because once the viewer figures out one frame, they can figure out all the rest based upon what they have learned. It escapes flatland by introd ucing time as a variable. They have an inherent credibility with the viewer because they show a lot of data, i.e., "I know what I'm talking about and I'm showing all my data to you." Keep the underlying design of small multiples simple and clear.
A good example of small multiples is on pp 30-31 of Envisioning Information which shows the NY Times report of the acquittal of John Gotti based upon a chart introduced by the defense. It was done with Lotus 1.2.3, but Lotus has never used it in their advertisements.
On pp 120-121 of Envisioning Information it shows how Galileo reported the first telescopic observations of the rings of Saturn. Modern texts would report this discovery by saying, "See fig 17b", or a modern designer would take the image, put it in Ph otoshop and wrap it around the book jacket." Galileo says, "It looks like this", and there is the illustration right in the text. We'll be able to do this real soon now with digital publishing. Saturn functions as a drawing, a word, a noun.
The Eighth Grand Principle: Don't Dequantify. Numbers have meaning. Use numbers or a graph that represents them. Don't reduce quantities to on or off, yes or no, here or not.
These Grand Principles are universal. Everything in this seminar is directly relevant to the first map scratched in stone 6,000 years ago and, heaven forbid, web pages. Sometimes the best that information design can do is not screw up good content.
The meta-principle over all of these Grand Principles tries to answer the question about how we derive principles of information display. The meta-principle is: Thinking and designing are as one. If the cognitive task is to make comparisons, t hen our display should do the same. The principles of information design are the principles of reasoning about evidence. It is visual thinking. Good design is a lot like clear thinking, made visible.
The converse is also true. Bad design is stupidity made visible. If a chart has three phony dimensions to compare four numbers it shows the person doesn't know what they are talking about. It doesn't just show that they used Excel.
Start by asking, what is the intellectual task that this display is supposed to help with?
A Tour of the book Visual Explanations
This book is largely about how we look at quantity. We tend to run with the latest technology and techniques even when they are inappropriate. Paintings from the 1500's have tile floors in stables, the wilderness, etc because drawing perspective was a new technique and everyone had to show it off.
NASA exaggerates the slopes of mountains on Venus, which are only 3 degrees, to make the images more exciting. They multiply the vertical dimension more than 22 times. These images now appear in astronomy textbooks claiming that this is what Venus lo oks like. Wrong. If you're going to stretch an image for a legitimate reason you've got to let the viewer know.
Magicians make their living making presentations to people. Magic takes place in the three dimensional space in which we live, over time and has a dimension of revealed/concealed; a total of five dimensions. Magic is about dis-information design. Mag icians never explain in advance what they are about to do, because the audience might begin to understand it. We should do just the opposite to assist understanding. Magicians never repeat a trick; we should use repetition wisely through shrewd, helpful redundancy.
Whenever you make a visual move make it as small and modest as possible, but still absolutely clear; just enough to make the point. The smallest effective difference. Don't overdo the effects of arrows, grids, annotations, etc on the design. When us ing shaded contour lines make the changes subtle, not from light to very, very dark. Don't use rainbow color schemes where a little is blue, medium is yellow and red is a lot (or whatever). You get abrupt changes of color for small changes of informatio n and its unintelligible, even by referring to a legend.
Showing Financial Data, see The Visual Display of Quantitative Information
Most financial graphs are about showing an assessment of change.
You probably won't be able to create financial report templates from a statistics or spreadsheet program, use a program that can see, not just count, like Adobe Illustrator. It has a spreadsheet input tool. Implement templates for all your standard r eports in a program like Adobe Illustrator.
The Design of Interfaces, Web Sites, Information Kiosks and Presentations, see Visual Explanations, p 146ff
Definition: BOGSAAT design process, a bunch of guys sitting around a table.
Tufte's heart sinks when he hears a BOGSAAT say that they need a metaphor for a design, like the desktop for the Mac, or an open book displayed upon a screen.
Design metaphor too often replicates hierarchy. Look at Newsweek magazine. It has big red rules that show the staff where their editorial turf is in the magazine, its not there to help the reader; they already know what they are reading.
Design metaphor too often replicates computer programming. The complex nested nature of programming is reflected in most menu-driven interfaces. You're forced to decide if you like choice number 3 or NOT.
People want flat interfaces, not hierarchical or branching interfaces. (Note the Alpha EZ Menu - ds). Excite offers people 154 choices on their opening page. People want as much up front as possible to see the breadth of the domain. People can see what they want to find, as well as other things that may also interest them. Hierarchical interfaces do not promote discovery.
The most common user activity at web sites is fleeing; leaving the web site. There are corporate web sites whose average download time exceeds the length of the average visit, which means that people are leaving before the corporate logo can load.
There is evidence that people bail out of web sites as they are asked to drill down into it. Each deeper level loses an order of magnitude of viewers. Start with a hundred viewers and four levels down you've only got one tenth of a viewer.
The only thing that makes your web site unique is the content. That's what makes people will come and stay. Everyone has seen a spinning globe and fancy buttons and mailboxes that open and close. If your web site looks like a web site you're in trou ble. It should look like a content site. The Bloomberg web site even says "web site" on the opening screen! What else would it be?
Most information kiosks are rather lonely; no one uses them.
The most important issue in the design of computer interfaces is the allocation of screen real estate. Less than 10 percent of the screen should be for computer administrivia overhead, operating system imperialism and marketeering, otherwise you have technology in contempt of the content. The screen is computing's most precious resource. Pay attention to how it is allocated. Screen allocation usually has more to do with organizational power struggles than the needs of the user.
Computer screens are about 1/15 the resolution of good paper images, such as nice topographic maps. It would take about 15 computer screens to render a page sized topographic map in similar detail. If you also allow for operating system imperialism i t might take 20 screens.
There are 18,000 characters per page in a phone book, that is 350 characters per square inch. By comparison you're lucky to have 3,000 characters on an entire computers screen, that's 5 - 8% the character density of best selling non-fiction titles, su ch as the phone book and Michelin Guide.
Overhead projectors are worse than computer screens. People show you their best side as they read aloud four trapezoidally misshapen words. It is even worse when the dreaded overhead projector is combined with the striptease act, where the four words on the misshapen trapezoid are slowly unveiled one at a time.
There is something even worse than an overhead projector. It is the graphics that were displayed in Pravda during the cultural resolution. They had a resolution of about half a character per square inch. They were about information denial and censor ship, and their resolution level ranked right there alongside the overhead projector.
The Visual Display of Critical Information, see Visual Explanations, p 110ff
A hospital patient's medical chart is a write only memory. Its purpose is for CYA, billing and legal purposes. No one wants to read it, especially the interns. They carry around 3 by 5 cards so they don't have to open the chart. The first question in a medical conference is usually about the data, when did such and such happen, not about the care. An alternative presentation for a patient chart is described on page 111.
Lessons on Evaluating Presentations of Data
Pp 38-39 of Visual Explanations talk about the space shuttle Challenger disaster. Shuttle booster rockets are recycled and their casings re-used. A half a second after ignition a hole opened in the side of one of the solid fuel booster rockets. At t hat point all was probably lost. The hole may have plugged itself for a minute, but then it opened again and the flame grew larger and ignited the big liquid fuel tank. The boosters are built in Utah for political reasons. They are built in segments be cause the completed rockets are too big to ship to Florida. Two rubber O rings are at the field joints between the two sections. The O rings move in flight and didn't adjust to the stresses of launch quickly enough in the Challenger disaster.
The very cause of the failure was debated for 11 hours BEFORE the blast off by NASA and the manufacturer. The engineers who designed the rocket voted unanimously not to launch because they felt the cold would make the O rings too stiff to move and adj ust. In 24 previous launches there were O ring problems in six, which were found when they recycled the rockets. The launch day was 27 degrees. Shuttle launches are usually launch at 60-70 degrees and the coldest they had ever launched was 53 degrees. The manufacturer made the first no launch recommendation in 11 years. NASA called the manufacturer wimps.
The manufacturer prepared a presentation with 13 exhibits as to why the launch should be called off (see page 40). The presentation was in deep trouble from the title chart, because there was no one's name on it. People do things, not companies. Peo ple have credibility to other people, even when companies don't always have credibility.
The question was, were the previous failures on cold days or warm days? The table they showed had a lot of data, but it broke up the damage data into small, stupefying fragments. There was no summary measure of the total amount of damage on each laun ch.
The handwritten chart of previous temperature experience left out most of the launches and included horizontal test motors fired in Utah. Only two of the seven temperature data points were from actual launches. There was no chart that included both d amage and temperature, just charts that showed one or the other.
Lesson 1. When somebody makes and arguement about cause and effect, ask them to show the alleged causality.
Lesson 2. When someone gives a presentation with fragments of data, ask tthem to show you the full data set. If they say its too much data, ask them to print it in six point type.
NASA ripped the presentation apart and pointed out that it did not show causality. NASA found some damage on a warm day that they used to counteract the report of damage on a cold day. A war story can always be cancelled by a counter war story.
If the data had been ordered by temperature it would have also been ordered by damage, which would have stopped the launch. There was always damage when launches happened below 65 degrees. The launch at 53 degrees barely survived.
The Challenger launch temperature was six standard deviations outside the range for which they had actual engineering data. A statistician wouldn't make such an extrapolation, why should a range safety officer?
None of the 38 people who looked at the data asked for the full database or asked for cause and effect to be brought together in a graph.
Lesson 3. When someone presents you with a display in an arguement, look out the window and ask yourself what you'd really like to see to be convinced. Ask for that. Think of it as an evidence selection problem. Imagine that chart is just one selected out of a pool of thousands. Ask to rummage through that pile.
How to Make Presentations, see pp 68-71 in Visual Explanations
Most of the information on making presentations is cribbed from books on how to teach. Many of them were not very much help. They said to be a better teacher you needed to work harder or have more charisma.
Show up early. Something good is going to happen. You can head off problems, like the room being double-booked. You can adjust the lights, etc. It gives a perfect time to say hello to people and advance your cause.
Early in the talk tell them what the problem is, why it is important and what your solution is. This will give people a framework for absorbing and judging your presentation.
An alternative is the stumblebum technique. Put an obvious error in the start of the presentation to get people's attention and get them watching intently for a reappearance throughout the rest of the presentation. Only use this carefully when your p resentation is otherwise flawless.
Never apologize at the beginning of a talk for your health, your equipment, etc. Talk about the content. Stay out of the first person singular if you can.
Particular-General-Particular is a technique for looking at complex information. Point out a bit of detail, pull back and take a global view, then go back in to another bit of detail.
Law: give everybody in your audience at least one piece of paper. It is the highest resolution device at your disposal. It will be a symbol that enforces your credibility. It is a document, testimony.
Think about your audience in terms of what they read. People haven't gotten stupider just because they came to hear you talk. Give them the same type of displays and language they are used to.
We can do better than the overhead projector. Too loud, too stereotyped in the corporate world. PowerPoint isn't any better. The bullet list promotes generic thinking. The narrative between the bullets carries the content. Witness the twelve-sec ond mission statement generator on the Dilbert web site.
Audiences are deserving of our respect and we should act that way. Never diminish the audience. Assume that people care as much as you do about the material and are as smart as you are. At least start from this perspective. None of this KISS stu ff. Don't let this advice get twisted into censorship. You can respect your audience and say what you think.
Humor makes things memorable, reinforces points and allows for exaggeration and hyperbole. If you have a joke you can use on any occasion, use it on no occasion. Use jokes only to reinforce points. Alienate your audiences on the merit of your content, not your jokes.
A way to alienate your audience gratuitously is to use masculine pronouns universally. "He moves the cursor to his menu bar and makes his manly choice." It is perfectly proper in speech to mix singulars and plurals. "They can then move the cursor to menu bar and make their choice."
People will judge your entire talk on your reaction to their tiny question or your brief interaction with them. If you ask for questions, wait at least 10 seconds for one to be offered. If you anticipate a shy audience, plant a question in the aud ience. Lay out ground rules about questions at the beginning to gain consensus from the audience.
If you believe in the material let people know it. Don't stand behind the podium clutching the podium. Let people see your gestures and enthusiasm. If they know you already the fact that you believe in the material may sway them.
Finish early. Something good is bound to happen. At a minimum people will be thrilled. How many times have you wished a presentation had gone on for another 25 minutes?
Practice, practice, practice. You will get better as you practice. Practice in front of a video camera for a real thrill, its ruthless. Watch the tape at normal speed and high speed to see repetitive quirks show up. Turn the picture off and liste n to just the sound and listen for ah, like, actually, basically, you known what I mean, etc. If you do these things too many times people will start keeping count. Don't end every sentence with an upward inflection like a question. Rehearsals are pain ful. Don't be discouraged by the difficulty and pain of rehearsing. Because you've practiced you can walk and make eye contact with the audience and assess what is going on it the participant's minds.
The two most dehydrating things in the world are making presentations and taking airplane rides, and the two often go together. Drink lots of water. You may be able to get out of bed the day after you return. Avoid alcohol in planes.
Most of what happens in a presentation depends on the content. To improve a presentation, improve the content. None of the above will save a presentation with bad content.