Notes from Presenting Data and Information
Notes from Presenting Data and Information, Edward Tufte
The following are notes from a conference presented by Edward Tufte, a leading thinker in effective methods of presentation of data and information. Though these notes are long, I think it pulls some high points from many of his books in a Cliffs notes sort of way.
General data design
“It usually takes a plurality of evidence to understand anything interesting.”
“In a display of information, chart every element in a non-redundant way.” As a negative example, in organizational charts, putting boxes around people’s names and titles is superfluous and takes up space you could use for something else, like adding more contextual detail.
“Every visual move you make, every pixel, and every drop of ink should be used to present non-redundant information”
Every display of information is designed to give us reasons to believe, at least until something better comes along. Every chart and graph should provide documentation and sources, providing some indication of credibility.
In graphic displays of huge amounts of information, the design should be straightforward. Everybody looks at different parts of the display, sees something different, and sees different relationships. The display should be content rich, but with an invisible architecture underneath.
When presenting tables of data, always give the units of measure.
“Every table, chart, and graphic you create should be accompanied by annotation.” When we’re talking and making gestures that is annotation. In writing, add annotation with an efficient design. All it has to do is do its job. We don’t need anything flashy.
Performance data is just a noun followed by numbers. Approach performance data like the newspaper sports section. Everyone can read the sports pages. We don’t need fancy tables or zebra charts. Good models of presenting large amounts of data in tables are the sports and weather sections of the newspaper.
Tables of data are useful if they put the data in context, show comparisons, and are ordered by performance, or quantity, or day. Ordering by performance is interesting. Ordering by alphabetizing is boring. People are good at scanning. And though we’re using the newspaper model as a metaphor, the sports pages, the weather pages, the financials pages, people are good at reading these; 80 million people a day are reading these. “People aren’t suddenly stupid just because they are looking at your data.” Newspapers are good models of presentation in general.
Use models and templates
A useful idea in analytical design: find a good model and copy it. Find an excellent, but conventional model. This will minimize the ‘figuring out time.” We should maximize the content-reasoning time and minimize the figuring-out time. We’re not trying to present a puzzle. We want people to start reasoning about the content right away. Also we might minimize the design-admiration time. So ask yourself, am I maximizing the reasoning time and minimizing the format figuring-out time.
Think of finding good design models as a research project not as a creative act. “Talent imitates, genius steals” – T.S. Elliot. Find a good architecture, a good template, and follow it. With four or five good templates you’re set. For scientific information, look at how articles in Nature present information. This is a solved problem. View design as a solved problem and find that solution.
He showed a video of the dangers of being glitzy. The vis-o-matic.
There are two profound issues in the theory of analytic design.
- Almost everything that is interesting is a multivariable problem (three or more dimensions) that must be presented in two dimensions (on a page or screen). This has been a problem of analytical design in its 6000 year history: since maps were scratched into rocks.
- Information resolution capacity and the scope of performance measurement is increasing. What is the resolution of information? In 1610 we maybe had 11 orders of magnitude, say from tiny visible specks to large views of the sky. Now we have 45 orders of magnitude with incredible data density.
He mentioned the book “Flatland” by Abbott. How do we present three dimensional geometrical objects in a book? Maybe you stick with three dimensions! In an actual copy of
Euclid’s Elements of Geometry (with a preface by John Dee) there are pop-up pyramids! Turning to the back of the book, in the largest type in the whole book, was information about the publisher along with a large picture in profile. It makes the point; always put indication of ‘creatorship’ on your creation. If you’re proud of your work, put your name on it.
Euclid’s Elements has some poor designs, too. The codes and legends, for example, A, B, C, “Refer to Figure 1a,” are impediments to learning. If you’re trying to figure out which line on the plot has two dashes and one dot, you’re not interpreting the data. Get viewers out of the box. Put names on the object, not in a box. Read content directly so you don’t have to figure out the codes. Maximize the content time, minimize the figuring-out time.
How best to represent solid objects? In
Euclid, using pop-ups represents the brute-force method of escaping flatland. The pop-ups in
Euclid still work after 436 years. “Do you think your website will last that long?”
He suggests footnotes should go on the side of a page….”This is where God intended footnotes to go.”
Grand Design Principles
From Beautiful Evidence, several grand design principles include:
- Show comparisons. When we display analytic data we are trying to show comparisons. Our display should help out. Show the context. For example, in the Napoleon’s March chart, the width of the line allows comparison of the number of troops.
- Show causality. For example mechanism and dynamics. Intervention thinking (when trying to figure out a problem) is causality thinking.
- Show multivariate information (three or more variables).
- Completely integrate words, numbers, and images. It’s all information after all. Don’t get all worked up about the differences between words, numbers, and images. Evidence doesn’t care what it is. Don’t segregate information by the mode of production or accidents of software. When we’re presenting information, it is testifying for us. Why should our intellectual architecture be segregated?
- Document everything and tell people about it. Fundamental quality control for credibility: say where the data came from. Build it into the image and words. This is part of the reasons to believe. When placing all this on a graph, this is what six-point type is for. Show care and craft in your materials. And as a consumer of information you should be concerned if someone is not telling you the data source. What are they masking?
Point 5 show the symmetry of these design principles, as a producer you need to use these principles, and as a consumer you should be looking for them. These are principles of thinking that become design principles. This is why they are symmetric. This is principled design. We’re all trying to figure it out. It’s not fashion. The point of information display is to assist thinking. These design principles are universal, and cross-cultural. Why are these cognitive tasks universal? What does a sentient being need in order to understand causality? The world is revealed by differences. What does a brain need to think?
He talked about its history of the map of Napoleon’s March. The map’s creator made the map as an anti-war statement. He was driven by his passion. Not once does it mention Napoleon’s name. It took him twenty years not to see that. This was to memorialize the soldiers, not the surviving celebrity.
6. Presenting data is a content-possessed enterprise. The quality, relevance, and integrity of your design can’t help boring data. Get better numbers if they are boring. Serious presentations stand or fall on content. If you want to make big improvements in presentations get better content. Our presentations should not harm the content. The fundamental problem with PowerPoint is that it harms a presentation. We should have a “Do no harm” statement for presentations.
He had a first edition of a book by Galileo. He said that reading Galileo is a lot like reading Feynman.
7. Detail helps credibility. When making comparisons, do important analysis adjacent in space, not flip-flip-flip, one damn thing after another. Humans can see 150 MB of information at a glance in 20-bit color, not to mention in a rich three dimensional depth of field. A computer screen displays at 1/1000 of human capacity. We process by differences, not absolutes. “We cannot waste any space that’s not content related.” What can you do…other than buy a Mac? Buy a big high-res monitor, or multi-monitors. We need more content real-estate adjacent in space. This is outside-in design; designing for what the user sees. In serious computing every pixel is content; there are less cartoon-like features and more serious content presented.
He showed sunspot pictures from Galileo’s book in motion. “This is the only moving picture from 1612.” From Letters on Sunspots is the line, “The annual movement of the Earth.” This is the only place he mentions the Copernican motion of the Earth in print. These are big words. A lady once asked him for the digital file from the text so she could get a tattoo.
During the break, E.T. spoke loudly at a lady talking on a cell phone, “CAN YOU TAKE THAT CELL PHONE OUT OF THIS ROOM PLEASE!” It didn’t even ring.
Also during the break we were treated to a slide show, showing the architecture and artwork on his property, “where architectural works finally break out of flatland.” It seemed a little eccentric.
From Galileo on, scientific theories are measured by how well theory describes visible evidence. We were liberated from wordy evidence. It’s the visible, empirical evidence that counts.
8. Grand design principle 8. Use small multiples. Put lots of information adjacent and at high resolution. Detail helps credibility, but it’s aided by comparability. Take advantage of the investment of time spent trying to understand your design. Small multiples aid credibility. It says to the audience, “I was watching the system all this time, and this is what I saw. All of it. All the data.” It makes the point that you’re not cherry-picking the data.
“The biggest single threat to learning the truth is cherry-picking of evidence.”
What to do? Don’t give up. Ask, “Am I seeing results of evidence or evidence selection?” What should we do as consumers? Few presenters give data files, but if it’s too good to be true, it probably is. We should stand up and say “It’s more complicated than that.” Real evidence has rough spots. Cherry-picking will be given away over a series of studies. Or, for example in medical trials of new medicines, the most favorable evidence comes in the first study. Then there is “evidence decay.” The early studies are by enthusiasts who “know the truth.” Later studies are more rigorous. It’s not a problem of the data; it’s a problem with data management. In some fields this is a real problem. In science, however, there’s always the ultimate truth of the laws of nature that needs to be satisfied.
What to do as a presenter? Say so. Say the findings have been replicated, we’ve looked at the data, we’ve had independent investigators look at it, and bring in authorities of the past. For small data sets show all the data. That’s very convincing. As consumers be concerned if you don’t see all the data. For big data sets, give a link to all the data. This aids credibility.
He discussed the chart that acquitted John Gotti. Sometimes we are economical with the truth.
What is the best way to best present information? Make a big large multiple display of information and a handout. The nice thing about a handout or a big chart: it lives on, on paper. Always give out handouts.
9. Design principle nine: scale. How to get a sense of the size of an image? We need to use whatever it takes to explain stuff. Moving images help, stills help, small multiples help. We have a pluralism of data and that helps explain things.
A sense of relevance is important to develop. We need ways to manage vast quantities of evidence and sort out quickly the 2% of relevant data, the diamonds in the sewer.
If organizations start lying externally, it often indicates lying within.
Regarding scale and scale transformation, the world is often multiplicative not additive, so log-log plots are often useful.
It’s hard to gauge variation in steepness when everything is steep. So charting near 45% is the best way to see differences in slope. In terms of aspect ratio – we want lumpy not spiky. Choose the vertical scale and use good scale transformations.
He discussed sparklines, or “data words” that are images used inline with text. It seems that the idea sort of started with close reading of Galileo where he had little images of Saturn and its rings
inline with the text. Sparklines provides context for individual data. All the data is presented within the eye span. Sparklines are data graphics that can be rendered with the resolution of current typography. They are graphics-like words. Columns of sparklines e.g. of performance data allow easy comparison. They might not allow five but maybe two significant figures, but “we want to be approximately right rather than exactly wrong.” Graphics can be everywhere.
He has problems with boxes around things in graphics. They are clunky, they activate negative space. Don’t use grids in spreadsheets. Every book has a grid layout but it’s not expressed. There are no grids in the sports section; there are no boxes around city names on maps.
How do you implement sparklines? There are ways to implement within Word, Excel, PowerPoint, etc. All the code is free. The best is sparkmaker for office. To install there are a bunch of macros. There are reviews on his website for different implementations of sparklines.
To make sparklines you need a statistical graphics program like Origins. The new version will have sparklines built in. It creates graphics that are of acceptable quality for a publication like Science. Then the data is taken from the statistics package to Adobe Illustrator.
A good newspaper is 10-12 x the resolution of a computer screen. The only advantage of newsprint is this high resolution vs. the Internet.
A flaw of current office software is that is segregates information by the mode of production. The original GUI concept was not concerned with the OS or applications. Information was not separate but all worked on within one document. It was user oriented, content oriented, and document oriented. We are not benefited by the fallacy of segregation by mode of production. And it’s self-perpetuating. We are forced to segregate numbers from text from images, so we design that way.
9. cont. …back to grand design principle nine. Scale, magnitude….put everything on the universal grid. Show “how big,” and put every piece of information you can within a context. The way people learn and understand information is by the context that surrounds it. From Envisioning Information, “to clarify add data,” or to simplify add information. We gain understanding by context: where it came from and the scale of measure. Everything has a scale. Don’t present any single piece of information. Put it in context. Clutter is not caused by too much information, but due to a failure of design. There is no relationship between the amount of detail presented and how difficult it is to read.
To do important reasoning, space out the information across a large reading area, for example on a wall chart: spaced large, not temporal, or “flip-flip-flip” like a book or a slide show. He related an anecdote of a PowerPoint slide presentation that was ‘taken down’ by a wall chart: at a shuttle RTF (return to flight) presentation the one-slide-at-a-time show was supplanted by one well-designed wall chart, where serious stuff was presented in space not in time.
Presentation and PowerPoint
We did some “Intense Reading of a Serious Slide” from a consumer point of view. This was a slide concerning the eventual destruction of the space shuttle
Columbia. We need to learn three things when viewing a presentation:
- What’s the story? What’s the argument?
- Evaluate the credibility, trust, and confidence. Is this the best we can do?
- Describe the domain. What’s the domain statement of the presented information? What is the information relevant to? Most presentations will over-reach. Nail down the relevant scope of presentation. Exactly what is this presentation relevant to?
The head of NASA said the foam that hit
Colombia’s wing was like the lid of a Styrofoam cooler flying out of the back of a pickup. But at 600 mph that packs as much kinetic energy as a 400 pound safe falling on you at 65 mph. Nasawatch is a web site that keeps an eye on NASA. We looked at the PowerPoint slide at length.
He said, “PowerPoint is mediaeval with its implementation of hierarchy,” in references to bullets and headings and outlines.
“The words on a slide make up something that is “almost” a sentence. It’s better than the average bullet grunt.”
He seems very bitter about the Columbia disaster.
“Significantly is an adverb substituting for a number.”
PowerPoint’s small real estate forces you to pare down sentences and content: “Instead of abbreviating the truth, why don’t you get a better presentation medium?”
Regarding the use of the word “It” on the
Columbia slide: “The difference between ‘It’ and ‘Die’ is very large.”
On the slide they use three different representations of in3. E.g. “cu in.” If the rocket engineers at NASA can’t figure out how to get PowerPoint to do superscripts, it’s a problem with the software.
“When you’re compromising the language of science through your presentation medium, get yourself a better presentation medium.”
“Engineering by PowerPoint”
He mentioned Feynman’s Challenger disaster comment, “For a successful technology, reality must take precedence over public relations, for Nature cannot be fooled.”
“The laws of nature are completely oblivious to what humans think of them.”
What are causes of presentation quality? The quality of the presenter, the quality of the content, and the technical investment are the sources of presentation quality. When PowerPoint leads, stupidity collaborates.
What are the problems with PowerPoint? It’s good for projected art 24×50. You can’t put much on a slide. The formats take up half the space, and that leads to even lower resolution. It’s fine for images but for words and numbers it’s low resolution. PowerPoint has attitude, and so do video games (see p. 158 in Beautiful Evidence). It’s the opposite of a wall chart. It turns information into a sales pitch. In text, behind every paragraph is a pot of data. If a scientific paper has about 1000 numbers per graphic, PowerPoint has only about 12 numbers per graphic.
Conway’s Law – For any organization that designs a system, the design will mimic the organization’s hierarchy.
With PowerPoint we have the insidious replacement of the sentence. “Sentences are much harder to write than PowerPoint grunts, but sentences are much smarter then grunts.”
What to do? Make smarter presentations using smarter tools. Only use PowerPoint as a slide projector, as in a course on art history.
How to give Serious Presentations
From now on your presentation software is Word. It can handle images, text, prose, layout…. In your presentations, give everyone an 11×17 piece of paper folded in half with all of your information contained thereon. This is the information equivalent of 50-250 slides, and it is a document that lives on, leaving traces, that requires you to take responsibility for the content presented. In that report, have a 200 word executive summary saying what the problem is, who cares about the problem, and what your solution is. Make the summary long enough so that you have the important stuff on the table right away. This is harder than creating a deck of slides. But it will be better.
Say you have a five minute opportunity to bring the big boss up to speed. Give him your 11×17 piece of paper and say, “Here, read this.” You can talk 120 words per second, but we can read four times faster. This is a high resolution data dump. Then you’ll get the questions after all the content has been read, not during the slides, and only where they come in the presentation, but when all the information has been digested and people are more up to speed. This is the best way to get people up to speed in a meeting. This will lead to a more discussion-style meeting with a more active audience that is learning more actively.
These ideas pertain to serious presentations, maybe 25-80% of presentations, depending on your setting. Use an 11×17 piece of paper folded in half.
Grand summary of presentation tips
Outside the meeting, preparation:
- Ensure the quality, relevance, integrity, and content…try writing that executive summary first. Simplify the design by adding content.
- Practice, practice, practice. Rehearsal improves performance. Practice in front of a camera. People practically fondle themselves during presentations, and this will cure you of that. Listen to yourself with the images off. This will point out the use of place holders, like ‘like’ and ‘um’ and ‘you know.’ These are the things that start people keeping score during your presentation.
At the presentations:
- Show up early. This is the best piece of advice. You can head of serious problems, you can say hello while people trickle in, and you are in a position to greet people. And the opposite is about as rude as it gets when you show up late to your own presentation.
The presentation begins:
- Tell people the problem, the relevance, and the solution, in about 200 words. A Nature article is a good model.
- In the introduction, never apologize.
- See how long you can go before you say “I” or “We”. Saying “I think” is ok, when you are expressing an opinion.
- Get into explaining a complex graphic as quickly as possible. Go in, then out, then in again: explain some particular detail, then step back to explain the context, the step back and explain another particular detail.
- Give everyone a piece of paper, 11×17 folded in half. Word is now your presentation software.
- Audiences are precious and they deserve to be treated that way. Think the best of them. Stay away from the K.I.S.S. principle. Don’t allow any nasty talk about the audience behind the scenes. In classes, if a teacher thinks the students are idiots, they will fulfill his expectations. These things shouldn’t shut you up – speak frankly to your audience. Usually people recommend ‘know your audience’ as a number one rule. Instead, ‘know your content’ is the number one rule. Respect your audience is number six. Don’t underestimate your audience. Be very careful in characterizations of an audience. It will affect your presentation. Humor? Be very careful. If you have a set piece joke that you can tell on any occasion, use it on no occasion. Don’t alienate anyone. Alienate your audience on the merits of your presentation, not with bad jokes.
- If you believe stuff, let people know that. Your affect, and your non-verbal actions are important. Show intensity, enthusiasm, passion.
At the end:
- Finish early. People will love it, word will spread, and keep them asking for more.
And with that, the conference was over.