A day with Edward Tufte

Edward Tufte is the author of several books on the presentation and visualization of information and data, as well as the presentation of evidence. He is a professor at Yale and apparently does quite a bit of touring doing lectures like the one that I attended this past Friday.

My overall opinion of the lecture was that it was excellent, extremely informative, and certainly time well-spent. I felt a surge of inspiration and a new mantra that I would no longer accept anything but the best presentation of information and data within my own applications. I'm sure that was the goal of his lecture, and so in that regard it was a complete success. There is a lot of really good content in his lecture, but there's also a lot of self-inflated egotism as well, as you'll see in the discussion below.

On Evidence

I totally agree with Tufte here. Visual evidence is serious, not something to be ruined by bullsh*t charts and meaningless tables and graphs and fancy clipart. If you are showing something, it had better mean something. People should be able to take a look at your evidence and reason with it. It should provide some meaning, and value. Tufte presents his grand principles of analytical design (there might be a 10th, he took his sweet time delivering these):

1. Show comparisons
2. Show causality
3. Show multivariate data
4. Integrate word, number, and image
5. Document everything. Cite sources, scale of measurement. Everything.
6. Presentations stand or fall based on the quality, integrity, and relevance of the content. Conversely, the presentation format (such as powerpoint), has nothing to do with how good a presentation is
7. Adjacent in space is superior to stacked in time - the classic example of this is showing Gallileo's solar observations in a flip book format where you can only compare two observations at once as opposed to lining them all up in sequence next to each other so you can see them all at once, or adjacent to each other.
8. Use small multiples
9. Put everything on the universal grid

Think of these as the 9 commandments of analytical design. Treat them all as rules that you cannot ever violate. In my opinion, the simple act of using his grand principles as a checklist for verifying the integrity of a presentation can dramatically increase the power and quality of any presentation, regardless of the format for delivery. I had no qualms whatsoever with his presentation on analytical design and evidence display. Learn it and live it.

On PowerPoint

Tufte makes no bones about the fact that he thinks PowerPoint is a hideous creation that needs to be banned from all corporate environments. He makes his case quite clearly by illustrating that PowerPoint had become so entrenched in the corporate culture at NASA that engineers were using incredibly low-resolution, weak, ineffective PPT presentations as a substitute for actual engineering and technical reports. The Columbia Accident Investigation Board (CAIB) was disgusted at the "pitch culture" that was so pervasive throughout NASA. Get Tufte's book, Beautiful Evidence and skip to the section on why Powerpoint is evil and you will be enlightened. I guarantee that he will convert you, and you will be forced to re-evaluate every presentation from that moment forward and use PPT only as an overhead projector to show images and other content that cannot be used in handouts or on paper. I know that I personally have completely re-evaluated my use of PPT, and I will be far more aware of its intelligence-sapping effect both on the presenter and on the audience in the future.

2 for 2 so far - I'm totally with Tufte on this point. PowerPoint should be used as an overhead projection tool - if you find yourself using it as a crutch so you can convey your point in grammatically incorrect bullet format, then you should stop right there and start over. Seriously. It might be tough, and you might suffer "bullet withdrawal", but in the end I think its worth it.

On Clutter and Hierarchies

Tufte's statement that Clutter is a problem to be solved by design, not by a reduction in information are words to live by. They should be in every single programmer's heads as they decide how to display large volumes of information - are you cutting out useful information as a crutch to compensate for poor design, or is there a better design that gives the user everything they need without stripping out meaningful data or information? Up to this point, I am completely in Tufte's corner.

Then Tufte drops the ball. He complains about hierarchies and forcing users to incessantly trudge through hierarchy after hierarchy. He claims that hiearchies are an artifact of a low-resolution interaction format like the computer screen because you can't display enough information. Ok, I agree there - everyone hates the automated voice-recognition phone trees you get sucked into every time you call a huge corporation. He then goes on to claim that one solution to this is to present the user with everything in a completely flat fashion. He lists sites that have over 400 links on the main page as a good example of this practice. This is where I raise the BS flag and stop being one of Tufte's sycophantic worshippers. He says something along the lines of, "Those sites get millions of hits per day.. Why aren't you setting your site up that way?" My response: Because I possess a modicum of sanity, thank you very much.

I'm sorry, but assaulting the user with so much "linkjunk" (to borrow from Tufte's chartjunk label) is just as bad as forcing them to click through 800 levels of a binary tree. He claims that the human "eye-brain system" is so powerful that it can handle scanning such a massive onslaught of text and links. While it may be possible, it may not be preferable. He compares this to the New York Times, which has a really high resolution. The difference - when people are navigating sites they are not reading articles, stories, or reviews. In short, NY times is prose and the navigation of a web site is interaction.

Swatting at the problem with a 500-foot-wide fly swatter and blanketing the entire problem of hierarchies by claiming that you should do away with hierarchies in favor of an assault of 400+ links per page is ridiculous. I'm surprised that such a suggestion comes from Tufte, who is otherwise lucid and sane. The happy medium here is to do some research, figure out what the goal of your audience is. Why are they coming to your site, what are they looking for when they get there - you give them that in a flat fashion in a simple, easy to access format that is free of nonsense flash animations, glitz, buzzwordiness, and foolish clipart. From there, you can lead them into a reasonable hierarchy. The reason for hierarchies is so that people only have to process a small amount of information at a time, making the data they are looking for easy to find. If you combine Tufte's zeal for throwing the entire site at you in a single page that scrolls ad infinitum with 8 billion links with the practical need for filtering information to reduce the amount of time users are wasting scanning through a maze of links - you might come up with something functional that your users will appreciate.

The sites that have 400 links on the home page that Tufte sited are not doing things right. The millions of people who use those pages are apologists (Tufte should take more of a cue from the Cooper books he cites as references). The reason millions of people use those pages isn't because of the brilliance of the flat link-assault design, its because those sites provide services that no one else does better. If another site came along that presented the same information in an easier-to-navigate fashion that was free of nonsense clipart, flash, advertising, etc - that site would have a chance at stealing some of those millions of site viewers.

Conclusion

In conclusion, I just want to say that this presentation was extremely valuable. The lecture was inspiring and I've gained a new-found appreciation for the seriousness of the presentation of evidence and information in my applications and in my work in general. From now on, I will have an entirely different checklist for self-evaluation before I send off a presentation document to a co-worker or before I give a presentation to co-workers (I am more likely to use handouts now instead of PPT!). The only area in which I had trouble believing Tufte was in his solution to hierarchies. In one breath he claims that clutter is a design flaw, not a flaw of the content. Then, immediately after that, he suggests deliberate clutter as a solution to incessant hierarchies. I think there's a happy medium, and I'm pretty sure I can achieve that in my work from this point forward, so in disagreeing with him I have also learned something new.

Regardless, anyone doing any programming with user interfaces, design, analytical design, or information presentation should seriously consider picking up his books and giving them a thorough read.

tags: presentation  tufte  lecture  information  evidence  analytical  design  

links: digg this    del.icio.us    technorati    reddit