Monday, January 11, 2010

Aspiring to Tuftian perfection - presenting data

Often financial, scientific or engineering analysis takes large amounts of data and numbers and crunches them down to a single number or decision. I know it felt that way when I was researching and writing my thesis. Much like 42 being the answer to "Life, the Universe and Everything", it doesn't seem fair to take all of the data and turn it into one number or decision. Being intellectually honest and the presentation of the data have a lot of influence on the final outcome.

That intellectual honesty and the importance of good data representation and presentation are the lifework of Edward Tufte. I have just finished reading Beautiful Evidence and Visual Explanations which are his third and fourth books on the topic of data presentation. I am going backwards in the series because that's what's available to me and because I needed to answer a question about presenting some data of my own, while keeping my intellectual integrity at the same time.

Edward Tufte describes his books in the introduction to Visual Explanations. He says that The Visual Display of Quantitative Information is about pictures of numbers, Envisioning Information is about pictures of nouns, and Visual Explanations is about pictures of verbs. Beautiful Evidence seems to be about the best examples of these pictures, where well presented data helps in decision making. He also has counter examples where poorly presented data hinders decision making. In the two books I have just read he takes especial umbrage with the presentation of the data for the Columbia and Challenger shuttle disasters. His opinion is that the data presentations failed to stop and potentially caused the disasters. He also hates Powerpoint. He has been quoted as saying "Power corrupts, PowerPoint corrupts absolutely"

In Visual Explanations, in his review of the reason for the Challenger disaster he notes how the data about O-rings should have actually resulted in a recommendation not to fly but the presentations did not get the message across. The flight day had temperatures and conditions that were far outside of any of the past experience in flight or testing that the O-ring testing and usage had, yet the data presented didn't clearly indicate that. He proposes the unclear presentation as a reflection of unclear thinking and suggests this is the cause of the disaster. There is some disagreement with Tufte's analysis of the presentation flaws leading to the bad decision to launch in cold weather.

In this book, Tufte also uses the famous origins of epidemiology story of John Snow discovering that the Broad Street well was the source of an 1854 cholera outbreak in London. Tufte cites the discovery of the pump well that caused cholera epidemic as good presentation of data but he cautions that if the data were presented in a different form it might give different conclusions. Tufte lauds the fact that Dr. Snow checked his conclusions by looking at anomalous data and finding the reasons for them to see if his hypothesis was supported.

In Beautiful Evidence, Tufte reviews the Columbia Space Shuttle disaster. When the Columbia Space Shuttle was in orbit after it appeared that the wing had been struck by debris during launch, the team on the ground had 12 days to determine if the debris that hit the wing was an issue. Once again the team could not present the information in a manner to convince people that the impact under discussion was probably far larger than any from data they had from previous testing. Tufte is extremely critical in these examples because he feels that the bad presentation was a sign of unclear thinking and was in the direct causal chain that led to the disaster.

Of course there are those who decry my "What would Tufte say?" enthusiasm as I critically analyze every chart in every presentation that I make. I wholehearted recommend reading all of his work, and I realize that Tufte's design of presentation criteria are excellent but perhaps a little perfectionist. Not every presentation I make is life or death, in fact none are. So perhaps I can get away with a "good enough" approach in which I try to keep some of his principles in mind when I must present data and still conform to a corporate fashion for PowerPoint or Excel charts of a certain type. The alternative is to despair of ever producing Tuftian presentations with a clarity of thought that only a few transcendent individuals possess. Like Salieri as portrayed in the fictional Amadeus, I can recognize his genius but can only produce mediocrity myself.

1 comment:

Anonymous said...

You may be interested to know that Tufte was just appointed to serve on the Recovery Independent Advisory Panel.

I suspect that he will recommend against the use of PowerPoint.