Join 3,366 readers in helping fund MetaFilter (Hide)


The Information Sage
May 17, 2011 3:53 PM   Subscribe

“If you display information the right way, anybody can be an analyst,” Tufte once told me. “Anybody can be an investigator.” - The Washington Monthly interviews informaticist Edward Tufte [via]
posted by Blazecock Pileon (45 comments total) 76 users marked this as a favorite

 
I'm a big fan of Tufte. In fact, I have a copy of his book A Visual Display of Quantitative Information. I can definitely say that reading his insights has informed my own work and helped me in my chosen profession.

It's also a beautifully well-put-together book. One of my prized possessions.
posted by jnrussell at 4:07 PM on May 17, 2011 [1 favorite]


If you display information the right way, anybody can be an analyst

FIRST ANALYST!!!
posted by It's Raining Florence Henderson at 4:08 PM on May 17, 2011 [1 favorite]


Does Tufte provide suggestions of software that helps people to display information "the right way"?
posted by lukemeister at 4:11 PM on May 17, 2011 [2 favorites]


Well, lukemeister, he tends to favor programs like Illustrator...which aren't too helpful for the average joe. That being said, there's quite a bit you can do in bog standard Excel to make things easier to read.

The folks over at Juice Analytics have a pretty helpful chart chooser in addition to numerous articles on how to get the most of of Excel.
posted by jnrussell at 4:19 PM on May 17, 2011 [5 favorites]


Wouldn't that just make the actual analyst job the one who figures out how to represent the information?
posted by delmoi at 4:23 PM on May 17, 2011 [5 favorites]


How'd I do?
posted by Devils Rancher at 4:24 PM on May 17, 2011 [26 favorites]


Wouldn't that just make the actual analyst job the one who figures out how to represent the information?

Ideally, it's a collaborative effort between the analyst and the designer. At least, that was the best way I found to work on informational graphics.
posted by Thorzdad at 4:27 PM on May 17, 2011


I welcome new and interesting ways to express ideas graphically. But I fear the fetishization of infographia will inevitably lead to abuses that are diametrically opposite to Tufte's ideal.

The other day I was looking at a very beautiful, striking graphic, but it was taking me a moment to wrap my head around the actual abstract information. What is this chart showing me? Then I realized that it would have been much simpler, much easier to interpret, as a boring old bar chart.
posted by Xoebe at 4:28 PM on May 17, 2011


That map of the French invasion of Russia is horrifically effective.
posted by It's Raining Florence Henderson at 4:32 PM on May 17, 2011


I'll just share the worst designed graph i've seen in some time.

(unfortunately, that'll probably be meaningless to anyone that doesn't play starcraft)
posted by empath at 4:33 PM on May 17, 2011


empath (and everyone else), here is a classic.

I think I found the blog linked above here on MeFi. Maps are a great way to examine the effectiveness of graphic communication. We know what a map generally represents, it's fairly intuitive. Adding abstract data to the spatial representation quickly reveals problems with the representations of either.

Hey, weren't there some Federal Government maps recently here on the blue that took a lot of flak? Yeah.
posted by Xoebe at 4:41 PM on May 17, 2011 [4 favorites]


I welcome new and interesting ways to express ideas graphically. But I fear the fetishization of infographia will inevitably lead to abuses that are diametrically opposite to Tufte's ideal.

Tufte isn't necessarily about new or interesting. He's all about effective, and whatever form the data must take to clearly and effectively communicate the message. But, your fear is correct and, in fact, it's already happening. There's no lack of terrifically horrible graphs and charts out there that mimic the visual style of Tufte-ish graphics, but completely missing the "effective" part of communicating the information. Morons armed with illustrator.
posted by Thorzdad at 4:44 PM on May 17, 2011 [4 favorites]


The folks over at Juice Analytics have a pretty helpful chart chooser in addition to numerous articles on how to get the most of of Excel.

Someone should give Fox News that URL.
posted by birdherder at 4:45 PM on May 17, 2011 [1 favorite]


My coworkers and I went to a Tufte seminar a few years ago and we still trade examples of bad charting via email with "Tufte Wept" in the subject line.
posted by birdherder at 4:47 PM on May 17, 2011 [8 favorites]


Ha! I think that might catch on, birdherder!
posted by It's Raining Florence Henderson at 4:49 PM on May 17, 2011


Oh, man:

One day, when he had been dating Druckrey for some time, his office manager came in to work, opened the shoe box, and saw a piece of torn paper that said, “Find out how to get married.”
posted by beepbeepboopboop at 4:54 PM on May 17, 2011


empath,

That is by no means nearly the worst graph- for one thing, it's only in 2D.

Xoebe,

That is an ugly, ugly graph, but it actually presents a large number of data very efficiently. Technically, it does follow Tufte's principles, particularly as they relate to dense, multivariate information presentation.

I am living proof that good data presentation, without necessarily a strong background in quantitative analysis, can go a very long way in a number of disciplines that claim to be data-centric.
posted by TheWhiteSkull at 4:55 PM on May 17, 2011


How'd I do?
posted by Devils Rancher


You just won charts. Game over.
posted by GuyZero at 5:10 PM on May 17, 2011


Excellent post. For anyone interested in the field, I'd recommend checking out Stephen Few in addition to Tufte. Tufte's books can be pretty inaccessible, especially if you're rather artistically challenged (as I am). Few tends to focus much more on psychology and cog sci than Tufte, so his work is a good supplement.
posted by graphnerd at 5:12 PM on May 17, 2011 [5 favorites]


I spent the entire day today at his 6 hour course in Arlington.

He said that this article was supposed to have been in the New York Times Sunday Magazine but something happened last minute so they put it in the Washington Monthly.

The class was amazing, BTW.
posted by k8t at 5:33 PM on May 17, 2011


Here's a copypasta of my Tufte class notes:

Show causality and testify for it
Is the analysis dictating the method?
Visualization do whatever it takes to understand
Don't pre specify things
Plurality of methods
Consumers want to see whatever it takes
Don't show off the method but what you learned
Models without boxes
Don't put quantitative data in your model or display unless you have to to demonstrate comparisons
the character of the relationships should be expressed
Arrows deny you verbs
Maps are the model for conveying info with no boxes
What's your story and are you credible?
Credibility is competence, not cherry picking (make data available), no motives (most people aren't evil).
Human behavior is less clear cut than hard sciences
Open mind not empty head
Super graphics are key so ppl can use their own cognitive style to understand information
Ppl don't need to be read to
No slow reveal
High res data dump so people can read with their own cognitive style which is faster than listening. You annotate super graphic
High res is interactive
Ppl don't have to wait for something relevant
Super graphic of penetration rates?
We want to see causality in things
But we are indifferent to mode
We shouldn't segregate info into different applications because this percolates up
People map of adopters and their demographics?
Paper is best for spark lines
Use trebuchet for tables
Nature journal best viz
Why can ppl read sports page but not data at work?
SE on p, 174 table display

General advice about non fiction reports
1. Find good examples and copy them. 6-8 templates. Design is a research problem. Google image search.
1a. copy Nature. Public library of science. Professional reporting trumps corporate needs.
2. Make a good table and use it over again.
3. Super graphic and find one.
4. Intellectual model for presentation - sports page. Reporting nonfiction. Perspective version of Nature. Then maybe nytimes will cover it. If your stuff is important.
5. Credibility is story and how you convey it. Demonstrate mastery of detail. don't talk like an MBA. Don't over do it.

Two concerns in showing info.
Multivariate analysis how to express. How many dimensions?
Info resolution. Codes are an impediment to learning. So be thoughtful in design. Directly label your line rather than have a legend. Have props in 3d.

P. 137 principles of analytical thinking and design

Ppl want spatial adjacency
Area and high res

Presentations:

Ppt is solely a projector operator system
Fullscreen is good

Word is a page layout device

Write in sentences

Have a summary on 11x17 with super graphic. What the problem is, who cares and what you're going to do about it. 100 words.

At beginning give people time to read because they can do it faster than listening and they can use their own cognitive style.

Walk them through the handout informally. Highlight important things.

If less than 30 people you'll have a q and a. Makes for a short and effective meeting. No bullet points.

Preparation:

Make a big improvement through better content. Spend your time on the content rather than the template.

Performance - rehearsal improves performance.

Arrive early to your own presentation. Advances your cause in that time. It is gracious.

Finish your presentation early. People will like that.
posted by k8t at 5:35 PM on May 17, 2011 [34 favorites]


Some additional facts because I came in a few minutes before it started and had to sit right in front of him... (He signed my book without me asking before things got started).

- Tufte smells amazing. I hate perfume-y smells, but his was fantastic.
- He had a cool belt with red stars built into the leather (I recall this being some symbol of a particular Asian leathermaker...).
- He is 60 but is incredibly physically fit and looks like he could be in his late 40s with great hair and he looks great in a suit.
- He had 3 lackies with walkie talkies doing things for him - water, nuts, coke with ice (He didn't eat or break to use the bathroom ALL DAY). The lackies had to wear light blue button down shirts, darker khakis, and brown dress shoes. One ran the MacBook Pro for the displays (and there was a backup MacBook Pro in case something happened to the first one) and also did "crowd control" during the autograph sessions, moving people along when they were taking too long. The main lackey also brought around the multiple 1st edition books that Tufte uses for illustration (Gallileo, Euclid) and the lackey had to wear white gloves while taking it around (Tufte did not wear gloves and in fact went off for awhile about how he believes rare books should be read and touched).
- He is totally obsessed with the iPhone and iPad (although he has a few critiques of the design) but believes that in 5 years everyone will have a tablet of some sort and that hi res "super graphics" will be common place. He used an iPad for his lecture a few times. He has an iPhone 4 with a sheep as the wallpaper.

- He gave off a very arrogant vibe.

- When I asked him a question about displaying a particular statistical analytic model visually he didn't understand what I was asking of him. It made me think that he is good at statistical analysis of the past (not the modeling that quantitative social scientists trained today do - HLM, SEM, latent class analysis) and he isn't really in touch with the realities of publishing in "normal" journals that have strict style requirements for presentation of data. That isn't to say that it wasn't useful for me -- and I'll probably use some of his ideas in my conference presentations -- but he's going for people that do more descriptive analysis than the sort of multilevel modeling that I do.
posted by k8t at 5:59 PM on May 17, 2011 [8 favorites]


Can you put that into a chart for me, k8t? Too many words.
posted by It's Raining Florence Henderson at 6:01 PM on May 17, 2011 [6 favorites]


(Seriously, though - thanks for sharing!)
posted by It's Raining Florence Henderson at 6:01 PM on May 17, 2011 [1 favorite]


I didn't write in my notes his argument that text and images need to be combined more and that they used to be when they were created in one place (one guy, one pen - writes the words and draws the drawing). Today we do our drawings in one place, write in another, analyze our data in another... this is not an okay way of doing things (and he went off on the operating system developers for creating such a situation).

He also talked up Chrome OS a lot. He really seemed to have his finger on the pulse of tech stuff. I was impressed by that.

He also bitched about how hard it is to manage forums and how his "people" strictly moderate his discussion forums.
posted by k8t at 6:02 PM on May 17, 2011


Well, as a technical writer, I feel like this thread was made just for me. Thanks to BP, and everyone who has participated.
posted by It's Raining Florence Henderson at 6:04 PM on May 17, 2011 [1 favorite]


Oh, and we got the books in an adorable little carrying case.
posted by k8t at 6:06 PM on May 17, 2011


I agree with delmoi, if you've decided how to display the information "the right way" that is the analysis. After that people consume it. If they are able to do further analysis on top of that particular presentation of the data it is largely coincidence.

Analysis of data and presentation of that analysis is hard and not everyone can do it. Reading the conclusion of the analyst is easy if they've done it right. Those are different things.
posted by stp123 at 6:07 PM on May 17, 2011


Wouldn't that just make the actual analyst job the one who figures out how to represent the information?

Depends on what you mean by "the actual analyst". I've been working on my application's graphing facilities the last couple weeks. You definitely want to design the right graph to highlight certain relationships. But the "actual analyst" still has to hunt around with baby steps after taking those big strides in the right direction to home in on an actual solution.

Also, I look forward to all future FPPs including information on how the subject smells.
posted by DU at 6:39 PM on May 17, 2011 [3 favorites]


Does anyone know if he has ever toured the southeast for his lectures? What awe inspiring work!
posted by cavalier at 7:07 PM on May 17, 2011


@cavalier: I know he travels around a bit, but I don't remember how large his circuit is (it seems to be a circuit). I don't see registration for any courses in the southeast at the moment.
posted by thebestsophist at 8:05 PM on May 17, 2011


cavalier,

Tufte's web site says he is giving the course in Arlington, VA tomorrow (May 18). That's sort of southeast by Mountain Time Zone standards.
posted by lukemeister at 8:13 PM on May 17, 2011


Tomorrow's class in Arlington is sold out.
posted by k8t at 8:17 PM on May 17, 2011


Of course, all the analysis and appropriate design in the world can't save a graphic from executive meddling.

For instance...
At my last job, I worked with an analyst on a series of graphics for the CEO to use in an upcoming presentation to potential investors and the board. There was one graphic he wanted that would illustrate the increase in profits (or efficiency, or something) for our customers over a six-month period following installation of our product. Unfortunately, no matter how many times the analyst and I attacked the numbers, the graph showed a straight line trending slightly down.

Of course, the CEO wasn't happy about this. He ordered us to attack the numbers once again and, this time, find a time-span that described an increase in profits. Eventually, we did find an upward-trending time-span...Approximately a three-week period that drew a strong upward line. The CEO was happy, and then he ordered the graphic to take that actual trend and draw a straight extrapolation from that line. Oh, and he told me to remove any references to the actual time-span (days, weeks, etc.)

In the end, he had a graphic that supported his agenda, but which was highly fabricated and misleading. The analyst quit right after that.
posted by Thorzdad at 4:32 AM on May 18, 2011


We know what a map generally represents, it's fairly intuitive

But dangerously not. What a geographical map loses is population density, thus, slicing anything dealing with people by geographical lines is a great way to distort the data if you're just summing events.

It's fine when you slice by events over population, say, events per thousand residents.
posted by eriko at 6:31 AM on May 18, 2011


For comparison purposes, here's how I did it wrong
posted by hank at 6:41 AM on May 18, 2011


Thorzdad: He ordered us to . . . find a time-span that described an increase in profits. . . . take that actual trend and draw a straight extrapolation . . . remove any references to the actual time-span

This is covered under 'Professional Ethics' rather than 'Information Design'.
posted by Herodios at 6:49 AM on May 18, 2011 [1 favorite]


I welcome new and interesting ways to express ideas graphically. But I fear the fetishization of infographia will inevitably lead to abuses that are diametrically opposite to Tufte's ideal.

Tufte isn't necessarily about new or interesting. He's all about effective


Rat own!

In one of his books -- sorry, I don't recall which one -- Tufte points out that sometimes, a table of figures is the most effective info display. Not the most exciting. Not the most creative. Not the most interesting. The most effective way, considering the appropriate information density needed to 'tell the tale'.
posted by Herodios at 7:09 AM on May 18, 2011


Does Tufte provide suggestions of software that helps people to display information "the right way"?

I went to one of his day seminars and it was splendid. This question was asked a couple of different times and no satisfactory answer was given. I think the reason for this is because the tools shouldn't dictate your process. He tried to get this point across by showing examples and explaining how they were created. Most of them were one-off jobs, which makes repetition of Tufte's principals somewhat difficult in practice.

Adobe Illustrator was recommended for high-resolution layout and graphics capabilities. I believe R may have been mentioned in passing, but really the focus was on fantastic displays of information that are fine-tuned for their specific goal; something you just can't do repetitively (easily) with Excel and some templates.
posted by odinsdream at 7:43 AM on May 18, 2011 [1 favorite]


Hmm, I went to his class here in Dallas and found it unbearable. Basically consisted of him reading sections of his book to us. His additional insight was often right there on the page, just him paraphrasing or rephrasing things that I had already read.

And I 100% agree with k8t's perception that he came across as arrogant.

Tufte did spend a chunk of time talking about PowerPoint. I found this ironic, because his personal presentation solution instead was to stop the class, blank out the screens, and have us read silently for a few minutes - for each distinct section he wanted us to review. At some point, I would have preferred the PowerPoint.

Basically, his message is not for me. I'm interested in practical application, not theory. I'm not working in a framework where hand-created Adobe Illustrator graphics are an option. The books are a wealth of theory knowledge, but I personally could have skipped the class. Really glad to hear others got good value out of it for their uses.
posted by TheFlamingoKing at 8:55 AM on May 18, 2011


A takedown of a Wall Street Journal editorial's bogus chart

posted by hank at 9:16 AM on May 18, 2011 [2 favorites]


Basically, his message is not for me. I'm interested in practical application, not theory. I'm not working in a framework where hand-created Adobe Illustrator graphics are an option. The books are a wealth of theory knowledge, but I personally could have skipped the class. Really glad to hear others got good value out of it for their uses.

While I did enjoy the class, I do have to agree with this point. There actually would be a great market for classes that taught how to repetitively implement Tufte's design principles in real-world scenarios with real software.
posted by odinsdream at 1:19 PM on May 18, 2011 [2 favorites]


"I told Edward Tufte that this box was The Google. Later, I opened the shoe box, and saw a piece of torn paper that said, 'Find out how to get married.'"
posted by moonmilk at 1:23 PM on May 18, 2011


'The Cognitive Style of Powerpoint' is worth a read.
posted by greytape at 2:22 PM on May 18, 2011 [4 favorites]


I had correspondence with his office this morning -- they have published a whole new tour itinerary out to the end of the year if anyone wants to check again.

Also, they confirmed that he has no plans to visit Florida. :P

Anybody interested in a meetup somewhere out there in Denver or Minneapolis next month?? :)
posted by cavalier at 9:33 AM on May 19, 2011


« Older The National Endowment for the Arts' new Arts in M...  |  Thundercats Are Go, for a rebo... Newer »


This thread has been archived and is closed to new comments