Graphics Fail

One of the classic mistakes in graphics is the height / volume fail.  This is how it works:  the length of an object is used to portray some sort of relative metric.  But in the quest to make the graphic prettier, the object is turned into a 2D, or worse, 3D object.  This means that for a linear dimension where one object is 2x as long as another, its area is actually 4x the other and its volume is 8x.  The eye tends to notice the area or volume, so that the difference is exaggerated.

This NY Times graph is a great example of this fail (via here)

The Tebow character is, by the data, supposed to be about 1.7x the Brady character.  And this may be true of the heights, but visually it looks something like 4x larger because the eye is processing something in between area and volume, distorting one's impression of the data.   The problem is made worse by the fact that the characters are arrayed over a 3D plane.   Is there perspective at work?  Is Rodgers smaller than Peyton Manning because his figure is at the back, or because of the data?  The Vick figure, by the data, should be smaller than the Rodgers figure but due to tricks of perspective, it looks larger to me.

This and much more is explained in this Edward Tufte book, the Visual Display of Quantitative Information.  You will find this book on a surprising number of geek shelves (next to a tattered copy of Goedel-Escher-Bach) but it is virtually unknown in the general populace.  Every USA Today graphics maker should be forced to read it.

  • http://sgreffenius.wordpress.com/ Steven Greffenius

    Good post! I think in the last sentence, first paragraph, you intend: "The eye tends to notice the area or volume," not "The eye tends to notice the area or value..."

    I like data visualization questions. Thanks for these observations!

  • Ted Rado

    To an engineer, this is all obvious. Many years ago(in Neanderthal times) in Engineering Drawing classes, I studied orthographic and isometric projection, and perspective drawing. I also learned descriptive geometry. I don't know if current engineering students learn this stuff, or if it is passe what with computer drafting. It sure taught me to visualize things in three dimensions.

    Non-engineers have comparatively little ability to see things in three dimensions. It is up to the illustrator to present the info in a manner that enables the unwashed masses to get the correct idea.

  • dovh49

    I don't have that one yet, but yes, I was planning on buying it. I have one of his other books on visualizations.

  • me

    I love how over time, this blog hits all of my pet peeves. Thank you, Warren! Now that we've had "bad infographics", I am wondering if "chewing with an open mouth" might be coming up next :)

  • DrTorch

    Alas, my copy remains virtually unopened.

  • terrence

    Waddya expect from the New York Slimes? Knowledgeable articles? Dream on, McDuff!

  • http://thegameiam.wordpress.com David

    Reading Tufte's work dramatically improved my ability to create drawings which would convey what I wanted without conveying anything I didn't want. Sheer genius.

  • I Got Bupkis, Critic Extraordinaire

    >>>> Every USA Today graphics maker should be forced to read it.

    BWAAAAAAhahahahahahahaaaaaaa....!?!?!?!?

    You're looking for something resembling FACTUAL VALIDITY from a major MSM source?

    Are you DAFT, man?

    :^D

    Validity is not a concern for them.

    P.S., methinks the Eli Manning figure is gonna get a bit bigger after this weekend, too.

    P.P.S., It's GREAT to be a Florida Gator. :^D

  • I Got Bupkis, Critic Extraordinaire

    >> I don’t know if current engineering students learn this stuff, or if it is passe what with computer drafting. It sure taught me to visualize things in three dimensions.

    Ted, computer drafting CAN help -- a lot -- with these kinds of things, if the instructor works it in as a topic.

    I've done a good bit of drafting for architects. The only type I've EVER done is using AutoCAD. I recall a couple issues with a building. I was doing elevations on one four-story building dating from ca. 1905 (renovations) -- we had no paper plans, much less CAD drawings, so I was creating a lot of stuff from scratch. I couldn't readily measure the visual detail heights, but it was brick, so i counted courses, using the fairly standard 3course==8inches. So I drew it that way. But the problem was, I DID have an exposed evacuation landing along one side that I DID get measurements from, and they didn't match. So i went back, turns out, for this building, the coursework was 1c=3", so every three courses I'd lost an inch. Had to shift everything up by a variable amount. THAT was a pain in the tukus.

    Same building, we were renovating it, subject to restoring it as much as possible to its original form (Nat'l Reg. Hist. Places) -- but had to also offer ADA compliance, chose to do that with an add-on service tower that would include an elevator, ramps, and just have a multi-level enclosed crossway to the existing renovated structure. Anyway, The roofs on this were unusually steep, due to a need to match the height rise/run of the original structure, which was already somewhat steep. I was doing the stairwell drawings at one corner of the service tower when I thought I saw a discrepancy between the joist requirements and the necessary ceiling height of the top of the stairwell area. I quickly threw together a block-wise mockup of the spaces involved, and, yep, the joist was going to protrude down into through the ceiling. Thanks to my spotting it before it got to the field, we could detail how it was to be handled ahead of time and it saved some expensive change-orders later on.

    While you had to be more cautious before, now you can actually throw stuff together in 3d to see how things actually fit. More and more, this is going to be the case -- CAD/CAM is getting to the point where you can actually get 3d-"pieces" of any existing part, and/or make as a matter of course any 3d-pieces you'll be making on the fly, and just fit stuff together in the computer and see if there are any obvious issues. Not gonna spot everything but it requires less powerful spacial visualization skills on the part of the people doing it.

    Another example I've heard, only in description, was the idea of making a building in 3d, an elementary school, say. You can do a walk-around in the model and see how things will actually look...

    ...And THEN you can do another walk around as a person only 40" tall and see how it will look to small children -- and notice things like the fact that all the water fountains were set to adult-level heights.

    You could do similar things with handicap-specific issues by moving yourself around a space at seated height rather than standing height, and at least recognize/make trade-offs to make things more accessible to anyone in a wheelchair (I'm not a major fan of going to truly great expense for people in wheelchairs, but I don't see any reason why not do things with them in mind when there is little direct cost, for sure. If shelving being 6" lower might be good for them while putting no one else out, then hell, why not do it?)

  • I Got Bupkis, Critic Extraordinaire

    Keep in mind that visualization is something that modern computers can do tremendously easily. Most of the real meat of Chaos Theory did not get developed in full until the 1980s when computers were readily available to do 3d-modelling which allowed the people to play with the data and get a real visual feel for what it was doing. They'd done halting stuff before that, but when you're dealing with a few million data points it's not easy to look at them as numbers and feel what is going on.

    This is no less true of actual physical structures than it is for random-form experimental data.

    Modern engineering is far more able to do true fault-tolerant work just because you can put parts together and actually stress-test things in the model. Errors can still happen but far less commonly.

  • EscapedWestOfTheBigMuddy

    The Visual Display of Quantitative Information is absolutely the most important book no one ever told me about. I found it because my father in law---who deals in antique toys as a sideline and struggles to find gift for me at the usual times---bought me a copy at a estate sale.

    Now it has pride of place on my work bookshelf next to Tufte's later work Beautiful Evidence, Horowitz and Hill, The C Programming Language, The Particle Physics Data Book , my planometer and sliderule.

  • http://www.weird-strange-facts.com ibra

    Nice info up there.
    check out some of the latest UFO sightings at

    http://www.weird-strange-facts.com/latest-ufo-sightings.html

  • bob sykes

    Like Ted, I learned my drafting in the pen-and-ink and pencil era, but before I retired I found myself supervising a CAD class.

    You can do everything in CAD that you used to do on paper/vellum, but descriptive geometry has been severely deemphasized. Sketching, of course, is still done by hand, as it must be, even if its on an iPad. Nowadays, only the art students and architects buy T-squares and triangles.

    Critic Extraordinaire has a good point, especially for civil engineers and architects. Our work is largely renovation and remodeling, although a lot of brand new stuff is built, too (Cincinnati + Columbus + Cleveland every year). The great problem is getting factual data on the old stuff. Original drawings and specs and permits are guaranteed to be lost, because government agencies and owners can't be bothered. I once worked for a company whose main competitive edge was their hugh library of drawings and maps for the as-built environment of many New England towns.

  • bob sykes

    Like Ted, I learned my drafting in the pen-and-ink and pencil era, but before I retired I found myself supervising a CAD class.

    You can do everything in CAD that you used to do on paper/vellum, but descriptive geometry has been severely deemphasized. Sketching, of course, is still done by hand, as it must be, even if its on an iPad. Nowadays, only the art students and architects buy T-squares and triangles.

    Critic Extraordinaire has a good point, especially for civil engineers and architects. Our work is largely renovation and remodeling, although a lot of brand new stuff is built, too (Cincinnati + Columbus + Cleveland every year). The great problem is getting factual data on the old stuff. Original drawings and specs and permits are guaranteed to be lost, because government agencies and owners can't be bothered. I once worked for a company whose main competitive edge was their hugh library of drawings and maps for the as-built environment of many New England towns.

    Tufte is marvelous. Everyone who designs data displays absolutely must read him. It is especially important to note that the defaults in MS Excel violate virtually every principle of good graphics. But then, Excel was designed for MBA types who need to obscure data, not explain it.

  • John Cunningham

    It's great to see you reference Tufte. each of his books is a masterpiece, highly worth regular re-reading. In a decent educational system, his books would form the basis for a mandatory course for all.