Does GE Think We’re Stupid?
The series of interactive data visualizations that have appeared on GE’s website over the last two years has provided a growing pool of silly examples. They attempt to give the superficial impression that GE cares about data while in fact providing almost useless content. They look fun, but communicate little. As such, they suggest that GE does not in fact care about the information and has little respect for the intelligence and interests of its audience. This is a shame, because the stories contained in these data sets are important.
Most of the visualizations were developed by Ben Fry (including the colorful pie that Homer is drooling over above); someone who is able to design effective data visualizations, but shows no signs of this in the work that he’s done for GE. The latest visualization was designed by David McCandless, who has to my knowledge never produced an effective data visualization. In other words, GE has gone from bad to worse.
My friend and colleague Stacey Barr, a leading expert in performance measurement, recently sent me a link to GE’s new natural gas visualization, designed by McCandless, along with these words:
I get these emails from Information is Beautiful, and while yes they are beautiful, I often struggle to interpret them. Have you seen them? Would you agree (or do I need remedial training in how to read graphs and diagrams)?
Take a look for yourself to see if you share Stacey’s perspective.
The only things you can’t see here that are available in the original visualization on GE’s website are the names of the countries (one per square) that appear only as you hover over them. What we see here looks like an oddly designed version of a treemap, with only one dimension of data (the relative sizes of the squares, which in this case represent the amounts of natural gas left per country) rather than two dimensions (using varying color intensities as well to display a second set of values). Unlike a true treemap, however, the squares do not fill the space, which raises the question: “What do the white spaces mean?”
If you’re like me, you turned to the legend in the upper left corner to learn the meaning of the white spaces. I interpreted the legend to mean that the blue squares represent amount of natural gas and the white spaces represent years of natural gas left. What followed, of course, was complete confusion. Only after clicking on everything in sight did I eventually discover that what I assumed was a legend is in fact a unique control for switching between quantity and years. The blue and white colors of the control, though they match the colors of the chart, mean nothing. The white spaces in the visualization are also meaningless.
Unless you enjoy hovering over individual squares one at a time and trying to compare their sizes to one another—something the human brain can’t do well—this entire display reveals only a single fact: The world currently uses 2.6 trillion cubic meters of natural gas a year and we only have 187.9 trillion left. Words and numbers tell this story; the graphic is little more than decoration.
McCandless’ design also provides access to other related views. Here’s the next:
Now, rather than squares with rounded corners and mysterious spaces in between, we have seven rectangles with straight corners, without white space, except for the odd section in the lower right corner. It’s looks as if McCandless’ calculations didn’t work to fill the space, leaving a bit left over.
Why not use a graph that actually makes it possible to interpret and compare the values without reading the text? The same values appear in the bar graph below, which I quickly constructed in Excel, but now we don’t need to read the text to determine or compare the values.
Are McCandless’ big blue rectangles, which vary in overall area (width times height) more beautiful than the smaller blue rectangles in the bar graph that vary in length alone? Actually, they’re not, but even if they were, would the additional beauty justify our reduced ability to read the graphic? Certainly not for anyone who cares about natural gas reserves.
The final view reveals the following:
Before anything else, notice where your eyes are primarily drawn. Most likely to the blue graphics. Unfortunately, these graphics, despite their visual complexity, are almost entirely without meaning and can only be compared in the roughest of ways. The fact that oil, natural gas, and coal have been displayed in different ways (a circle, square, and a triangle) makes comparisons especially difficult.
This visualization includes some interactivity, letting us increase or decrease each source of energy’s annual amount of production to see how it alters the number of years that resource will remain. Here, I’ve increased the production of oil, which decreased the size of the blue circle.
Unfortunately, because the graphics can’t be meaningfully deciphered or compared, the only useful piece of information that we can get from this is the number of years left, which is presented as text.
How would you respond to Stacey’s questions? Are these visualizations of natural gas and other sources of energy easy to interpret? Do they do a good job of telling the story of depleting resources? Do they suggest that GE or David McCandless actually care about this information? Do GE or McCandless respect our intelligence?
I’ll leave the answers to you. You already know what I think.
Take care,
P.S. Why such passion?
Some of my colleagues in the field of data visualization, especially those who work predominantly in academia, don’t fully share my strong reactions to poorly designed data visualizations. I’ve spent some time wondering why I feel more strongly about this than they do, despite the fact that we base our work on the same evidence-based principles. One could argue that I’m by nature more reactive, more apt to raise the banner of battle in the face of harmful practices. There might be some truth in this. Despite my penchant for speaking out against perceived wrong, upon reflection, I’ve come to think that my strong opposition to harmful data visualization practices compared to the greater tolerance of my academic colleagues stems from the different worlds in which we live and work.
Most of my time is spent working with people who use data visualization to support decision making in their organizations. They don’t live in the realm of theory but in a daily struggle to squeeze real value from their data. In many cases, their professional success or failure is measured by the degree to which they are able to find, understand, and then tell the important stories that live in their data. I watch them labor to exhaustion with poor data visualization tools or delight in tools that extend their reach. I applaud their deep desire to solve real problems in the world through the use of data. I empathize when they report the frustration that they feel as they struggle to show their organizations a better way, hindered by horrible examples of data visualization that have captured the imagination and set the expectations of those in charge (for example, a CEO who loves his pie charts and dashboard gauges). In this world, the quality of data visualization—the culture that supports or thwarts it, the skills that extend or limit it, and the tools that support or undermine it—really matters.
Because this is the reality that I see almost everyday, dispassion would be heartless. This isn’t just a job for me; it’s a mission. There are certainly more important matters in the world, but this is one that I can do something about.
P.S.S. A second opinion
My wife, Jayne, is a neuropsychologist. She diagnoses and treats brain disorders and injuries. Other than a shared interest in aspects of the brain, our professional lives don’t overlap. I asked Jayne to read this blog post to see how someone outside the field of data visualization would respond to GE’s natural gas visualizations and my critique of their merits.
As Jayne began to read my review, she soon reached the place where I wrote that I’ve never seen an effective data visualization by McCandless and sighed. Her thoughts were, “Here he goes again,” fearing that my words might set off another firestorm of angry responses. Then she continued reading, and her attitude changed. Here’s what she said about the natural gas visualizations after finishing:
They make no sense whatsoever. I actually found myself getting angry as I reviewed them—primarily because it seems like a deliberate attempt on GE’s part to deceive—to make us think they care about an issue by posting pretty visualizations that, in reality, say nothing. GE and David McCandless should be ashamed! After reviewing these visualizations, it seems clear that GE doesn’t really care about the issue, they just want us to think that they do.
You might be tempted to interpret Jayne’s comments merely as a wife supporting her husband. If you knew her, however, you wouldn’t make this mistake. She would have made no bones about disagreeing with me. You might not share her opinion, but you shouldn’t dismiss it.
She asked me: “How can people defend these visualizations? Why would they do so?” These are questions that I can’t answer. I’m just as puzzled as she is.
47 Comments on “Does GE Think We’re Stupid?”
Personally, I can’t figure out why GE doesn’t capitalize on this. With as much time staring at these graphics as is required for a person to make any sense of them, you’d think they would sell ad space on those pages and try to make a profit.
As for how I’d respond to Stacey, I’d tell her that these graphics don’t appear to have been created with any intention of conveying information. Because they don’t. Period. They are nothing more than pretty pictures to look at.
As art, interesting, even attractive.
As a means to convey information, most wanting.
They should go ten years back – less fun, more information http://www.ge.com/annual01/letter/cockpit/
Tomas,
More information, but horribly designed. I use this Digital Cockpit by GE in my dashboard design class to illustrate what not to do. It’s nothing more than a poorly designed table with traffic lights. This and probably five times as much information could be displayed on a dashboard much more richly had they used appropriate graphics.
Have you checked out GE’s attempt at making sense of the Census data? It’s slightly better than the visualizations in this post — the scale bothers me — and wasn’t designed by McCandless.
http://fathom.info/projects/indicators.html
I get the impression that many David McCandless infographics are based on small or incomplete datasets.
For business users, it’s pretty straightforward to to a simple presentation of small data and say “we need more data.” For informational artwork or illustrations, a simple bar chart is unacceptable because it immediately turns the audience off.
Thanks for writing this Stephen, it’s good to see information design critique happening publicly.
Agree with what Bill said above, this might pass as art, but hardly as good information design.
As UK-based designer Paul Gyford put it “Just because it’s graphical, doesn’t mean it’s useful.” (http://www.flickr.com/photos/philgyford/4505748943/)
GE capitalizes big time on these, this is first of all marketing / image building for them (compare it to other companies who simply slap a ‘eco/environment-friendly’ sign on their site).
Therefore, if the visualization is attractive, and the perception for their average customer is that they are concerned with the environment, then their goal is reached.
IMHO the problem lies foremost in the term ‘data visualization’, which is too abstract. As it is used by a broad interdisciplinary group of people, each group/discipline interprets the actual goal of a visualization in another way.
Giovanni,
It is definitely true that a great deal of the confusion stems from the fact that people aren’t defining “data visualization” in a uniform way. In the past, even though various people defined data visualization somewhat differently, they all defined it as something that above all intends to inform. We could all agree that a data visualization succeeds to the degree that it leads to understanding.
Only in recent years have people with predominantly artistic intentions begun to produce something quite different without giving it a new name or clearly declaring their intentions. This has led to a great deal of confusion and contention. I could appreciate what some of them do as “data art”, but I can’t appreciate it as “data visualization.” If data artists don’t like their work to be evaluated based on its ability to inform, then they should call it something other than data visualization and declare their intentions. We could then evaluate the success of their efforts by how well their intentions are achieved.
@Russ: “For informational artwork or illustrations, a simple bar chart is unacceptable because it immediately turns the audience off.”
What _informational_ mechanism would you suggest as an acceptable replacement for a simple bar chart? I have yet to see any myself.
I generally agree with your opinions on bad and good data visualization but don’t have the balls to say it the way you do. I am glad that you are there to say it – clearly.
I fully agree with you concerning the argument. But why always so negative? when i click on Examples on your Website, first things i read is:
“Each of the examples that appear below illustrates quantitative
information that is poorly designed for communication.
Click on any of these examples to see an analysis of its problems
and my proposed solutions.”
Why not start with a good example and a positive statement?
Stephen
They do have a use. By their very nature we can all add them to our list of examples that do not work. They become the evidence of what you and others preach. Without them where would you and this blog be?
So in this respect they are “good” visulisations – but for the wrong reason.
“people with predominantly artistic intentions” aka. graphic designers* :)
The intention of a visualization can/should be deducted by it’s context, I totally agree that a visualization should first-of-all inform, but if the visualization fails to reach the target audience, it never reaches the point that it can inform.
In the context of mission critical visualizations, any design must limit itself to facilitate the interpretation of the data, there is no need for (superficial) esthetics.
When the goal of a visualization is to reach a new/broader/non-dedicated audience, a certain amount of esthetics is needed in order to attract them, albeit, in order to inform correctly, the data/metrics should be honored.
Unfortunately, certain designers have been poisoned by marketing (thus added deception to their toolbox) the correctness of exact data/metrics has become less important to them to conceive a message, (superficial) esthetics have become dominant (to simply transmit a feeling).
For those, ‘data visualization’ is simply another design style, which they use as visuals w/o understanding the (cognitive) purpose of it, often used out of context.
Out of this comes a contradiction: there is a (valid) skepticism towards graphic designers from eg. the academic corner, whilst on the other hand, they struggle to decently formulate there research findings in order to inform an audience eg. on why they need further fundings.
* in general, graphic designers cringe when McCandless calls himself a graphic designer, he’s a good journalist / story-teller, but many times his narratives do not benefit from his design skills.
Matthias,
The point of the Examples section of this website is to teach people why many of the ways that graphs are typically designed don’t work and how they can be improved. To do this, I must begin with the problem, explain why it’s a problem, and then proceed to a solution. People don’t fully appreciate solutions or even recognize them as such unless they first recognize the problems.
In more general terms, the reason why I write more unfavorable reviews than favorable is because there are far more bad examples of data visualization in the world than good. Given the ratio, what’s surprising is that I write as many positive reviews of data visualizations, products, books, articles, etc., as I do. I relish opportunities to write positive reviews and appreciate every opportunity that presents itself.
I was looking at the Gas Giants visualisation (the one with the bit missing) and cringed a bit. I personally prefer the Fewnian approach of bar charts, but I would have also added an ‘others’ entry to represent the capacity of the rest of the world bring the dataset to 100%.
I would suggest that a better way to illustrate interactively how long our reserves will last would be a line chart with cubic metre on the y-axis and time on the x-axis with the lines diminishing to zero in a number of years. You could argue that the only crucial piece of information is the time when reserves run out, but I would argue that the gradient of line would also illustrate that if production levels grow we will run out faster each year.
I had thought about having each country as a separate line, but that could get cluttered very quickly.
@Giovanni “…I totally agree that a visualization should first-of-all inform, but if the visualization fails to reach the target audience, it never reaches the point that it can inform.”
I don’t think anyone has argued that a visualization shouldn’t be designed to attract an audience. If I recall, Stephen has suggested many times before that visualizations can be informative as well as visually attractive. Nobody is criticizing McCandless’ work for being aesthetic; his work is being criticized simply because it fails to inform at all.
Tell GE (or Ben Fry) that a Nightingale rose is not a pie. A pie has constant radii with varying angles while a Nightingale rose has varying radii with constant angles. The figure above varies both. Anyone know the name for it?
@Andrew “…McCandless’ work … is being criticized simply because it fails to inform at all”
I think we can all agree on that :)
I was actually beyond that point, on the matter of ‘data viz’ vs. ‘data art’, the role/purpose/responsibility of the graphic designer etc…
Just being honest here… but my first reaction to this post was “duh!”.
This is marketing. GE happens to be using a tool (data vis) that you have a connection to, but every tool used in marketing (language, art, statistics, etc) has been misused repeatedly. In this case, ultimately the viewer is not supposed to learn anything about natural gas. The goal is to build GE’s brand as being associated with science and data. And (outside of data-wonks) it’s effective at doing so.
Pointing this out as bad data-visualization is fair – if it were meant as a visualization; but it’s not.
Geoffe,
Pointing this out as bad data visualization is definitely fair, because, contrary to GE’s true motives, they are presenting this as data visualization.
Later today I’ll publish another post in this blog that will poignantly demonstrate why it’s important to critique poor examples like this publicly. They are harmful in their influence.
You said that GE is using a tool to which I have a connection. What tool is that? If the name of the tool is “data vis,” as you perhaps meant by your parenthetical remark, I’m not familiar with a product by this name. I assume that GE has used multiple tools. For instance, I doubt that Ben Fry and David McCandless used the same tool, given the fact that Fry tends to use a tool called Processing that he developed.
Stephen,
Indeed – by “tool” I meant the field of data visualization. Just as language is a tool, GE is using data visualization (the field, the concept) as a “tool” to demonstrate: “hey, we’re science-y! We’re not just a washers, driers and light bulbs – we’re worldly and smart!”
I certainly won’t argue that as a means of conveying information, this thing fails (it’s awful) – I’m merely pointing out that it’s not designed to convey information in the first place. So holding it up as even a failed visualization seems like an exercise in futility.
In fact, to me, the more interesting point is that data-visualization-as-marketing is now very much main stream.
I’m approaching this as a marketer (who has interest in data visualization – thus my having read your blog). After browsing GE’s other “visualizations” it’s quite clear what the intention is: and it’s not to inform at all. So I don’t mean to critique your critique as much as I mean to suggest that outside the four walls of this website, the context is really quite different – and perhaps that warrants mention.
Geoffe,
Regardless of GE’s intentions, people look at these absurd examples and set their expectations about data visualization accordingly. Unlike you and I, many others don’t recognize these examples as absurd and ineffective. People are influenced by these bad practices, and this influence can lead to great harm. Until this is no longer the case, or much less the case, I’ll continue to expose these bad practices for what they are and do my best to show people a way that better serves their interests.
Dear Sthephen,
Just today I was reviewing the gas visualization and without reading your post thought to myself that it was hard to interpret the data.
And thanks for the legend switcher, i had the first impression that blue and white where displayed at the same time.
Cheers,
Manuel
Stephen,
would you consider it a ‘black-white’ world, where there is either data visualization to effectively inform on one side and data-art as data-driven artistic expression on the other side?
There also seems to be a gray area in between, where a hybrid solutions both informs a little, and are also created for its aesthetic beauty (not merely treating form as a means). Would you have a name for hybrid solutions with dual intentions?
Stephen,
agree that GE Digital cockpit is not the best dashboard ever, but: 1) it still has more information than the reviewed data visualization, 2) it is ten years old. Do you have any example of better dashboards made before year 2000? Would be interesting if anyone can collect historical pictures of “best practice” dashboards. I’ ve found some pictures of tableau de bord dating back to 1960 and then some really simple dashboard from +/-1990 (made in Lotus 1-2-3).
Jan,
Hybrids of data visualization and data art do exist. The problem with these hybrids, however, is that they compromise intentions on both sides of the spectrum, resulting in displays that fail significantly in both respects. For example, McCandless’ displays could be much more artistically pleasing and interesting if he focused entirely on aesthetics without trying to communicate specific information, and they could be many times more informative if he focused primarily on presenting the data as clearly and accurately as possible. Rather than setting clear goals for his displays, he arrives somewhere mid-way between artistic and informative without evaluating what’s he’s accomplished. He doesn’t seem to have standards for measuring the success of a display based on clear communication objectives. Because people have praised his work as it is, he churns out more and more of the same with abandon.
There is a difference between creating a hybrid of data art and data visualization on the one hand and creating data visualizations that are aesthetically pleasing on the other. I advocate the latter. The goal is to present information in a way that fully, clearly, and accurately informs to the degree that the audience needs to be informed, and does this in a way that is pleasing to the eye.
Attention to aesthetics in data visualization shouldn’t attempt what artists strive to achieve, but instead strive to draw people into the data in a meaningful way, without distraction. Ugly visualizations are off-putting. For instance, if a visualization looks cluttered or the colors play off one another in visually annoying ways, people won’t be drawn into the data. In this context, aesthetics in and of themselves don’t need to produce awe in the viewer (“My god, that’s beautifulâ€), but should invite the viewer to become engaged with the data in a way that allows its story to produce awe, or at least interest, understanding, and a desire for action.
Hi Stephen,
that’s a great reply which gave me some thought. And after some contemplation, I wonder what you think of the following 3 categories:
1) data visualizations: based on the research and theory of information visualization and visual perception. The visualization is used by a specific (private) audience, and the data is relevant for the audience (they have to make decisions based on the visualization for instance), and the visual representations are used as a means enhance cognition. Visualizations for business, science, serious analysis etc. fall into this category
2) hybrid visualizations: does inform but the audience is a more general public. The data is not extremely relevant for the audience (the don’t make decisions based on the data), but the one publishing the visualization wants to engage with the audience. The purpose is more about creating awareness and engagement. In this context visualizations can (but not necessarily so) be more experimental, less compliant with visual perception, but not misleading.
3) data-art: does not inform at all, and is an (aesthetic) artistic expression based on data
I think there is definitely a place for the second category, and you may be right that calling that (pure) data visualization might be confusing with the first category. I think visualization contests from Visualizing.org for example fall in this category, and also what large organizations like GE, WorldBank, UN, etc. attempt to do when they create visualizations for a large audience.
It may indeed be a problem that there is not really a name for the second category. If a visualization falls in one category, and you evaluate it using criteria (which may need to be defined more specificly?) from another category, then your evaluation will not be right. I think you may find this similar to your line of reasoning that calling everything a data visualization means that everything will be evaluated as a data visualization.
If this second category would exist, I think McCandless work would fall into this category. And I think his work should be judged with the rules and criteria for this category. And within this category his work could just as well be judged good or bad of course…
What do you think?
Hi Jan,
We seem to fundamentally differ in that you find value in what we’re calling “hybrid visualization,†but I don’t. I find it hard to imagine circumstances when any form of visualization that is supposed to inform would benefit from doing so in a way that aligns poorly with the abilities of human perception and cognition or augment them in some manner. Why should we inform the general public less effectively than smaller audiences?
You stated at the end of your comments that hybrid visualizations, such as those created by McCandless, should be “judged good or bad†according to the “rules and criteria for this category.†The problem is, this doesn’t exist as a recognized category and no rules or criteria for evaluating its merits have been defined. If practitioners such as McCandless would give what they’re doing a new name and declare their objectives clearly, then rules and criteria could be defined and we could judge their work accordingly. Until then, because they now call what they’re doing data visualization, I’ll continue to hold them to the standards of clear, accurate, and enlightening thinking and communication.
You defined “data visualization†as “used by a specific (private) audience,†but there’s no reason to limit it in this way. Data visualization can serve the needs of all audiences, including the general public. Much of my work is designed for the general public and many of my clients use data visualization to inform the general public.
I think your notion that the general public cares little about data and doesn’t use it to make decisions, but is drawn only to visually engaging graphics with little concern for the data that those graphics supposedly represent, suggests a low opinion of the general public. I think we often underestimate the degree to which people in the general public desire accurate and useful information as a means to expand their understanding, improve their lives, and do what they can to create a better world. If they’re not already interested in the information that’s being displayed, sexy graphics won’t change this—they’ll only create an appetite for sexy graphics. Shouldn’t we be nourishing people’s desire for information with meaningful displays?
You described “hybrid visualization†as “more experimental.†Actually, data visualization is much more experimental in the true sense than the work of people like McCandless. Universities and research labs all over the world do serious research, involving experiments, to improve and extend the reach of data visualization. I’m not aware of any actual research being done by people who produce the hybrids that we’re discussing.
You also mentioned that “data art does not inform at all.†I suspect that most artists would take issue with this statement. Art can inform and often does, but in a manner that speaks more to human emotions and usually concerns itself less with the communication of facts.
These are my thoughts.
I am so glad other people feel the same way I do about McCandless “data visualizations”.
After viewing a TED presentation by MsCandless, I recently purchase his book “Information Is Beautiful” – he seemed like a nice chap who perhaps had something to add to this rapidly evolving field.
I got the book and it is very “pretty†but his visualizations are incredibly poor at presenting information. In most cases they actually obfuscate the underlying data and impede proper understanding.
I got more and more despondent as I turned every page, I thought “these aren’t really data visualizations at all more like pieces of art parodying the work of the data viz worldâ€. I started writing critical notes on the pages of the book but it was just too easy to find fault, I got bored and returned to Tableau to cheer myself up ;-)
I am toying with the idea of sending McCandless a “charity pack†containing copies of Stephen Few’s books – he could use the education.
Hi Stephen,
thank you for your great reply, and making me aware of some of the wordings I used were not specific enough or correctly expressed my thoughts. My response:
I do agree that we disagree on finding value in hybrid visualizations. I think evaluating visualization from a research and infovis theory is not always the only perspective to evaluate a visualization.
Of course, I do agree with you that data visualization is not limited to a private audience.
I don’t have a low opinion of the general public! And saying that people want accurate information to create better lives is something I can only agree with. In fact, I think one of the reasons people like Dave McCandless are creating work like this is to inform the general public, and finding ways to reach more people to get their message across, to make them more aware (whether he does a good job is something else). And I also think that people like Dave McCandless are wondering if for example a bar chart, although visually effective, is the best way to reach many people.
Experimental is perhaps not the right word. Like you said, various universities and research labs are experimenting with new types of visualizations. And, correct me if I am wrong, they are mostly concerned with finding new ways of using information visualization as a means to enhance cognition and effectively inform, in a way that complies with visual perception, to communicate effectively. I don’t think hybrid visualization producing people are doing similar research, but I do think people are trying to find ways to create visualizations that appear to be more engaging than conventional statistical charts. Again, I don’t have numbers to back me up on how successful they are.
And of course, I also must full heartedly agree with you that data-art does often inform, but indeed on a more emotional level.
A final question: are there according to you situations possible where you could step away a little from creating the most effective visualization (from a perception theory point of view). Would reaching a larger audience for instance be a valid reason to do so?
Jan,
Bear in mind that I don’t only evaluate visualizations “from a research and infoviz theory†perspective. More than anything else, I evaluate them based on how well they work in the real world. My perspective is well informed by constant experience with people of all types and many organizations who use data visualization to do their jobs.
I’ve seen no evidence that David McCandless is actually evaluating the merits of various approaches to visualization. Have you? He’s simply doing what looks cool, without taking time to test it. Unfortunately, what he’s doing doesn’t work very well for known reasons. If he’s going to challenge evidence of what works and what doesn’t, he should do so thoughtfully.
In response to your final question, my goal in data visualization is always to give people the information that they need (that is, what will be most useful to them) in the best possible way. The goal is to communicate information in a way that results in understanding the truth as well as possible given the circumstances. This does to some degree require different visualizations for different audiences, but never in a way that produces inadequate or faulty understanding.
Stephen,
that’s a fair response I can agree with!
I am not aware of Dave McCandless evaluating the merits of various approaches to visualization. And for this particular piece, I must agree with you that I myself struggle to interpret these images, so yes, they do a poor job informing, and it’s impossible to compare…
Thanks for the discussion, it helps to shape my thoughts on this!
I can see your point about a lot of David McCandless’ work. I think it can be amusing but is often lacking in any information that can translate to wisdom.
But what do you think of the Billion-Pound-O-Gram?
I actually found that a pretty enlightening graph, especially when he used it at his TEDx talk and laid onto it the cost of the economic crisis (it took up the whole graph).
That worked for me. It does what most visualizations probably should do: provide answers by comparing data (How much has Iraq cost the US? Compare it to other things governments and non-governmental organizations spend their money on. That makes sense to me.)
Other examples of his work might be less useful, but Hans Rosling rightfully highlighted that visualization in his Joy of Stats program for BBC4. Have you seen this?
BTW, I really like your book “Now You See It” (I’m using it for a paper, among works by Colin Ware and Edward Tufte and many other sources that you use in your book).
You also might be interested in reading this post as a preparation of Eyeo Festival, written by Mike Migursky from Stamen Design. He also criticizes the same piece of work by McCandless: http://mike.teczno.com/notes/seeing-like-a-state-lying-with-pictures.html
Christian,
The Billion Pound-O-Gram, like McCandless’ other charts, fails in that it presents values in ways that we cannot easily compare. We simply can’t compare the areas of rectangles very well. If this were a case of needing to compare more values than a bar graph could handle, using rectangles of various sizes would be an acceptable compromise. This doesn’t apply, however, for a simple bar graph, with the bars arranged from the biggest to the smallest, with color to break the bars into separate categories (spending, fighting, etc.), could present these values beautifully and in a way that we could easily compare. To demonstrate this, I’ll write a separate blog post in a day or two with examples.
Two parts of Rosling’s “Joy of Stats” are unfortunate: (1) the fact that he featured McCandless’ work, which is hardly a respectable display of statistics, and (2) the demonstration that Rosling did using one of his signature animated bubble charts, which was severely undermined by silly video effects. You can read my critique of the latter in this blog entitled “Rosling–Where’s the Data?“
Steve, you are my hero! Whenever someone asks me why I am so passionate about getting the visualization correct, I just point them to your blog. I enjoy the comments and your responses immensely, as you prove to have a “real conversation” with your audience. Whenever I am mentoring a younger colleague in the ways of BI, Stephen Few and Edward Tufte are required reading! Thank you so much for the leadership you have provided us mere practitioners!
Thanks Tricia. I appreciate your kind words of support. Together we can make things better.
Point(s) taken, Stephen.
I’ve been following #eyeo on Twitter all week. (sigh) I hope to be smart enough to follow this advice some day (which was presented as the last slide):
http://post.ly/2IwL0
Stephen,
Wow… I’m a noob when it comes to data viz, but your explanataions now make me feel like an expert, even though I’m far from it. I guess what I’m saying is… wow… :)
This is very well written and informative. We should all know that businesses like this have been doing this forever; confuse the little people and make them only understand what we want them to.
Great job standing up for us.
Stephen,
I am taken by your initial PS. I live in Western NY state where there is a current, important policy/political debate re: natural gas extraction via “hydraulic fracking.” I have my views on that matter but will leave them aside. I also grew up a company town that GE abandoned as an environmental wasteland with PCB (etc) pollution of waterways. And, of course, NY State has spent decades cleaning up the Hudson River, in large measure degraded by GE policies.
So, there is some reason for passion in this particular matter – GE is a company with a less than stellar environmental record, mucking up the informational ecology on the matter of energy production and policy. The data-graphics you call attention too are not mistakes. They seem, from my vantage point, to be designed to obfuscate. It would be wise to ask whether GE has a stake in gas exploration …
At a more general level, you might go back and read the exchange in the 1920s between Walter Lippmann and John Dewey on democracy and its vicissitudes. Lippmann bemoaned the ill-informed, easily mislead electorate and recommended a mode of technocratic rule by experts. Dewey responded by fingering the difficulty the public has in seeing shared interests and predicaments in an increasingly complex political-economic-social world. He insists, plausibly enough, that our problem is how to make reasonably technical information about reasonably complex matters accessible to common citizens. So Dewey insists on the importance of being able to artfully communicate the results of scientific (including social scientific) inquiry in ways that the public can use and debate. [The final page of Tufte VDQI on clarifying the complex seems germane here.] But it is important to note that Lippmann too is preoccupied with whether the members of the public can see or envision complex matters in clear ways. I guess the point of this ramble is that your preoccupations are central to the operation of democratic politics and that that is so regardless of one’s political/partisan views.
So, if one is committed to democratic politics generally or worry about the relation of “experts” to common citizens, there is reason for passion.
Best,
Jim
PS: I teach political theory and one of my students excitedly sent me the link to your web site and copied in a bunch of his peers. So you are reaching a younger generation – perhaps our best hope!
You’d probably be interested in this review of current information visualization by the statistician Andrew Gelman and mathematician Antony Unwin:
“Infovis and Statitical Graphics: Different Goals, Different Looks” (unpublished, June 11, 2011)
http://www.stat.columbia.edu/~gelman/research/unpublished/vis9.pdf
More particularly they discuss their perceived difficulties with most popular information visualizations and suggest a dialogue as to defining the appropriate goals of displaying data (or better, Tufte’s term “quantitative information”). Some very thoughtful insights. These various visualizations for GE could well be critiqued in the context of this paper.
I do not understand Stephen Few’s anger towards use of less data. The most important thing to be kept in mind while representing data is who is this data addressing to and what is that the consumers of this graphics are looking to gain from this. I do not believe GE is presenting high level strategic data so that CEOs of fortune 500 companies can analyse these graphics and take tactical and strategic decisions.
GE is presenting the simple data in a fun way. Thats it. This data is not for high level analysis of any type. So why the fuss?
GE might be internally using a different set of charts & graphics to drive business results. It is sharing some simpler stuffs with you. Do you expect GE to share what its CEO is using?
Your article is titled “Does GE Think We’re Stupid?”, but it seems you say GE should be stupid and share with us the valuable data it uses.
~XIX~,
The problem has nothing to do with the quantity of data or with a legitimate unwillingness by GE to share confidential information. The problem is that what they did share, they presented poorly. If the information is useful to the general public, then it should be displayed in a way that is meaningful, accurate, and clear. I believe that you overestimate the intelligence and interests of most CEOs and underestimate the intelligence and interests of the general public.
What is not obvious, if you haven’t worked directly with CEOs of large corporations, is that information is often presented to them in impoverished ways. Not as silly infographics of the McCandless variety, but as horribly designed, information-lite tables and graphs. Whenever my students complain that they are being forced to display information poorly in their organizations (e.g., bright spinning pie charts and gauges), it is almost always the CEO that they point to as the worst offender. It is doubtful that an organization such as GE that presents information to the general public in ineffective ways is presenting information to one another internally in ways that are much better.
I agree with Stephen. The proper term for this type of advertisement is “misinfographic”.
steve, i dont think you’ll ever be out of work or things to write about as long as ppl keep visualizing data :)
Does GE Think We’re Stupid?
Simple Answers to Simple Questions:
Yes.
* * *
Actually, it’s more subtle than that. They’re MAKING us stupid; it helps their business model.