Teradata, David McCandless, and yet another detour for analytics

While catching up on the latest data visualization news yesterday, I ran across a blog post by Mario Bonardo of Teradata Corporation. Bonardo’s account of a recent Teradata event in Europe caused me painful dismay. He praised a presentation by infographic artist David McCandless, suggesting that what McCandless does is precisely what needs to be happening in the realm of analytics. Bonardo wrote: “Being asked about the differences between traditional information graphics and his own ones, McCandless said he is aiming to remove as much irrelevant information as possible to get to the core of things, discover new correlations and challenge traditional views.” My dismay does not stem from McCandless’ words, but from his actual practices, which don’t deliver what he claims and certainly don’t point the way to a productive future for analytics. Here are two typical examples of McCandless’ work from his website:

[Notice how hard you must work to figure out what the various mountain ranges represent. Don’t bother actually trying to determine or compare the values, because it can’t be done.]

[I bet that any one of your reading this can think of a better way to display these values for easier, more accurate comparisons.]

You can also read my critique and redesign of a well-known visualization by McCandless—”Colours in Cultures”—in my March/April/May 2010 Visual Business Intelligence Newsletter article titled “Our Irresistible Fascination with All Things Circular.”

Stripping away the irrelevant—McCandless’ stated goal—can only be done once you’ve found a way to display the relevant. Too many of his visualizations display information in ways that hide much that’s relevant and essential, leaving little of value for the viewer to see. McCandless rarely chooses forms of display that our eyes and brains can perceive with ease and precision. He selects what will appeal superficially to the viewer (lots of circles, swirls, and vibrant colors), not what will most effectively express what’s essential and meaningful. His displays rarely draw viewers into the data in a thoughtful way, but entertain in a way that delivers a simple message, which is often anemic when compared to the richer, subtler, and more complex stories that live in the data.

McCandless is a creator of infographics: combinations of words and graphics that are designed to communicate specific messages. On those rare occasions when his infographics lend themselves to data exploration and analysis, they do so in limited and awkward ways, asking viewers to perform data surgery with blunt instruments. Infographics of this type attempt to tell a story. In McCandless’ case, the stories that they usually tell, if communicated in words alone, would require only a short sentence or two. They make a simple statement in a way that looks lighthearted and fun. As such, they invite viewers to accept the message superficially, not to explore or contemplate deeply. This is not the true realm of analytics.

According to Bonardo, the other members of the panel in which McCandless participated at Teradata’s event made statements in support of his approach, but their comments make me wonder—assuming that Bonardo represents their intentions accurately—if they’re actually familiar with his work.

Stephen Brobst of Teradata said: “Whereas traditional reporting tools were providing answers to pre-defined questions, todays analytical environments offer opportunities to think out of the box and help to find new questions to develop cutting edge business strategies.” Most of today’s so-called analytical environments fail to do this, with only a few exceptions. I’ve seen no evidence that McCandless’ work attempts to address this opportunity at all.

Oliver Ratzenberger of eBay said: “We offer specific visualization classes for our data analysts. We really want them to play with data so that they may actually find new nuggets.” I know this to be true, because I’ve taught information visualization courses privately for teams of analysts at eBay. Trust me when I say that I did not promote the practices of McCandless, which don’t support the rich exploration of data that eBay needs.

Daniel Rodriguez Sierra from Telefonica said: “Human beings are very good at recognizing patterns. The challenge is to display data in a way that it allows human beings to make full use of this natural strength.” How true, but McCandless’ work usually disregards human perceptual and cognitive abilities.

There is a world of difference between the simplistic (that is, overly simplified), artistic infographics of McCandless and the engaging, thoughtful interaction with data that analytics requires. If the business intelligence (BI) industry is now choosing to swap its primitive, dysfunctional charts for the artistic approach of McCandless, we’ll be in for several more years of BI failure. To date, Teradata has partnered with some good data visualization vendors (for example, Tableau) and some embarrassingly bad data visualization vendors (for example, FYI Corporation, which is now defunct, and Bis2). Featuring McCandless’ work does not bode well for the direction Teradata is heading.

In closing these remarks, I want to challenge the merits of the theme that Teradata chose for this event: “Innovate to Differentiate.” Innovation is useful when it leads to a better way of doing things, not just a different way. Innovation in and of itself is highly overrated. There’s been relatively little real innovation in BI for many years. Much of what’s been marketed as innovation is old news repackaged, in the hope that no one will notice. Rather than trying to innovate in the realm of analytics without the required expertise, it’s time for BI companies, consultants, and thought leaders to first learn the basics of data sensemaking and communication—analytics—and only afterwards to try their hands at creating something new.

Take care,

30 Comments on “Teradata, David McCandless, and yet another detour for analytics”


By Andrew. April 15th, 2011 at 1:12 pm

If not for the extra label on the “chart” itself, I’d have thought based on the colors that the global media was crazy scared of asteroid collisions around ’05-’06. Good article!

By Alex Kerin. April 16th, 2011 at 6:35 am

As usual, you pull no punches – an approach that is appreciated. While there may be room for pretty looking information displays as an art form, the fascination with McCandless’ work has always left me confused. I move from confused to concerned when his work, BI, and solving business problems are mentioned in the same sentence.

Innovation in business data delivery comes from a thoughtful and deep understanding of the business problems we are trying to solve, ensuring we highlight the issues (and the bright spots), and tailor the information to the end user. Such is the sad state of data delivery from most of the top vendors that ‘thinking outside of the box’ could be as simple as providing more than 5 pieces of information on a single dashboard.

By Michael Coyle. April 17th, 2011 at 2:04 pm

I agree wholeheartedly with you analysis. While I am not a data visualization expert, I’ve studied it at a university level. One of the most simple ideas is that knowledge of human perception in choosing colour can assist you in not accidentally lying to the reader. For instance, use of a rainbow palette is a cardinal sin in data analysis.

It would be an anathema for an artist to be told what colours to use, and Mr McCandless shows every sign of being an artist, choosing colours for their aesthetic value and not for analytical worth.

By Stephen Few. April 17th, 2011 at 11:01 pm

You might be interested in reading and participating in the discussion that my comments have generated in the blog flowingdata.com.

By Stephen Few. April 18th, 2011 at 10:09 am

If you’re interested in seeing an example of the trend in data visualization that stems from the influence of David McCandless and others that promote similar approaches, watch the YouTube video that introduces entries to Google’s Data Viz Challenge.

By DR. April 18th, 2011 at 10:15 am

First, thanks for taking the time to post your thoughts.

But I believe you’re missing an important nuance. You say the following:
“There is a world of difference between the simplistic (that is, overly simplified), artistic infographics of McCandless and the engaging, thoughtful interaction with data that analytics requires.”

You’re explicitly differentiating between “overly simplified” and engaging/thoughtful. You’re creating a split where none exists. Often simplifying is the *best way* to be engaging and thoughtful. In fact in a previous post (http://www.perceptualedge.com/blog/?p=903) you argue as much: “eloquence through simplicity”.

I don’t know McCandless, but I’m familiar with his work – and if anything it’s engaging – often because it’s over simplified. You can argue there are technically better ways to present information, and you’d be right. But are there (obvious) more engaging ways? Not often. And if a presentation is not engaging, it will have no audience, in which case the work is just an exercise. Perhaps a personally useful exercise, but not one worth further pursuit by others.

Another item I take issue with is your comment:
“[McCandless’s graphics]…invite viewers to accept the message superficially, not to explore or contemplate deeply”
I wonder whether you have read the comments that often follow his graphics? (Either on his website or elsewhere.) Quite often the conversation is quite contemplative. Could his graphic have answered more questions? Perhaps. But they often spur good conversation. I personally feel it’s much more valuable to prompt dialog than pretend to know all the answers.

Anyway – my 2 cents there. Thanks again for your contribution and for providing a forum for conversation.

By Stephen Few. April 18th, 2011 at 10:48 am

DR,

I actually chose my words quite carefully when I described McCandless’ work as “simplistic (this is, overly simplified)”. I advocate “simple”, not “simplistic” displays. By simplistic, I mean that either part of what’s essential to the data’s meaning has been stripped away, thus overly simplifying it (that is, robbing it of essential complexity) or the data has been expressed in a way that looks simple on the surface but is actually quite difficult to interpret. So yes, I am differentiating between “overly simplified” and “engaging/thoughtful.” Simple (rather than simplistic) displays pare the essence of the story down to what’s essential and presents that in a way that is as accessible as possible.

There are definitely more engaging ways to tell the stories that are contained in many of McCandless’ infographics. The winners of the recent Malefiej 19 competition that I recently blogged about are fine examples of thoroughly engaging and beautiful infographics. They engage people in the information in ways that fully inform and invite rich thinking about it.
Presenting data as clearly as possible is not pretending to know the answers. It respects people’s intelligence by assuming that they’re actually interested in the data, not in being entertained by the pretty shapes and colors.

It is true that the very act of discussion, whether in the context of a blog or elsewhere, invites thinking and interaction. When discussion about the data is prompted by one of McCandless’ visualizations, to what degree does the infographic inform the discussion? McCandless is providing a service by creating an opportunity for discussion about important issues. How much richer could these discussions be, however, if he presented the data more effectively? You can use a Tarot deck to spark ideas, contemplation, and discussion, but the cards aren’t providing useful information, they are merely inviting people’s minds to explore ideas. I want the realm of data visualization to provide rich content for the discussion, not just an opportunity.

By Rob Beal. April 18th, 2011 at 12:07 pm

The brutal honesty of your review here is refreshing. Your review here captured my thoughts much more eloquently than I could have stated. I’ve seen the work on McCandless’s site, and I have to say it baffled and confused me…art heavy and data light. He took elements of Tufte, twisted and distorted them, and came up with something that looks pretty (sometimes) but fails to convey complex data. Effective information graphics require bridging the artistic with the scientific, the right brain and the left brain. McCandless has the right brain part down pretty well but fails to bridge to the analytical. Companies that try to follow his guidelines (if you can call them that) will end up with confused employees and less effective communication.

By Sarah. April 18th, 2011 at 12:26 pm

Interesting. McCandles’ work seems designed to sell – more the sort of thing you put in a brochure to peak a person’s interest, or to communicate one specific, pre-defined idea. The choice of colors, shapes, and highlighted data force a specific interpretation of the data.

However, if your audience is analysts and you want those analysts to come up with their own interpretation, you need to find a way to present more of the data (clearly), without elements that will psychologically or visually sway the interpretation.

By Stephen Few. April 18th, 2011 at 12:43 pm

Sarah,

As you’ve pointed out, McCandless’ work is certainly not designed for analytical purposes, but the truth is that, even for the purpose of communicating “one specific, pre-defined idea”, his design choices are usually ineffective. It is often appropriate to design a display to tell a specific story, but McCandless ignores many of the principles for doing this well.

By Andy Cotgreave. April 19th, 2011 at 2:26 am

This a great debate, Stephen. The simple/simpistic aspect of this debate got me thinking: how would McAndless rework Charles Minard’s Napoleon 1812 graphic:
http://www.thedatastudio.co.uk/blog/the-data-studio-blog/andy-cotgreave/if-mcandless-reworked-minard

By Andre. April 19th, 2011 at 11:13 am

He designs for clients not for people.

By Stephen Few. April 19th, 2011 at 11:43 am

Andre,

I’m not sure what you mean. Are you saying that McCandless only creates data displays that are commissioned by clients, based on their specific requirements, and that this limits him in some way?

By Andrew. April 20th, 2011 at 1:52 pm

Stephen,

After reading through some of the comments over at flowingdata.com, it seems to me that this would be a great opportunity to create some infographics for your “Examples” page to demonstrate just what McCandless’ data displays are lacking, and what they could be. Trying to discuss poor design only seems to be met with stubborn insistence that “nontraditional” is enough to be “compelling”, or that “overly simplified” is the only way to “tell a story.” If you demonstrate that the data could be displayed in a thoughtful way that more efficiently “tells the story” and can still be just as engaging, people may be more hard-pressed to disagree.

What are your thoughts on this?

By Stephen Few. April 20th, 2011 at 2:23 pm

Andrew,

I’ve done this with similar work by others, but only once have I taken the time to recreate one of McCandless’ displays. It doesn’t appear on the Examples page of this site, but in the newsletter article titled “Our Irresistible Fascination with All Things Circular.” I agree that it would be useful to do this with other examples of McCandless’ work as well, and I’ll do so if time permits. This involves more work than is obvious, because I wouldn’t be able to base my design on the little bit of information that McCandless reveals in his displays, which usually don’t present values in a way that can be decoded, but would have to find the data on which his display was based and explore it to find the important stories and actual values.

By Sean. April 21st, 2011 at 7:42 pm

I’ve often seen David M.’s work. I work in BI field and deal with data everyday. His work has often puzzled me. I often have to spend quite a bit of time to figure out what the heck is he trying to show. I was thinking that may be it’s me who is not “getting it” and missing the value of his work.

Thank you for this article. Now I know I’m not alone. As far as people including the higher ups in most companies eye candy equals value. I’ve come across that too many times.

It’s a sad state of affairs in BI World.

By Andrew. April 22nd, 2011 at 10:44 am

I’m not trying to beat a dead horse, but I had one other comment on this topic. There is an interactive version of the Mountains Out Of Molehills that I think you should look at (if you haven’t already):

http://www.informationisbeautiful.net/play/mountains-out-of-molehills/

Improvements include the ability to level everything at the bottom of the chart, and to highlight legend items (making it easier to see mountains normally obscured by other mountains). Still no numbers on the value scale. But the greatest mistake of the interactive version is the “Scale by total deaths” option:

1. It scales the whole chart by total deaths, not timeline deaths; this is very misleading. Do sudden increases in related deaths correlate with spikes in fear or not? Did the same number of people really die from Killer Wasps every summer since 2004? We’ll never know based on this graphic.

2. Even worse, consider what happens to the message by scaling the chart by total deaths directly. I would think media scare stories with few or zero related deaths are critical to the message, but scaling by total deaths removes them completely; only fears that were apparently validated by a whole slew of casualties remain. Is that the story this infographic was supposed to tell?

It seems to me that while the ability of McCandless’ infographics to inform is important (as it is being debated over at flowingdata.com), it’s not the only issue here. The “Scale by total deaths” option not only serves no purpose, but distorts and confounds both the data and the message to the point that its mere inclusion suggests a serious failure of the author to understand either. Without a clear understanding of the data and the message, I don’t think anyone can make an effective display.

By Stephen Few. April 22nd, 2011 at 10:59 am

Andrew,

Thanks for these insightful observations. While focusing on the visual design of quantitative data display in this discussion, we’ve disregarded other factors that also influence the effectiveness of data visualization. Presenters must know the data and they must have a basic understanding of statistics, in addition to an awareness of how people perceive different forms of graphical display.

Too many examples of so-called “data visualization” today fail in each of these respects, often because their designers have never taken time to learn these skills. When these skills can be easily learned, there is no excuse for ignoring them. Sometimes people argue that they’re inventing new ways of visualizing data that don’t follow the old rules. This argument is naive and potentially harmful. Whenever you get involved in field that is new to you but well known of others, your potential contributions are not hindered by learning what the experts already know and then innovating from there. In fact, your success is magnified and streamlined by avoiding known mistakes.

By Stephen Few. April 24th, 2011 at 10:41 am

I need to correct an error in my description of David McCandless’ background, which was based on an assumption that turns out to be incorrect. McCandless does not have a background in design. Apparently he has worked for most of his professional life primarily as a programmer. This clarification explains a lot. McCandless has probably never studied human perception or cognition, which is not a surprise given the typical failures of his designs. What has always puzzled me about his work, however, is how that, even from a graphics design perspective, his visualizations are often not particularly pleasing to the eye. For example, his color choices often seem arbitrary and not particularly well matched. I now understand why. His work as a visual designer is recent. He has much to learn. In the meantime, however, he has been popularized as an data visualization expert, which is where the problem lies. His diagrams should not be seen as exemplars of data visualization, but as the learning exercises of a relative novice. While being promoted as an expert, however, will he take the time to develop the skills that he lacks, or will he spend all of his time producing more of the same and presenting it to the world? For his sake and the sake of the world that’s influenced by his work, I hope for the former.

By Anders Halvorsen. April 25th, 2011 at 3:22 am

Hi Stephen.

I did see this TED talk by McCandless, where he states that he worked as a journalist before getting into design. He shows his CV as a infographic at 07:50 (but as most of his graphics, it’s not very informative). I noticed is that he claims that “When I started designing, I discovered an odd thing about myself, I already knew how to design” (08:15). This attitude does not sound promising for your hope that he will take the time to develop the skills he’s lacking.

The first graphic in your article is discussed here (at 3:00) and it actually contains some really interesting stories. Too bad the graphic fails to commute any of them clearly.

There are a couple of other charts there which clearly illustrates some of the failures in McCandless graphics. At 17:25 there is a chart of the CO2 emission from the Icelandic volcanoes. I did not see the size difference before he added the numbers to the chart, and he even manage to mess up a simple subtraction so that one of three numbers are wrong (so much for quality control and caring about the data); a mistake that would have been easy to spot with a properly designed bar chart.

At 12:30 there is a supplement chart where he compares popularity and evidence for different supplements. A biplot would have shown the relationship between the two, both accurate and easy to understand. But McCandless chose to plot how strong the evidence as the y-axis and the size of a bubble to correspond to popularity. I’m clueless about what the x-axis and colors are coding.

By Sathya. April 25th, 2011 at 4:05 am

Hi Stephen,
(I wanted to post in the heated discussion at FlowingData, but the comments seems to be closed. I wish to get your comments on this Stephen)
Easy to use visualization techniques of Stephen Few – as he most of the time points out ‘it took me just 15 minutes’ or ‘within 10 minutes, I did this’ – I find this more appealing for audiences who want to interpret data into readable or communicable format. McCandless is simply not replicable (MS Excel is the only graphic program I could afford to use!).
And to whoever who gave the Radiation dosage example by McCandless – I don’t think that is the best example. Please it has only one dataset, i.e., the radiation dosage and its source. I can simply use a table, ranked ascendingly or descendingly will work (why do you need the Cone or the rainbow colors anyway? And it is also not print friendly graphics). Most of the time, table is the best form of presenting the data – in fact Stephen Few uses this in one of his examples (http://www.perceptualedge.com/example4.php).
In his TED talk, McCandless says it is not right to compare absolute value of ‘Defense expenditure’ among nations, but ‘% of expenditure to the total GNP’ is a more apt data to compare – I think there lies the genius of McCandless (though it was obvious, many failed to make the connection) – but he lost me when he presented the infographics for that – again an ascending or descending or a need-based ordered bar/column graph would have sufficed. I think Mr.Few would have loved to have that in his examples pages.
Again repeatedly it has been pointed out that ‘bubble graphs’ are beautiful, but it makes visual comparisons hard and mostly misinterpreted (Few and Tufte had repeatedly argued on that).
I think that many a times ‘infographics’ is simply over-rated. I think that purpose and clarity should come first. Thank you!

By Philip Hodges. April 26th, 2011 at 7:08 am

Hi Stephen

I was going to put this link on Flowing Data in reply to the last comment made regarding an interactive version but the comments seem to have been closed off.

Anyway, last month I created an interactive version of “Colours in Cultures” which you can view here:

http://lab.zoho.co.uk/lab/interactive-colours-in-culture/

I don’t really want to get involved in the argument as it is not my core field and the interactive mentioned above was done purely for fun, not for profit but I thought you might be curious to see it.

As another aside I also found it amusing when I noticed yesterday that your books sit next to David McCandless’ book on the shelf by my desk.

Have a good one.

Philip

By Stephen Few. April 26th, 2011 at 5:57 pm

Hi Philip,

The fact that McCandless’ book sits on your shelf next to my books illustrates my concern. Too many people today see McCandless’ work as the way data visualization should be done, without discriminating between effective design practices that have grown out of deep study and experience from the experiments of a novice. I believe that taking the time to learn best practices springs from a greater appreciation for information and the rich stories that dwell within it.

Your interactive version of McCandless’ “Colours of Cultures” diagram make it possible to explore the data more easily and meaningfully, but a matrix of columns and rows with the same functionality would work much better.

By Philip. May 4th, 2011 at 6:34 am

Hi Stephen.

Thanks for your feedback but I must confess I am not sure that I follow your logic that my comment about the “books on my shelf” does actually illustrate your concern.

Surely the fact that I have multiple books on the subject representing different points of view is more inclined to be evidence of the opposite i.e. that I am taking the time to explore and learn from a multiple set of resources rather than just adopting one “popular” approach as a best practice.

Though to clarify it was just meant to be a throw away comment on what I thought was a funny coincidence. I wasn’t trying to indicate anything deeper.

Anyway the main reason I am back is I was curious to see if an interactive grid version would “work much better” so last night I amended my interactive adaptation so that the user can now switch between two views.

Once again just doing this for my own “fun” but thought you might be interested in viewing the results:

http://lab.zoho.co.uk/lab/interactive-colours-in-culture/#/?shape=square

Philip

By Stephen Few. May 4th, 2011 at 11:37 am

Hi Philip,

There are a few ways in which your new tabular design can be improved. Eliminate the legends by labeling the columns and rows directly. If labels next to the rows seem to be stacked too tightly, you can free up some vertical space by either eliminating what you currently show above the table or by moving it to a different location, such as to one of the sides. You can still provide the ability to select color meanings or cultural groups to highlight selected data, but do so by allowing people to select the column or row heading rather then items in a separate legend. When you eliminate these legends, there will be no reason to label columns or rows with letters (A, B, C, etc.) or numbers, which are meaningless. The only set of values that need to be defined as a legend is the list of colors. Make the means of unselecting items more obvious and easy. Once something is selected, it isn’t obvious how to unselect it without selecting another item. For exploration purposes, you might want to provide the means to sort the columns and rows in various ways.

These are my initial thoughts. Have fun.

Steve

By Andrew. May 9th, 2011 at 9:51 am

@Philip

If I may offer some suggestions as well:

Looking at the display for a couple minutes, the only message I get from it is that “many colors have many meanings in many cultures”, which shouldn’t be news to anyone. After more time goofing around with it, it becomes nothing more than a “fun” reference table with no real message. Stephen’s grid version suffers from the same flaw as far as I’m concerned. The viewer ends up doing all the work to make sense of the information, and personally I don’t like work that much.

What’s the point? What story are you trying to tell? You want to have a lasting effect on the viewer, don’t you? You want them to take something away from seeing your work, right? It should be critical to the designer that the display make an impression quickly, because most viewers won’t pay attention to a graphic for any more than a few minutes before moving on. What do you think they will take away from the “Colours in Culture” graphic?

My advice: make the infographic-as-it-is secondary (for further investigation; drilldowns, for the dashboard-enthusiasts) and instead focus on the patterns and outliers. Emphasize meanings that are mostly consistent in color across cultures (Ex: Passion, Evil). Or meanings that vary greatly in color between cultures (Ex: Happiness, Love). Or colors that have a large variety of meanings (Ex: red, blue; which occurs more often, and are you really going to force me to count them by hand to figure it out?). Or meanings that have little or no color-representation (Ex: Gratitude, Calm). There are many interesting stories to be told here, but the graphic isn’t telling any of them. Instead, the viewer has to spend considerable time digging to discover them.

The most important question any designer should ask himself/herself when designing any data visualization is “Why?”. Why am I creating the display? Why is the viewer looking at it? If the only answer you can come up with is “because it’s beautiful”, then the graphic is likely to fail to inform most of its audience of anything.

By Stephen Few. May 9th, 2011 at 10:30 pm

Andrew,

I appreciate your comments and wholeheartedly share your belief that data visualizations must be designed with particular intentions in mind. Based on this very point, the fact that my tabular arrangement serves only as a reference table is not a flaw, if that is its intent, which is the case. Otherwise, an alphabetical arrangement of the color meanings would serve no purpose. McCandless’ original display did not tell a story, other than the one that you mentioned: colors mean different things in different cultures – not a surprise. At best, it could serve as a reference, but it couldn’t do that well when arranged in a circular fashion. With the addition of some interactive capabilities, such as the highlighting that Philip included and the sorting that I suggested, my tabular arrangement could also serve as a platform for exploration and analysis, albeit limited, because additional data would be required to understand the reasons why certain colors have particular meanings.

Data visualizations that are intended for exploration and analysis must be designed quite differently from those that are meant to tell particular stories. The suggestions that you made to Philip are appropriate for storytelling, but not for exploration and analysis, because they would limit the reader’s view to the stories that you’ve already identified. Providing the means for viewers to discover these stories and perhaps others as well on their own is often the intention of data visualizations.

By Andrew. May 10th, 2011 at 10:33 am

Stephen,

Fair enough; I did not mean to imply that a tabular layout is inherently-flawed by being nothing more than a reference table, especially if that was its intent. However, while that is the case in your tabular version of “Colours in Culture”, do you think that is the case with McCandless’? If so, then why?

Personally, I’m not so sure about McCandless’ intent. His supporters suggest that story-telling through “data art” (as much as it pains me to use those words) is the purpose of his work. Not to let them put words in his mouth, but let’s assume for the sake of argument that they are correct. I do not believe the “Colours in Culture” graphic serves that purpose, circular or otherwise. If it was intended to tell a story, then it clearly was not designed with any particular story (or audience) in mind. This is what I meant when I said the graphic is flawed; circular or tabular, it fails to achieve the goal that McCandless’ supporters are claiming.

Regardless, I do believe that data visualizations can be designed for exploration and analysis and still tell a story. My recommendation to Philip is to focus on the story by emphasizing patterns and outliers that support it. But I did not suggest removing anything; I certainly wouldn’t want to limit the user’s ability to discover other stories on their own. I think the graphic should be able to “captivate the audience” (as McCandless’ supporters claim he is doing) through a compelling and quickly-identifiable story while still allowing for exploration.

By Stephen Few. May 11th, 2011 at 2:18 am

Andrew,

Everything that you’ve said here makes perfect sense. Thanks for the thoughtful comments.

By Chuck Pirrello. June 9th, 2011 at 11:01 am

As I ponder this debate over chartjunk, I wonder if it’s more of a debate between attention grabbing vs. delivering useful content. The purpose of a graph is to inform and/or persuade. It can persuade in one of two ways – either showing data in such a way that people can immediately draw conclusions and recognize the urgency of action, or by grabbing attention so dramatically that people cannot but stop and take notice. If your intent is to grab people’s attention, visually appealing graphs adorned with unnecessary decoration will certainly do the trick. But the more serious onlooker will certainly want to examine the details more closely, especially if an important decision hangs in the balance. A more casual observer, who is constantly bombarded by visual stimuli, will more likely accept the message based on its face value and question very little since there may be little at stake.
However, this is not always the case. I see many chartjunk graphs that are laced with the bias of their creators. This presents a more sinister issue, for if casual viewers become accustomed to such graphs and don’t interrogate the data being presented, they can be terribly misled. I’m not saying that transparent graphs cannot mislead, but it seems the more decorative ones provide the author with more ways to lead a viewer toward an invalid conclusion.