A Defense of the Unknown in Infographics

We’re inventors – we’re creators. And that’s the most important thing about what we do. And I think we should welcome failure every once in a while.
Hannah Fairfield – NYT Graphics Editor, Malofiej 18, March 2010

Last year at Malofiej, one of the major awards ceremonies for infographics in journalism, The New York Times took home ‘Best in Show’ for their work on box office receipts from 1986 to 2008. I’m sure most of you saw it. It was non-traditional. It was an adaptation of Lee Byron’s streamgraph, which he had previously applied to last.fm music listening habits – a smoothed stacked area chart at the core.

What followed was a lot of back chatter among the infographic community. Many didn’t like the interactive at all, despite winning an award voted on by peers. Some called it one of the worst graphics NYT had ever published, that it was too complicated for readers, and that it was too hard to read.

This surprised me.

It’s something I’ve come to expect from academics and the stat crowd but not from graphic departments that report the news. I had the impression that they were more open-minded, but I guess not all of them are.

This, from Hannah Fairfield, a New York Times graphics editor, is no doubt a response to the haters at this year’s Malofiej.

I obviously strongly agree. Although Hannah sort of implies that the streamgraph was a failure. I’d argue that it was a success. Hundreds of thousands of people, millions maybe, engaged with the box office data and there’s no obscene misrepresentations. Were the patterns too complex to understand for some people? Yeah, probably, but how else is the general public supposed to learn? They’ll get there eventually.

Your turn. Does the box office streamgraph work?

[via VisualJournalism]


  • It works. One of the reasons that it works, is that it looks so marvellous that people actually study it, to find out what on earth it is. Doing so, they actually learn something. Quoting a colleague of mine: “It’s almost erotic!”. This may not be perfect, but it makes more people stop, think and learn.
    It gets the message across to more people, which to me seems like a good idea.

    It also leaves the readers with two thoughts: “summer/christmas is the best time” and “what goes on inside Amanda Cox’ head?”

  • I’m no fan of that infographic, and I’ve posted my analysis in public, why I don’t think it works. I’ll invite you to join that discussion anytime.

    I find it noteworthy how Lee Byron reacted to my analysis; compared to how you in this very short post find the space to call the people you disagree with for haters and not open-minded – and the valid discussion following the award for backchat.

    You say that there is no obscene misrepresentations in the graphic. I’ll disagree as I think I pointed out very clearly that the sexy and curved shapes are telling the completely wrong story about the individual movies.


    • @Gert – I read that too, and those are good points in that critique. but i still don’t think the curves are a huge misrepresentation. like lee says, it’s a product of interpolation. it’s estimation, and that’s going to happen when you try to fit any model to a dataset. It’s never going to be perfect.

      I don’t think I was referring to you in this post. Critiques and discussion are always good things.

      However, those who go around smashing the graphic in every talk and calling it the worst ever, I think, are out of line. Besides AC is the nicest person ever. Why would anyone do that? :)

    • I fear some of your valid points are getting completely obscured by your high horse.

  • This is not an original thought but visualisation is the selection and design of a display that communicates a particular message about some data. It is this message element which opens up most interpretation and debate about the effectiveness of a visualisation. It cannot be everything to everyone, yet this ‘everyone’ will be capable of seeking an entire spectrum of different messages.

    As Nathan says, the volume of debate and discussion about this graphic demonstrates the reach it has achieved. If nothing else, it has put bums on seats. It appeals on an aesthetic level and stands out as something novel and different and, as communication objectives go, this has to be the first requirement.

    Occasionally, I find the academic principles of information design (and their basis in understanding visual perception and processing capabilities) do tend to weigh heavy, even though they are the principles I generally swear by – you must create awareness and grab attention but the follow-up objective should be to impart insight. And this is where we have to remind ourselves, as viewers, that not every message we would hope for can be incorporated into a single graphic.

    As I’ve discussed here (http://www.visualisingdata.com/index.php/2010/03/if-only-the-egyptians-had-built-bars/) we have to find a way to better appreciate and accept the varied tastes, instincts and preferences that we all demonstrate from time to time otherwise we will stifle creativity and box ourselves into too small a corner.

  • I think that the Flash version (which has a horizontal orientation) of the graphic is much harder to understand than the newspaper version which was vertical. It was only after reading Gert’s article and seeing the vertical version that I finally understood what the graphic was trying to show. It was a lot a work to discover several trivial bits of information that every person who took the trouble to understand the graphic should have already known.

    Here’s what I “learned”:

    Some movies start out strong and then sales peter out.

    Some movies start small and grow over a period of weeks/months before declining

    Summer and Christmas are good times to release movies

    An Oscar nomination can increase ticket sales (or not)

    All this “information” was found by reading the text associated with the graphic. I would have learned more, faster, by the elimination of the graphic.

    Here is the link to the newspaper version


    • Contra Dan, I found the graphic easy to understand. All points that Dan derived from the article, I derived from the graphic without reading the article.

      To me, the streamgraph looks like a very useful tool for presenting certain datasets, and I plan to add it to the toolkit I use for scientific and engineering data visualizations at work.

  • Hannah is right. There must be room to experiment with visualization. What I think the NYT got so right with the film viz is that, regardless of success or failure, they were experimenting with subject matter that was interesting but not in a serious way.

    Would that same presentation have been received as it was had NYT been dealing with health data? I don’t think so. I am sure that the team working on it learned lessons in the creation of that viz that carried over into other visualization.

    I noticed that when they made their visualization about how people spend their waking hours, interactivity had improved. Perhaps this was a result of the interactivity in the film viz.

    Failing is frequently underestimated as a technique for learning, but is I think, so important.

  • in internet lingo, I would call Lee Byron’s graphic an epic win and won’t elaborate of why for concision’s sake.

    I’m not suprised on the discussion given that:
    – infographics are partly artistic, so it’s difficult to determine a “best” piece of infographics. Imagine the louvre’s conservationists walking down the aisles and picking the museum’s greatest work, with all different cultures and epochs represented there. same dilemma here.
    – if we assume infographics can be ranked, then what are the criteria? appeal? public response? technical mastery? mathematical innovation? interest of dataset? traffic generated? rigor? all of these could be valid response, yet they contradict each other.
    – finally, most people interested in infographics are not just passive viewers, but practitioners as well. success of a display can increasingly annoy those who haven’t received such exposure. I’m the first one to confess that I don’t like trendalyzer even if (or maybe because) the original Hans Rosling presentation had over 20m views.

  • I use this graphic in a basic “math in the real world” course that I teach for under-prepared college students. I hand them the URL, and ask them to look for interesting patterns and write a couple of paragraphs about what they see in the data. These are students who do not, generally, engage with numbers or graphs at all. Every semester, most of the class come twith their paragraphs excited by this graphic and the stories they’ve found within it (although a couple of students are usually confused and bewildered.)

    The graph is interesting and inviting in a way that a huge collection of bar charts could never be, and for that alone, I think it is very successful.

  • Jerome/Jessica – really interesting comments

  • The only “infographic” that could be interpreted completely correctly by 100% of readers is just “X is bad and Y is good” (which is obviously useless for most concepts). ANYTHING more complex than that WILL be misinterpreted by some people. That is absolutely not a valid argument against using more complex representations of data.

  • There are a variety of remarks in the comments about how visual design choices are often made in ways that highlight a particular slice of the data in order to advance a specific argument/narrative. It’s an important thing to remember about visualization that (much like data collection), bias is often part of the mix. Sometimes that will work, sometimes it doesn’t, so it’s hard to define best practices on this topic.

    Also, this piece is totally beautiful, but highly derivative in terms of visual display. That is certainly okay, but there should be mention of the innovator(s) of this technique. I’m not sure if it originated here, but there is an earlier work from the Department of Energy’s Pacific Northwest National Laboratory called Theme River. If they originated the concept, they should get some mention. http://infoviz.pnl.gov/research_themeriver.stm

    • byron and wattenberg mention themeriver in their paper. the difference between themeriver and streamgraph is the algorithm behind the organization.

  • It’s like the last thing you want to be said at your funeral: “He was a good guy.” If you don’t have any critics, you aren’t trying hard enough.

    So the boxoffice steamgraph is not going to help us make some new decision about social policy or even explain all the data that the graphic represents and…so what? There is a place for visualizations that make you consider just a little bit more about a small facet of the world that many people take interest in (boxoffice receipts) while looking shockingly pretty. Like say, the the type of casual stimulation you might want on a Sunday morning.