Minard, Tufte, Kosslyn and Godin (and Napoleon)

Do you prefer the full report:


Or the executive summary?


For Tufte’s fans, Minard’s map plays a central role in Tufte’s iconography, and the way he praises it (“best statistical graphic ever”) is quoted endlessly (974 results in Google as of today, to be precise). Tufte discussed The Map in his first book (The Visual Display of Quantitative Information) and in Beautiful Evidence he uses it to illustrate six Fundamental Principles of Analytical Design:

  1. Show comparisons, contrasts, differences
  2. Casuality, Mechanism, Structure, Explanation
  3. Multivariate Analysis
  4. Integration of Evidence
  5. Documentation
  6. Content Counts Most of All

But, of course, they violate Seth Godin’s Principle of Make a Point in Two Seconds for Lazy People (here, about minute 2:40):

…this is one of the worse graphs ever made. He [fusion_builder_container hundred_percent=”yes” overflow=”visible”][fusion_builder_row][fusion_builder_column type=”1_1″ background_position=”left top” background_color=”” border_size=”” border_color=”” border_style=”solid” spacing=”yes” background_image=”” background_repeat=”no-repeat” padding=”” margin_top=”0px” margin_bottom=”0px” class=”” id=”” animation_type=”” animation_speed=”0.3″ animation_direction=”left” hide_on_mobile=”no” center_content=”no” min_height=”none”][Tufte]’s very happy because it shows five different pieces of information on three axis and if you study it for fifteen minutes it really is worth a thousand words. I don’t think that’s what graphs are for. I think they try to make a point in two seconds for people who are too lazy to read the forty words underneath. And to make me spend fifteen minutes studying it doesn’t make sense.(…) The kind of person he wants to reach they want to read a complicate, difficult to understand graph and get the satisfaction of figuring it out, because then they get it…

In this post he uses pie charts to further illustrate his principle:

Pie charts are a great example of how people go wrong. [First pie chart, eight slices] It’s accurate. It shows more than a half a dozen places that traffic come from. It’s also useless. It’s ungrokable. It doesn’t have a point. [Second pie chart, three slices, one of them exploded] Here’s the same data, grouped to make a point. “We get our traffic from three sources, one dominates the other two, but only one of them is under our control in terms of our ability to scale it directly. So let’s talk about how we grow that slice.”

Godin is not alone. Minard’s map also violates Kosslyn‘s Principle of Capacity Limitations and Principle of Compatibility:

The hazards of not respecting the limitations of human mental processes are nicely illustrated – ironically – by one of Tufte’s favorite graphics. (…) This display has never captivated me for the simple reason that given human processing limitations – I needed several minutes to figure it out. (…) I can’t agree that this is an effective way to communicate; the display doesn’t present the facts so that they’re clear or easily absorbed. If you are in the mood, you may enjoy taking the time to study the display for the fun of solving a puzzle, pondering intricate details, or appreciating the graphic devices employed. But if you want the facts and want them in a clear, easily understood way, this display is not the solution.

So, in order to design a better graph, Kosslyn proposes a new design that could take advantage of the Principle of Salience, Principle of Relevance, Principle of Perceptual Organization and Principle of Discriminability e makes a better use of the Principle of Compatibility. The end result doesn’t seem much better than some revisions of Minard in Michael Friendly’s Gallery.

(By the way, I’m also posting a series on design principles for better charts…)

I find all this slightly absurd.

The divide here is, of course, the level of detail you need or are prepared to accept. A good/bad chart is not defined by the time it takes to read it, it is defined by the insights you get from it and how efficiently it tells the story. Minard’s map is much better at this than the typical dashboard or presentation slide we are used to, a War and Peace in a single image against a boy-meets-girl-boy-loses-girl-boy-gets-girl plot in fifty slides.

Tufte says “to clarify, add detail”, but detail is a dangerous thing. It is the detail you choose that defines how you interpret reality, and it exposes your biases. Minard adds detail to explicitly link deaths to cold temperatures during retreat, but around 75% of men died on the way to Moscow and he doesn’t tell us why. You can seriously undermine your analysis by adding detail to the wrong places.

No details for Godin, please. He contradicts himself, but I suspect he couldn’t care less. He wants less data and a chart that makes a point in two seconds. But then he needs “to talk about how we grow that slice” and that means more data. Wouldn’t it be nice to have in the same chart something that helps him? Of course, that would require more than two seconds to read the entire chart…

At the end of the day, what really matters is how you manage your data. If your skills are poor, your charts will always reflect that, even if you are a Tufte fan and no matter how much detail you add…

So, do you believe that this is “the best statistical chart ever” or that’s something that we shouldn’t take too seriously?[/fusion_builder_column][/fusion_builder_row][/fusion_builder_container]

9 thoughts on “Minard, Tufte, Kosslyn and Godin (and Napoleon)”

  1. I wouldn’t call it the best statistical graphic ever. I feel like Minard’s is labeled best much the same reason that Citizen Kane is considered number one. Albeit, I’m not sure what graphic I would call the best.

  2. Minard’s chart requires an effort to be readen, but this effort is much less than the effort needed to read all the information in a table or in a text. The redux-pie version is worse, because reading the chart requires more effort than reading the sentence “2% survived”.
    Aside from that, Minard’s chart wasn’t made to nowadays newspapers. So it’s incorrect to think on it as it was published in the New York Times.

  3. Arguing over “the best statistical graphic ever” is kind of useless, and only takes away from a much more interesting point: the inherent story-telling of graphics, even static ones.

    If a visualization or visual representation is only a pipe that pumps data into the brain, then the pie chart perhaps makes more sense. Or better, just use numbers. The “executive summary” can be put much more concisely in a simple sentence or a table of three numbers (total, survivors, deaths) than a pie chart.

    But to engage an audience, to find out more about what happened, and to actually care about what the data depicts (the death of 390,000 people!), we need a bit more than a pie chart. That is the power of graphics, that they tell you a story that you will remember. One of the reasons Tufte is so popular is because he tells stories, and these stories stick. A pie chart that looks like any other pie chart will never stick, but a well-designed data-specific graphic like Minard’s will.

    I have incidentally posted somewhat similar thoughts (on the subjectivity of visualization) on my own website recently.

  4. In the same way a writer needs to know the audience he is writing for, a visualization designer has to keep the audience he is presenting to in mind. If the audience is interested in quick facts then a minimalist presentation is better, but if the audience wants to know the detail, then the detail should be presented. This doesn’t make one visualisation better than the other, because the visualizations are saying different things. The inclination to consider Minards visualisation as superior to the pie chart might also be due to the fact that, we tend to give more credit to designs that condense a large amount of detail into something elegant and readable, as this is more difficult to achieve than the minimalist approach.

    I agree with Robert in that it’s pointless to try to find the best visualization in a domain that is so subjective. Being the best visualization depends on too many variables and eliminating all these variables is both impractical, and not necessarily desired. The subjectivity of visualizations especially in an uncontrolled environment is one of the reasons why scientific evaluation of visualizations is difficult. On the other hand, this same subjectivity is what makes visualizations so powerful, because the same display can mean different things to different people (even if this is not intended).

  5. I think Minard’s graph is really magnificent. Think a parallel approach: Can you create a statistical graph about the World War II? Or by using the Black Book of Communism? What are the measures then? The number of victims of course, what else?

    The issue makes the chart (series of charts) worth studying. As we all know, global warming is the issue nowadays, so it’s not a surprise the charts made available by Al Gore were rewarded with the Nobel Prize.

  6. Godin is thinking in PowerPoint terms: He is imagining himself in front of a screen giving a pitch.

    Long thoughts are not welcome at that party.

    (2 seconds, indeed).

    Minard’s chart (and much of Tufte) is about *visual reasoning*, not about delivering key bullet-pointed figures.

  7. Perhaps the two second visual bullet for people who are struggling with ADD is about the worst way to present information to anyone who actually needs to make a real decision. I find the kind of chart produced by Seth Godin’s approach mostly useless. I would take Tufte’s approach any day. Is the Monard graph the best? Best for what? I can learn from it though, which is enough for me.

  8. Context is everything. Who was the audience? Why were they assimilating the data? What are you expected to “carry away” from it? Is it a substitute for a verbal description, or a supplement to it? What were his audience’s expectations?

    I think it’s magnificent, as did most of the network engineers who stopped to peer at it pinned up on the outside of my tech writer cube. But this is hardly relevant to the task- definition questions I listed above.

  9. The best, who knows, but it is mighty fine. At a glance (in two seconds) you see the decimation of Napoleon’s troops over time and distance and then a direct comparison of the ‘before’ and ‘after’ as the retreat passes the attack. For the more detail oriented, there are also numbers, locations, dates and elevations.

    A Powerpoint pie chart or bar graph could only show elements of this data but would never capture the ‘futility’ this chart conveys.

Comments are closed.