Monday, July 13, 2015

The stickiness of visualization conventions

Visualization conventions are as sticky as any other kind of convention. But stickiness isn't a proxy for goodness. We shouldn't use a graphic form uncritically just because it has become the default to represent a specific kind of data. Tradition alone isn't a good textbook.

At the bottom of this post there's a before-and-after exercise*. The original, on the left, comes from Spain's Just for the record, I have mindlessly done tons of parliament charts like that in the past. The “mindlessly” part of that sentence is critical, as I was mimicking what many news infographics designers were doing.

But one day I actually tried to read the parliament-looking chart, and make comparisons. I couldn't. I needed to read every single figure to be able to mentally visualize variations.

You can see the evidence yourself. Read just one or two figures in the first chart below, force yourself not to read the rest, and then try to estimate changes accurately. Good luck with that.

So the parliament chart is nearly useless; a table would've been better. This chart is more a nice-looking abstract illustration than a visualization because a visualization should be an aid to understanding, not a hurdle.

*NOTE: I made a slope chart, but I'd be fine with a bar chart, a dot plot or even a divided bar chart (suggestion inspired by Clement Levallois)


  1. Well, there is one major benefit of the arch-visualization: when it comes to coalitions, it is immediatelly visible if two or three parties together do reach a majority in the parliament...

    1. That could also be achieved with a rectangular graph. In fact I remember seeing those in the Guardian coverage of the UK general election this year.

      This same chart, with the number of representatives would be equally good.

  2. The divided bar chart is a better choice for that, Štefan, and only if parties are grouped by possible coalitions. See example above, suggested by Anonymous

  3. There are at least 2 aspects I am interested in: a-the "flow" of representatives (alt. percentage) for each party along time, and b-as Stefan says, the ability to form a coalition based on the political wing or other convenience majority.

    For (a) I'd use a "flow" chart -Stankey diagram or bump diagram (draft done with RAW)
    For (b) I find the horizontal layout of divided bar charts better

    With (b) is also clear that, if polls are true, no single left or right coalition will reach absolute majority. Even with the incumbents Podemos and Ciudadanos in play, PP and PSOE will have to rely on "Others" - as they traditionally did- to reach majority.