Below is a great example of how to misguide the viewer’s interpretation of data in a Sankey diagram. Found this one on presentation slides somewhere on the web.

The two arrows branching off to the top in a 90° angle do not maintain their magnitudes, which supposedly represent the quantities, and are drawn at a deliberate width. On top of that, the bases of the arrowheads are about two times as wide as the actual arrow width, thus overemphasizing the flow. Look at the 40% thermal losses which look much larger than the 50% useful work to the right side…

I did play around a little bit with this tiny example, and came up with a number of alternative versions.







Not sure which one is the “best” one, and each has its pros and cons. #1 (hover the mouse pointer over the image to see the number of each alternative version) is very close to the original version. The arrow head size in #3 is more modest. #4 has no explicit spike arrow heads at all. #6 has grey divider lines on most of the horizontal section. I kind of like #7 with color differentiation best, but then again, it is energy that is displayed in all flows.

What do you think? Let me know your favourite or suggestions for improvements in your comment

8 Responses to “Lying with Sankey diagrams (4)”

  1. hamish Says:

    the spiked arrowheads add noise, because they add another dimension to the diagram–now the viewer has to decide–do they judge value by the width of the flow, or the size of the arrowhead.. even if both are in proportion, the data is being duplicated…

    So the 4th diagram is the only one to get my vote.

    good work though!

  2. michele Says:

    i agree with hamish – the arrowhead is useless. Not only useless but even dangerous if the arrowhead isn’t proportioned to the flows (as in original & first image) or only on some flows (as in 2nd image).
    I think the best is the 4th, but using curved angles as in 5th image and colors as in #7.
    It is true that all flows are energy, but there aren’t other units of measurement, so there isn’t any conflict using colors.

  3. Gabor Says:

    I also agree that version 4 is the best. I’d even make a version 4 with rounded flows, to give it a more organic feel of a flow. However I don’t judge the use of arrowheads to be so bad as Hamish and Michele. Especially for small flows the use of arrowheads is useful to indicacte flow direction. But in the original and your version 1 the heads are definitely too large to be useful.

    I consider the use of color, as in version 7 Ok to emphasize contributions (like in bar charts or pie charts). I like it better than using outlined flows, as in v 1 + 2. Keeping colored contributions and tracing them all over a large network (obviously not here) can be detrimental.

    I think the main problem with the original chart is that someone just used the built-in drawing shapes of Powerpoint (or Excel) and stretched them out until something with the approximately correct flow widths resulted (I could reproduce this exactly). The arrowheads could be shrinked, which was not done. It took the person probably just one minute to draw this. It’s a lazy effort, a lousy diagram, but I don’t think that the intention was to deliberately lie.

  4. phineas Says:

    Thanks for your comments. Interesting to see that #4 seems to be a clear favourite.

    @Gabor: Of course I didn’t want to offend the author of the original diagram, or suggest he is deliberately lying. It’s just a catchy post title, and the purpose of this little series is to show how to make better Sankey Diagrams.

  5. Prince Myschkin Says:

    I’d prefer a version with no arrowheads, a thin black frame and rounded flows.

    From a philosophical point of view, presenting data is never done intentionless. Lying is a hard word if we don’t know the intention of the presenter. But I agree that, for a scientist, the diagram is phony.

  6. Will Stahl-Timmins Says:

    I’m not too worried about the (smaller) arrowheads, although I tend to produce Sankey diagrams without. I quite like the colour separation in the last one too, it helps when making the size comparison, but I take your point about it all being the same pool of energy that is being presented… Perhaps at the left hand side of the diagram, the whole arrow could be the same colour, but with a nice gradient leading to the separate outputs in different colours?

    There are probably as many different ways of making a Sankey as there are people in this world to draw it…

  7. Sankey Diagrams » Blog Archive » ‘Worst Sankey Diagram’ Award Suggested Says:

    […] my last post, which led to quite a number of comments, one reader of this blog has suggested to create a […]

  8. Matt Says:

    #5 without arrowheads = #4 with rounded flow = the best

    Odd you didn’t make one of these…?