* not “ever”, just think about it though where precision is required!

This week I was “volunteered” to cover for Andy Kriebel on #MMVizReview. This was a great opportunity for me, and meant I got to chat with newly-crowned Tableau Zen Master Eva Murray for an hour or so while we constructively appraised a couple of dozen #MakeoverMonday submissions. It was great fun and I’m grateful for the experience I gained from it.

What stood out for me was the number of submissions which leveraged sized circles in their analysis. It was quite understandable as medals tend to be round, so it was a relatively obvious visual metaphor to use. The problem is – circles suck for a number of use cases. Us humans really struggle to interpret them in comparison to something like a humble bar chart, as Andy Cotgreave expertly elaborates on here. It’s important to make sure that when you use them, the context and data supports this visualisation choice.

I’m stubborn and I wouldn’t let #MakeoverMonday this week pass without making a dig at this visualisation faux-pas, so I banged out a quick example of why this means of articulating data is usually the wrong one.

Screen Shot 2018-02-18 at 18.18.34.png

So…..what was the value of German exports? Swiss? Belgian? What was the difference between the export values across these countries? It’s absolute carnage and I switched off tooltips as well, to be even more obtuse. If you’re interested, then you can download that workbook here.

The alternative option is available here. That second iteration looks like this:

Screen Shot 2018-02-18 at 18.07.21

Now I’ve obviously layered on additional cues here to aid interpretation, so it’s far from a fair comparison, but if I’d added the Export values into the circles, would you really have been able to understand from the size of the circles the magnitude of difference? No – you wouldn’t.

The second viz is very simple, but still utilises a couple of useful tricks, so I thought they’d merit a word or two. Firstly, I created a Set to allow users to choose the number of Exporting nations they wanted in the view. This is because originally I was just going to create a Pareto chart, but I soon realised that this would lose sight of the nations who were contributing to the majority of exports and that seemed a really crap idea.

Screen Shot 2018-02-18 at 18.35.14

By Exporter, I’m pulling through the Top N nations based on the USD value of those Exports. The [How many expor…] “thing”, is a Parameter:

Screen Shot 2018-02-18 at 18.35.25

Again, it’s simple. Let the users select whatever value they want and then we control the number of nations to explicitly see in the viz, and the remainder will be grouped together.

Once I’d ditched the Pareto idea, I realised I could simply re-purpose this Set. It started with an extra calc:

Screen Shot 2018-02-18 at 18.38.51

Essentially: If the exporting nation is in the Top N Set, then call it by its name, else lump it together as one entity called ‘The Rest’. In the grand scheme of a workbook, it looks like this:

Screen Shot 2018-02-18 at 18.38.59

Basic but required in this context. It makes the viz and the colouring dynamic, and anything that automates a process is a good thing.

This idea of easing processes extends to the second minor tip of this post – the sort order. Note how the exporting nations are sorted in descending order by the value of exports? However, also notice how “The Rest” sit at the bottom, in spite of having a greater value that the topmost nation? That requires a calculation too:

Screen Shot 2018-02-18 at 18.42.23

What does that look like in a table?

Screen Shot 2018-02-18 at 18.43.18

By inverting the export value for “The Rest”, it inevitably becomes the lowest value in the view, hence it assumes a position at the foot of the view when sorted based on this calculated view.

As an addendum to this post, Lilach Manheim rightly posted that the tone of this article was pretty absolutist and implied that sizing of circles should NEVER be used.

Clearly that is not the case, but I’m comfortable standing by the assertion the sized circles in isolation should not be used where precision is required, and that they should be complemented with a more traditional means of interpreting the underlying data (a discrete value or label, as an example) in these instances.

As a means of creating impact where precision is not required, they certainly have a place.