Guide: Radial Fundamentals

So you found a cool chart and followed the blog to make it, but there is more to it than following step-by-step instructions. 

A recent encounter with one of my #DataFam where I noticed something strange in their new viz has led me to write this post.

Radial charts - Coxcomb, Radial Bars, Radial Time Series, Sectioned Radial Jitter, Sunbursts, the list goes on,  all of these charts are visually interesting but they can be like marmite in the data viz world.  Where some folk love them (myself included!), there are others that find the choice to include them in a visualisation an unnecessary complication for the reader to understand, when other chart choices would be a better fit.

For data analysis I rarely use these charts - if at all in a business context, but where I want to push for something a little more visually progressive these charts are my go to.  It is that attraction, of something that looks so different to what is normally produced in our working lives, that I believe draws people to using them in their hobby vizzes; such as the chart below that I created to show the complete history of the Six Nations championship.

These charts do have their place in the viz world, but to use them correctly you need to make sure you set them up in the right way.

I will try to explain more...

There are essentially two types of radial chart from the view point of how the reader is expected to understand them: 1. Change in mark distance from centre, 2. Change in size / mark round the radial.  The first you are expected to read from inside the radial to outside concentrating on the length of the marks (i.e Radial Bar or Coxcombs) and the other you are required to read round the radial, in bands, and look at the marks changing in size or colour (i.e. Radial Time Series, like the Six Nations viz above).

There are a few fundimentals that you must adhere to when setting up certain types of radial - those that rely on interpreting the distance of the marks out from the centre (so, type 1 radials above).

1. Check your axis

This has everything to do with displaying the chart in the correct way so that interpretations are not skewed.  The scale coming from the centre of the radial should remain consistent all the way round the radial.  Therefore ideally the X and Y axis should be of equal lengths, so that the radial is plotting within a square area.  Any devation from this will effectively stretch the radial along one of the axis, which will only add to the complexity of reading a chart that some already find awkward to read.

This is why...

If you take two bars of the same value and plot 1 vertically and the other horizontally, they should appear on screen in equal length from the radial centre.  Yet, if you make one of the axis longer than the other, lets say the y axis, you are infact squashing the bar plotted vertically, so it is no longer equal to the horizontal bar on screen despite both being the same value!

Therefore, you must always fix your axis to have the same length.

You can play with the axis values though to suit your composition, look at the below images and see how changing the axis (but always maintaining the same overall length on each) can move the centre point of the radial.

You may ask youself why would you do this?

If you are using a visualisation tool such as Tableau, you may want to create some space on the radial worksheet to apply an annotation, or you may want to float the radial near some text on your dashboard but without having part of the worksheet falling off the side of the dashboard or covering another chart.  Just understanding this can help you compose your visualisations.

If you look at the bottom middle chart above (where the radial is in the bottom left corner) this set up provides an opportunity that allows you to increase the size of the chart to fill the chart space better - due to the alignment of the vertical and horizontal bars, whilst maintaining that the bars fit within the chart area, unlike the other options above.  To fill the chart space you would only need to reduce the length of the axis (the positive value on each axis), like this...

2. Maintain a 1:1 axis relationship when placing the radial on a dashboard

This is essentially the same principle as above.  You have now gone to all the trouble of sorting out your axis within your chart, so lets not mess it up now!

There are times when you see a published visualisation with the radial charts squashed in between text; they are pretty easy to identify, as the area on the dashboard afforded to them can appear rectangular.  So unless the chart has been floated over the top to fit over the rectangular white space, you can pretty much guess that the chart has been tiled and so the worksheet will be filling the rectangle.  There are other 'tells' which point to a radial being squashed, but they are far more subtle.

You need to be aware of this, because allowing the worksheet to fill a rectangular space is effectively the same as stretching or squashing your axis!

I have fallen foul of this myself, with one of my first visualisations on Tableau Public - 'The World's Most Expensive Prime Property, where you can see below int he image the radial worksheet dimensions are 1144 x 610   - eek!  Rather amusingly, this visualisation actually won "Viz of the Day" (Tableau Public Award).  At the start of this post you can see the same radial but set up with equal axis ranges, look how different the two charts appear!

The last radial chart fundamental I wanted to mention as part if this post is more specific to a couple of chart types...

3. Coxcomb & Sunbursts overlaps

Dependant on the tutorial you follow on how to build these chart types you may or may not run into some trouble with overlapping.

By 'overlapping' I mean, where the polygon areas drawn, for each bar / section can overlap the adjacent bar / section.

Why this is an issue is because overlapping gives a distorted view of the chart and potentially one or more polygons become more dominant over the others.

Take this coxcomb chart below, all appears well, the lighter section at the top is the largest (longest) and which looks to be the main takeaway.  Notice how it dominates the other wedges.  Yet, all is not well here, and the 'tell' here, is that the angles of largest wedge at the point at the centre of the radial is larger than the other wedges - they should be equal.   Admittedly, it is very difficult to pick that up.

The reason for this difference in angle is actually overlapping.  If we take the fill opacity down to say 60%, it becomes apparent the extent of the overlap.

If we correct, the build of the coxcomb and plot the wedges without overlap, the largest wedge becomes much less dominant visually than before.

All three of the points discussed in this post impact the presentation of the radial charts and the data they contain  To not correct these issues, then you are potentially misleading your audience.  Radial charts  can be difficult enough for the general reader, but keeping the above points in mind will hopefully keep the reading difficulty down.

I started this post saying I am partial to a good radial, so I hope to see more of these charts being presented correctly and chose your radial chart type wisely - the chart that will be present your data.

Lastly, and most of all... have fun with it!

I hope you found this post useful.

Thanks for reading,

Sam