Lessons learnt visualising CHAID
One thing I've been looking at recently is data visualisation: how it's used in market research and how it can be improved. At one end of the scale you have a 'high design' approach which aims to produce the most beautiful, clear, data visualisation possible. At the other end of the scale you have a 'pragmatic design' approach - how can you teach your average person to visualise data more effectively.
As part of the pragmatic design effort I've started to try and come up with some general techniques/approaches for visualising common research questions and methodologies. As a starting point I thought I would look at how we can visualise CHAID, there were a couple of reasons for this:
- It's a relatively well known and widely used technique
- The output you tend to get from CHAID is pretty visual to start off with, which I thought would be a good starting point.
So I started fooling around sketching out bits and came up with the idea of hierarchical drooping bluebells (strange I know). Based on this rough concept I started (but didn't finish) the graphic above and I learnt a few things:
- 'Hierarchical drooping bluebells' - WTF? Don't know where the idea came from, but I wished it had stayed there. The concept sucked, which is why I didn't bother finishing the graphic above.
- It's difficult designing something out of context without a subject matter. I tried creating the fake subject matter of visualising magazine subscription drivers, but ultimately I knew that I was striving for a more general approach so it didn't help very much.
- We normally try to put to much on our graphs/visualisations and I found myself trying to find a space for everything a researcher may want to add to a graphic like this. This type of thing definitely works better when you have clear point (not points) you want to make.
- I thought it would be easy to look at something like CHAID because it's output is visual to start off with. I wasn't easy, I couldn't stop thinking about hierarchical designs, having a visual basis to work from seemed to cloud my thinking a bit.
So there you have it, one naff graphic and a few things to be thinking about.





