KPMG conducts an annual survey of automotive executives from around the world, and this year, it visualized the results using Tableau. The report design is definitely slick, but I had problems with the way the author(s) chose to display some of the data.
I found the graph for the results of the business model disruption question especially puzzling:
First, this is clearly Likert-type data, so why is the “Neutral” category placed at the end, instead of in the middle, between the “likely” categories and the “unlikely” categories?
Also, while the color choices of lighter and darker blue for 2015 & 2016 are fine, setting these colors against a white background creates the problem of making the labels of the small values hard to see (see the 3% values for “Extremely likely” & “Not very likely”).
Finally, and most important, a series of back-to-back bar charts (used often for population pyramids) is an odd choice to display data from different years. I knew I was looking at survey data with a Likert-type scale, so when I saw the back to back x-axis, I expected to see diverging stacked bar charts. When I didn’t, I was confused, and it took me a while to figure it out.
I decided to try my hand at revising this visualization to one more appropriate for the data. First, I tried to see what the diverging stacked 100% bar charts would look like (h/t Stephanie Evergreen):
There are a couple of problems with this alternative:
1. The fact that 2016 has a 0 value for the “Neutral” category throws off the balance, and almost the entire reason for choosing this graph, which is to see the like categories grouped together for better comparison.
2. Also, ideally, the category names would be written above the bars themselves, making the key unnecessary; however, because of the imbalance of the bars, and because, as mentioned above, a couple of the extreme categories have very small values, writing the category names right above the bars was not an option.
Next, I tried a standard stacked 100% bar chart:
This is slightly better than the first revision because, at least the like groups can be compared from the same starting point (the ends, in this case). It still doesn’t seem ideal for the data.
So, I decided to try a slope chart:
The appeal of this chart is that it emphasizes the most important feature of this data:the change in each category from 2015 to 2016. The proportional totals are lost with this graph, but I don’t know how important they are.
What do you think? Which of these charts do you think best displays the KPMG survey data?