Greenhouse Gas Emissions - BBC Redo 2
Treemaps are an excellent visualisation given the right situation. Created by Ben Shneiderman and Brian Johnson at the University of Maryland in 1990, they were originally designed to show the size and hierarchy of files and folders on a shared hard drive. Dissatisfied with how limited lists of data were in how much information could be displayed at one time, they set out to create a new way visualising large sets of hierarchical data in an intuitive and efficient way. Here is the classic stock market example created by Stephen Few:
This lets us see in a very space-efficient manner, the relative sizes of each category, the largest individual stocks within each sector, and the conditional formatting means we can even see how those individual stocks or whole sectors are performing.
However, is far more common to see Treemaps in the wild used in the same way as this recent BBC article on greenhouse gas emissions:
This is the wrong visual choice and has therefore has a few issues:
Accurate comparison of categories is hard. Humans generally aren’t good at accurately estimating the difference in area between two shapes.
Categories are not defined. Our understanding of the visual is hampered by not really knowing what falls under “Business” or “Transport”.
Ugly. The colour choices are just awful and would distract readers even if the visual choice was better.
So how can we make this better? Well as always we should start with what questions we want to use the data to answer. In this case it’s fairly simple:
Which sectors are cause of most greenhouse gas emissions?
What is responsible for emissions within each of these sectors?
It was at this point that I decided to go to the original source and see how the data was displayed there. To my pleasant surprise, a perfectly serviceable column chart had been used:
This is the correct visual to use, as bar and column charts allow easy and precise comparison of multiple data values. So why change this to treemap?
The only reason I can think of is that a column chart wasn’t deemed “visually interesting” enough, but this is the completely wrong approach to take when presenting data visually as we saw with the Women’s World Cup Performance visual.
“Visually interesting” could just as well be a synonym for “unintuitive”. As you add more uncommon or varying elements to a page, the more work readers have to do to comprehend it. The reason column charts and line charts are so ubiquitous is because they are so efficient at displaying data in an intuitive way.
If readers are care about the data and its story, then they won’t get bored. Show the story clearly, and let the data speak for itself.
There’s only a couple of things I’d change about the original chart:
The x-axis labels are a bit cramped in this column chart format, so we should switch it to a bar chart instead
As there’s a small number of bars, I’d label each bar directly to show its value instead of make users look between the y-axis and the bar
If we want to make this more pleasant to look at, then we can subtly use some colour and text formatting for emphasis
Here is my version created in Power BI.
This is a combination of bar chart for the main visual and cards for the labels. I’ve used separate cards as data labels as it gives more options for text formatting, for example here I’ve made the data value bold without changing the category text. If a dynamic version is required for because the data changes or can be filtered, then regular labels will need to be used instead which unfortunately don’t have this option (yet).
So far this only answers the first question, as for the second, I’ve added a report page tooltip that provides more information on what specifically each category refers to:
I prefer putting supplementary information as close as possible to where people will actually want it, instead of hiding it away in an “Info” page at the back of a report.
Let me know your thoughts or how you would improve the treemap!