A Misguided Visualization Critique

Tableau posted a critique of the BBC's coronavirus visualizations, stirring up a good discussion and mixed reviews on Twitter. I wanted to dive a bit deeper into my reaction to the kerfuffle.

In the post, Andy Cotgreave’s primary assertion is that the following BBC bar charts should be set to a scale of 0% to 100%.

picture2_33.png

He makes the following points:

“Our eyes stretch to the right, and the risk is we perceive that to mean 100%."

Why would that be? To me, the first thing we gather from a chart like this is that some population segments have a much higher “case fatality ratio” than others. When you see the actual value labels, they clearly show the scale going to 15% (this scale could be better positioned at the top of the chart).

“There’s no space for us to visualize the people who survived the disease.”

I don’t see why this would be a necessity. Every visualization makes judgements about what data to include or exclude. The point of the chart is to focus on fatality rates, so it does. His suggestion would imply that all of the Covid-19 dashboards (here’s ours) should start by showing the population of the world.

“We don’t actually know where this data is coming from in the first place.”

There is a data source label on the chart, which is a reasonable place for such information.

“Our responsibility, right now, is to communicate all the data.”

Again, I disagree. The data analyst (or analytics translator) is explicitly not doing their job if they try to show all the data.

He goes on to create a couple less useful versions of this data. The first shows a bar chart with a scale of 100%, which makes it nearly impossible to see the differences in fatality rate by population segment (the original point of the article).

picture4_19.png

“While this approach (full scale) is interesting to show issues like gender gap, I don't think it is useful with rare events where small variations have a huge impact. You need more resolution to read the differences between, say 40-49 and 50-59.” — Jorge Camoes

A second iteration focuses on the survival rate, trying to paint a rosier picture of the situation. Good news, you can hardly see the percentage of people who died!

Which led my brother to do a little research and find that the Covid-19 survival rate for 80+ year olds is comparable to a shark attack but better than a lightning strike. This is why context is so critical — whether it is 15% or 2%, these death rates are high enough to deserve the spotlight in the original BBC visualization.

And to complete our journey of charting critiques: it might be better if Tableau had picked on a chart that really needed some help, like this one from the University of Florida about the fatality rates of deadly snakes…