What you asked

Let’s start with a question that’s all about getting started...with data viz that is. You asked:

Could you describe your workflow when creating a data viz -- does the question or data come first? Or neither?

"It depends. Sometimes a good visualisation idea appears when you’re exploring data sets at leisure, just to see what you can discover, particularly with the help of experts who know much more about the data and the topic than you do. In other cases, you may begin with a preconceived story you may want to tell, then you try to test it with the data to see if it has any merit; if it does, then you tell it."

With that in mind, what is the most underrated or undervalued skill for data visualisation practitioners?

"The capacity to think as data professionals, storytellers, and visual designers, all at the same time. It’s a difficult balance to achieve, and very few visualisation practitioners are good at everything (I am certainly not!), but when someone is able to balance those up, the result is usually a top visual communicator."

What do you think is the most misleading visualisation type and why?

"All of them can be. One of the reasons why a visual may lie is that we tend to project onto them our existing beliefs and biases. If I already believe that, say, a particular policy is good for the job market, and I see a chart that shows that the job market improved after the policy took place, my brain will immediately infer a causal connection between the two phenomena, even if they may be completely unrelated, and their proximity in time may be just mere coincidence."

So then how do you feel about pie charts?

"Pie charts receive a lot of hatred for no good reason. A single pie chart with three or four segments is harmless. The problem is when we misuse pie charts, and we end up designing them with twelve or even twenty segments! As any other graphic form, the pie chart can be used —or misused. The key is to always think about the purpose of the visualisation you’re about to design, and then assess whether the graphic form you’ve chosen meets that purpose."

Since many journalists have no background in data science, how should the field address this challenge?

"I think that you don’t need to be a professional data scientist and statistician, you just need to educate yourself a bit in numerical reasoning and then try to make friends with a few true statisticians so they can help you make sure you don’t screw up!

Nowadays there are plenty of excellent popular science books that can help anyone educate themselves in numbers. I’d begin with Charles Wheelan’s Naked Statistics and Ben Goldacre’s Bad Science. Then, try to read a few intro to stats textbooks, and do the exercises. Some coding doesn’t hurt either, some R or Python."

Our next conversation

This December, the European Journalism Centre has something exciting in the works. We’re releasing an online beta version of the Data Journalism Handbook 2!

It’s going to be a preview launch, with only a first batch of chapters available, and we’re giving you the opportunity to submit questions for our editors, Jonathan Gray and Liliana Bounegru.

Until next time,

Madolyn from the EJC Data team