A case for more sophisticated visuals

As head of data journalism at Deutsche Welle, Gianna-Carina Gruen often decides which graphics to include in articles – and has to defend her choices in the newsroom. Recently, a rather uncommon chart caught our eye. We asked her how it got there, and why.

Within our data-driven journalism team at DW, we try to experiment with chart forms that aren't part of DW's standard repertoire yet. In some cases, you might have to find a trade-off between a fancy look and effective visual communication:

We usually lean way more to the effective side, but also give visually different options a try. One example of that is a small multiple radar chart we did for our piece comparing the real cost of travel of trains and planes. The data analysis was done by one of our freelance data journalists, Tom Wills, and we developed the ideas for the visuals together.

When we plan out our stories, we think about whether it rather consists of a series of charts or whether there's one chart holding the entire story. Trains vs planes has both: Several simpler visuals explain, step by step, the different aspects that together compose the cost of a flight or train ride.

The small multiple radar chart can be seen as a visual summary of all the other charts - and with that summary function comes in this particular case a very high information density.

As a data team, we do have the editorial liberty as well as backing from DW's design department to publish these more complex visuals, but we still consult with the respective editorial desks before publication. "It's definitely a heavy one, but let users work it out", the science & environment desk editor on duty wrote as feedback on the radar chart.

When pitching the story and visuals to our social media desk, the immediate reaction to the radar chart was "this is way too complex." The question of complexity or when something is judged as "too complex" is one dear to my heart (as in: I can see myself writing a PhD on that), because I believe it heavily depends on your own context: How data literate you are, how experienced you are in reading charts, what types of charts you are regularly exposed to, how familiar you are with the topic or in which context you are seeing the chart.

Is the focus on "understandable" or "interesting"?

I can see where our social colleagues are coming from: On platforms like Facebook, you might want something that is easily and quickly understandable. And without a doubt, bar charts are often a more effective choice than a small multiple radar chart. However, they are also very abundant: Visually, you won't catch people's eyes with yet another bar chart.

It is a question of platform strategy: In the article, several graphics build up to "the big one", which probably makes the latter easier to understand. If the radar chart is your first point of contact e.g. on social media, it can surely be overwhelming. But then again, you should ask yourself what your goal is: Do you want your Facebook fans and Twitter followers to look at it and understand it fully right away? Or do you want to catch their eye, their attention, and raise interest to actually click through to the article?

In the end we don't know how our entire audience would answer all these questions, so we always try to push for doing it and see how people react. Only then can we figure out, one visual at a time, what works well on which of our platforms.

In terms of self-critique, we would do something differently next time: We wouldprovide a little legend with visual cues on how to read a radar chart on the graphic itself, rather than explaining it in the text.

PS: Turns out, users did like reading it: The trains vs planes article has seen one of the highest reading times of all our data-driven pieces.

You can find the article here: p.dw.com/p/33h4K
You can find data and code behind this piece here: github.com/dw-data/travel-cost


Gianna-Carina Grün

Gianna-Carina Gruen is a data journalist and currently leads the data-driven unit at DW. You can see the team’s work at dw.com/data and her personal profile on gcgruen.github.io

Runs on:

How many stickers do you have on your laptop?
Zero. Am I still eligible to be a data journalist?

How many pie charts have you built?
In all life time, probably less then 10 (so I hope).

How many times per week do you have to explain what "data journalism" is?
Let’s say: Not as often as I had to about a year ago.

How often do people use you as IT-Support?
Never, but I’m often being understood as the research desk for requests like “I am just looking for this one number of migrants/electric cars/AfD voters in Germany”. Luckily, DW has a desk for that with great researchers who seem to be able to dig up virtually any number.

How big was your biggest data set?
Probably half a million rows, which I only know because I once took a screenshot for reference how much information I have when I talk about “data” versus what people want from me when they call me “looking for this one number” (see IT desk question).

snow flake
© 2018 Journocode