Summer 2021

Conversation with John Burn-Murdoch

As Covid-19 spread across the globe, it became clear that data visualisation would become ‘the language of the pandemic’.

John L. Walters: I imagine that when you joined the Financial Times (FT) in 2013, epidemiology was not kind of high on the list of things you thought you would cover.
John Burn-Murdoch: No, although one of the first sizeable things I did for the FT was in 2014 when Ebola was starting to kick off.

At the start of last year, I started to see similar contacts from Liberia coming back into the address book.

We have a clear style guide for all our charts, which is constantly refined. We build all of our charts using the JavaScript library D3. In most cases it’s heavily templated, so we say, ‘This is going to be a line chart or a column chart’ and we pipe our data into that. There’s a stylistic approach that we take in general, which is we try to ensure that our charts contain no more information than they need to. We’re opposed to “chartjunk”.

On the flip side, we do think hard about annotations and the use of text in our charts and we will go back and forth on titles and subheadings and exactly what we want to say. There’s a lot of research that shows that when you’re using charts to communicate, the text is critical.If you have a title which says, ‘Number of Covid cases by day’ people will understand the chart but they won’t be prompted to do anything. Whereas if you give it a title that makes a point, you’re going to get a stronger reaction and it is more likely that someone will share it.

It’s like any news story or feature – you need a strong headline to sell it, to grab people’s attention.
There was a delayed realisation that we have all these brilliant headline writers at the FT. As things become more digital, there are challenges to think about what a headline means online versus in print. But print headlines and chart titles serve the same purpose in trying to be concise and clear and maybe even funny. What’s been good is bringing some of the subeditors and headline specialists into our team to work with us on chart titles.

How big a team do you work with?
It’s about twenty of us, and developers, designers and the statistics team are about half. The other ten are split roughly in thirds: those who work on quick turnaround graphics for the print newspaper and online; and others who work on more expansive online graphics or more customised graphics, such as big detailed maps. The final group is the one I’m part of where we’re Jacks-of-all-trades.

We do some live front-end coding and design; we are comfortable sourcing data from anywhere and making charts; and we also do a lot of the writing. So I will make lots of charts and then there will be other cases where I’m writing a story and someone else’s charts are going into that. We see our role as finding things out and communicating them – it’s just that the means through which we do that uses data and charts, whereas our colleagues in the rest of the newsroom use words.

Were there major decisions you had to make about the pandemic? Or was it more of a day-to-day thing of trial and error, finding out what worked?
Back in January 2020, it was a conventional world news story for us. The world news editor and other people on that team were saying, ‘OK, there’s this epidemic in China, spreading into other parts of Asia. We should probably have some kind of page tracking this.’ So we started with a world map showing the numbers of cases in each country back when you could count them on a few fingers. We didn’t think much beyond that, just a table and a map on a page that changed over time.

Two things changed. In the first week of March it became apparent that the pandemic was an enormous data story. Our team was getting more and more requests for data and charts: ‘How is the stock market reacting?’ and so on, and we were working longer and longer into each evening.

As March wore on it was hitting Europe and the US, and everyone – FT readers or not – wanted to know whether their country was going to look like Italy. What set the tone for the year was this one chart in early March based on a conversation I’d had with a news reporter. She was working on a story and asking about data on cases. So I made this one chart, essentially showing that most Western countries were heading the same way as Italy. We thought this was a chart that was going to live on our page here and update from time to time, tracking things, but it just got such a huge reaction online, on social media, that it was a moment when we thought, ‘Oh wow, data visualisation is essentially going to be the language of the pandemic!’

Are you talking about the so-called ‘hairy chart’?
Yes, lines going up for each country. At the time that chart wasn’t interactive as it is now. It was just a static image each day with a lot of annotations.

Was this chart informed by the approach you already took with financial data?
Elements of it – the fact that we were comfortable using logarithmic scales in charts and thinking about exponential growth. There are techniques that we used for the first time on financial data and reused here, but there had to be a lot of original thinking. The data we use in charts rarely comes down in a nice clean format. There’s a lot of transformation that we have to do. The techniques we use to do the preparation and transformation are things we have learnt from financial data. So there has been a lot of thinking about the best way to show a clear picture with messy data.

Were there certain approaches you dropped?
We talk about these pages as ‘tracker pages’, so we have our coronavirus tracker and a vaccine tracker. The audience has grown familiar with certain charts, but when the story changes it doesn’t make sense to keep them all – we have had to constantly assess this stuff. At the beginning there was a clear geographical pattern clustered in Asia and then Europe, but by now it is pretty much everywhere. So we had the main version of the hairy chart which had multiple countries on it, and another version which was a static substitute for the interactive one where we had small multiple charts for individual countries. Those did a great job for a while, but it got to the point where we thought the small multiples took up too much space, so we got rid of them. Each time we did this we’d get emails saying, ‘Where have these charts gone?’ It’s nice that people have developed an attachment, but we can’t have these pages getting more and more bloated.

'No different to a bad flu season?' Financial Times' animation demonstrates how England's winter Covid surge compares to flu seasons.

The animated hospital bed occupancy visualisation was an ingenious, dramatic response to the misleading and fake statistics of ‘Covid deniers’ and NHS denigrators. Was that analogous to something you’d done before?
Yes. Animation is something we try to be sparing with just because it takes more time to do it well. You can’t template an effective animation that will just work when you chuck a dataset into it. In this case it was all about the gradual reveal and the surprise at the end.

When the point is to draw a few lines to show someone, ‘This is normal. This is what you expect’, and then bring in the thing that is different, you can’t do that in static form. You could do it in words, write a long page with a shock at the end, but in chart form the reader will see everything at once so you have to animate it. I was thinking about the most effective, the strongest way of making the point that 2020-21 is completely unlike everything that’s come before. Thinking about social media, animated graphics resonate more with people. When you’ve got a very crowded feed, something that’s moving can catch your eye. It was an important message. Having it as a loop, which will just play again, means that if someone missed something at the start they can come back to it.

Where does something like that sit within the paper?
If we do something which has been planned for a week or more ahead of time, then we’ll almost always have a joined-up approach, so we know exactly how it will appear in print, online, on social media. With the pandemic we’ve done some pieces like that, but a lot are more quick-fire. Animations are a good example. We do them where they need to be done first and then work out where else we can use them. This one [hospital bed occupancy] went out on social media and the news that day, but I got straight into a discussion with the video team and we’re working on an expanded version with a voiceover that will go out on the website and YouTube.

Do you have a daily conference to discuss this?
There’s a newsroom-wide video conference every morning which at least one person from our team attends. This morning [in early January] I was fielding questions about what we should be doing to track vaccine rollouts. So we’ll come out of that meeting knowing what everyone in the newsroom is working on and therefore things that might come our way, but also ideas or topics where our team could proactively pitch in suggestions. Then we have a fifteen-minute meeting just within our team to talk about what we’re going to be focusing on that day. And it’s all on Slack from then on.

How are you dealing with the vaccination data?
There are two questions the reader has: who has distributed the most vaccines, and how quickly they’re doing so. You’ve got to get those two elements in right now but also show how quickly things are changing. The problem with a line chart is that after a while it gets too crowded, so we could do heat maps where each country is one line and the colours show the rate of rollout. This is a fast-moving story, so we’re taking an iterative approach. We’ve got a simple bar chart out already, just showing the percentage of population vaccinated in each country, but that was built with a view to taking different forms over the coming weeks.

Is this an example of your minimalist approach, making sure there’s not too much for the reader to take in?
There is a very skilled design and data visualisation team here. We could build fancy, arcane charts all the time if we wanted to, and we have done that in the past. When we produce those more freeform creative charts, we will get a very intense positive reaction – but from a very narrow set of people, generally people who work in this same space. And everyone else will either say they didn’t get it, or they won’t say anything. Whereas if we make a clear line chart, hundreds of thousands of people across the world have got something useful from what we have done. So we do flex our creative muscles from time to time, but we are not afraid to do a story which just has five line charts in it if we think that is the best way of telling that story.

You also have the authority of the FT as a news source …
Trust is a big part of it. Lots of people are visualising data to do with the pandemic and some of that is from people who are setting out to mislead. The fact we have the brand of the FT stamped on our work is huge. If people know anything about it, it is that it’s an old-school, serious financial organisation and that helps us.

Having the colour pink as part of the FT identity is useful – you can spot it in a crowded landscape.
When you think about design and accessibility, it would be easier for us to have a white background – because of the range of colours that have a sufficient contrast – but, in terms of branding, the fact that anyone can spot an FT chart a mile off is a huge benefit.

You mentioned building everything in JavaScript. Is that the case whether it ends up in print, video or online?
Yes, the D3 JavaScript library essentially manipulates SVG files. Everything that you draw in this library comes out as an SVG. So we can convert that SVG to a PNG to use on our website, but we can also take it straight into Illustrator or straight to print. We also do a lot of animation using software like After Effects. We have made the templates to be compatible with all our outputs, so when we make a single chart (and this is another of the benefits of doing it through JavaScript), we only code one chart but the output of that is a mobile phone-sized version, a desktop-sized version, a print version, a social media version and a video version. So we only do the task once but get all of those outputs.

John L. Walters, editor of Eye, London

First published in Eye no. 101 vol. 26, 2021

More on the topic of Covid-19 data visualisations in:
The pandemic that launched a thousand visualisations’ (Eye 101) by Paul Kahn
Visual systems of life and death’ (web-only) by Paul Kahn

Eye is the world’s most beautiful and collectable graphic design journal, published for professional designers, students and anyone interested in critical, informed writing about graphic design and visual culture. It is available from all good design bookshops and online at the Eye shop, where you can buy subscriptions and single issues.