How We Accidentally Wrote Our Most Popular Story Yet and What We Learnt in the Process

Our readers have spent over five million minutes on our graphical overview of Covid-19, more than 2500 of them decided to buy a paid subscription. The article started out as a way to recycle a few charts and ended up highlighting four crucial issues of data journalism.

In the final days of January 2020, the NZZ newsroom decided to publish an in-depth piece on a new disease affecting the faraway Chinese province of Hubei, leading to a lockdown of around 40 million people there. The earliest charts we published showed a total of about 800 cases in China, a map showed seven countries affected by Covid-19.

It took almost a month for the virus to reach Switzerland and for the NZZ Visuals team to come up with the idea of a graphical overview. On February 25, the Swiss Federal Office of Public Health registered the first confirmed case in the country. By then, the number of cases in China had risen to over 75'000 and Johns Hopkins University had created a dashboard and a database documenting the global spread of the disease.

The database enabled the data journalists in our team to update charts with ease, sparing the newsroom having to manually collect the data. Quite a few charts had now been created and with the disease crossing Switzerland’s southern border, reader interest was intense. So on February 26, the Visuals team decided to recycle some of these charts in a small piece to give readers a quick overview on the numbers.

By the end of the day we realised, this was a quick win: The virus would stay with us for some time and the JHU database made it easy to quickly update many of the charts. What we did not realise was that we had just published what would become the most popular piece of journalism ever by the NZZ Visuals team.

That realisation came only a few days later — by March 6, we counted over half a million views and what previously had been a collection of code fragments was consolidated into a script that made it possible to update dozens of charts within 30 minutes, freeing a up a lot of time to evaluate the data, put it into context and look for new aspects of the spread that could be of interest.

In March, an early version of our piece made it to the NZZ print edition.

While some publications published dashboard-like overview pieces, we decided to go for a semi-automated piece. Although requiring a bit more journalistic work, it allowed us to shift the focus of the piece as the spread of the virus progressed from Asia to Europe and further from the USA to emerging economies providing a new angle every day, e.g. on the regional spread of cases in Switzerland, racial disparities in the US or rising numbers in Brazil, Mexico, India and Russia.

With the crisis unfolding before us, we constantly evaluated new developments and discussed the best way to implement them into the piece. We revisited our existing charts with the goal of delivering their message as clearly as possible and with as much context as needed. We eliminated every aspect that was no longer essential to our story, yet the number of charts continually grew from nine to 28.

With the scope, the number of people involved in updating the piece grew. What was initially done by one data journalist now features work of the entire data journalism and graphics staff of NZZ Visuals as well as many foreign correspondents and editors, science writers and newsroom journalists.

After three months, four lessons have become clear that will certainly remain valid well beyond Covid-19 and guide our work in the future. All of these points were true before Covid-19, but the global pandemic has once again highlighted their importance:

1. Visual and data journalism are indispensable in today’s newsrooms: With data so readily available, the need for people who can assess its quality and relevance, analyse it with the appropriate methods and visualise in the clearest way possible is evident. The graphical overview pieces on Covid-19 are among the most-read stories on record for many publications, and NZZ is no exception. Charts and data analysis can offer insights well beyond government reports or expert judgement and can also be used to hold these actors accountable. Any publication without a strong visual and data journalism team is now at a serious disadvantage.
2. Data quality needs to improve: For analysis to be precise and charts to be accurate, journalists are reliant on data that is correct, measures the right thing, is published in a consistent, open format and complemented by a detailed description. Data without context is very difficult to work with. Many of these points were not met in the Covid crisis, making it harder for journalists and other stakeholders to draw the correct conclusions efficiently. This affects policy making as much as journalism, so data quality and making that data machine readable need to be a top priority for government agencies everywhere.
3. Collaboration is essential: How did China contain the virus to a single province while many European nations fail to do so? Why is Covid-19 spreading more slowly in Brazil and why is the curve barely flattening? Why are the cumulative numbers in Spain suddenly much lower than yesterday? Data alone might not provide the answers to these questions, but for a foreign correspondent, they might be obvious. Why is the death rate in Germany lower than in Switzerland and how serious is a rising reproductive number? Health and science writers might know. Furthering the mutual understanding of different areas of expertise within the editorial staff of any publication is crucial to developing its full potential.
4. We need to explain ourselves: Journalists are accountable to the public as much as any government official or scientist. Data journalism often involves decisions and assumptions that are not obvious and need to be made transparent and to be communicated effectively. In the course of the last few months, NZZ Visuals has published pieces on why we use certain data sources and how this data is collected, on the advantages and drawbacks of different chart types and our thinking in selecting one over the other, on the limitations of different statistical indicators and on our method of estimating the number of patients who have recovered from Covid-19. We have also published the code behind our overview piece. Feedback from readers and colleagues was often spot-on, leading us correct serious mistakes such as displaying absolute numbers on a choropleth map or providing per capita figures without adequate contextualisation. It has also helped us to spot inconsistencies in the data and to improve the way we visualise and describe the data.

These remain high ideals — correct and detailed government figures were often hard to come by, there is certainly still a lot of unused potential for collaboration within the newsroom of NZZ and communication of our methods was often delayed. But being thrown in at the deep end by this virus hopefully helped these issues to be identified so that they can be addressed by the time the next crisis comes around. And hopefully for data reporting in the future as a whole.

If you have feedback, questions or want to share how your newsroom has taken on these challenges, please get in touch.