COVID-19 at afp: local and global DATA JOURNALISM February 1, 2021

Agence France-Presse (AFP) is a leading global news agency providing fast, comprehensive and verified coverage of the events shaping our world drawing from a newsgathering network across 151 countries. One year ago, when photographer Hector Retamal sent this picture from Wuhan, we sensed that we would need to trigger the most ambitious global data journalism project we had ever launched.

As early as January 20, we checked the data on international flights out of Wuhan, and realized it was a major hub: more than 2,000 international flights in a single week. That, we feared, meant that the new coronavirus could rapidly expand.

We quickly discovered that covering the Covid-19 epidemic would mean data gathering operations in dozens of countries.

As a major news organization, we decided we could not rely on secondary sources to cover the path of the epidemic. We needed to provide fast updates, alerts and trends. Relying on other organization’s sources, schedules and updates was risky. Johns Hopkins University for instance, had a very good base, but waiting for their updates could significantly delay us.

As Covid-19 spread across China, the data team embarked on a combination of old school and sophisticated data journalism. On one hand, it coordinated data gathering on the ground, sometimes in countries where PDF's and even faxes were still the norm and having a “Rolodex” was more valuable than knowing about scraping. On the other hand, we had to be able to interpret the data in a timely and reliable way.

For over a year now, our newsroom of over 1,700 journalists across 151 bureaus collected data providing the raw material that would help us create our own database, and later, automated daily tallies, graphics, and videos.

Thanks to that labour intensive work, we were able to provide verified updates several times a day in six languages. We also had what we needed to create new features such as our Covid data-driven videos.


We had a great Data and graphics department we could rely on: 39 people from 10 different nationalities and backgrounds: web designers, developers and data journalists based in Hong Kong, London, Beirut, Berlin, Montevideo and Rio.

Four data journalists, two web designers and one database specialist were appointed “guardians” of the base. They vetted the data and cleaned the base, calling bureaus and correcting mistakes when discrepancies were found.

They were also in charge of sending alerts for major figures. Some sleepless nights were had, with an alarm clock ringing every hour when major thresholds, such as the tragic "one million deaths" mark, were about to be crossed.

Basic data journalism questions arose every day in the newsroom and within the team: how should we compare countries with different populations ? How are cases counted by official sources ? How should we report the infamous "Monday surge" — a day of high figures as weekend data collection lags behind.

We did compare our tallies with those published by Johns Hopkins University and the World Health Organization, and, when discrepancies were found, went back to the sources to make sure there were no mistakes.

The database has now been up and running for more than 330 days. Our base only relies on official sources, with one exception: for the United States, we use Johns Hopkins tallies and cross-check them with the data provided by the Centers for Disease Control (CDC). In Afghanistan, we use WHO data as there is no daily data available from official sources.

The database is then used by our reporters back on the ground to detect trends and provide context. The base now has several "children", including one to follow trends with its own dashboard used by the newsroom to anticipate outbreaks and detect upward and downward curves.

AUTOMATed content

Once the database was up and running, we created automatic content in text, graphics and video.

As the world’s streets were falling silent, our data journalists were creating semi-automated text tallies for our wires in French, English and Spanish.

The tallies were directly extracted from the base before one last edit from desk editors.

We were also able to create daily and weekly automated graphics, such as these:

Our tallies, threshold alerts and regional angles have made headlines around the world. We have produced over 16,000 Covid-related graphics delivered via satellite, FTP and our AFP.com platform.

The number of downloads only reflects part of the metrics but it does show high levels of use. 10, 496 graphics were downloaded in Spanish; 8,172 in English; 1,624 in French, 1,077 in German, 1,531 in Arabic and 202 in Portuguese from January 2020 to January 2021.

Our interactive team also designed Data driven videos (DDV), connected to the Covid database with special templates to produce 10 to 15 second video sequences. These feature either the number of cases or deaths or the trend of the pandemic for any given country, region or the world. The videos can be created with three clicks by video or social media editors, without motion design or dataskills, a convenient tool for breaking news content.

We built the tool using technology developed with a Google DNI grant for the AFP Data project. In previous versions of the DDV, we created short videos for tweets and location. Data driven videos can be produced in French, English, Spanish and Portuguese. Those short videos can be used by broadcasters alone, or as part of a story. They are also used on Twitter.

DATA for the public

The Covid-19 data we gathered was made available to the public via our AFP Interactive page, where cases and deaths are automatically updated in five languages (French, Spanish, English, German, Italian). The embeddable version of this interactive graphic has been viewed 96 million times since March 2020.

The page also shows data for the 10 worst-affected countries and their trends

In March and April, our documentation department also maintained a database to count the number of people under lockdown.

We wanted to measure how the world had come to a halt and monitored four variables: mandatory lockdowns, curfews, quarantines, and "people invited to stay at home". To calculate the percentage of people invited or forced to stay at home, we used UN population projections for 2020. Our lockdown tallies made headlines in many countries.


We also looked for new graphic ways to tell the story to fight against "Covid fatigue".

Last but not least, we produced graphics detailing the impact on people's daily lives.

Eventually, our media clients could rely on us for timely data updates and therefore spend more time focusing on more specialized issues. We believe no other international news organization provided such a massive amount of Covid data texts, graphics and data-driven videos in six languages.

Thank you for your time !