One of the highlights of the 2019 Canadian election were the two political leaders debates, organized by a consortium of national media outlets. Since Canada is a bilingual country, one debate was in French and the other one was in English. For the French one, we produced this data visualization during the night, broken in three parts, to explain what happened to the voters.
This project was the front story of our national website the morning after the debate. At first, the plan was to have a simple text article to summarize what happened during the debate. We proposed this project instead. The stakes were high because the leaders’ debate is a very critical event in Canadian elections. But we’ve been able to demonstrate that it’s possible to publish a complex data visualization story very quickly. The public responded very well to it with great engagement time and many positive comments on our work. We were the only media outlet to have a data-driven story published only a few hours after this live event on TV, which really put us in the spotlight.
During the debate, four people were in charge of filling a Google Sheet with the details of each intervention by a leader. They worked on a recorded version of the debate, to be able to listen several times to some interventions or to ask second opinions from someone on the team if needed.
For the analysis and the data visualizations, many things were coded before the debate. A Python script connected to the Google Sheet and calculated several indicators (number of attacks, most talked topic by the leaders, etc.) to help to point to the most interesting facts of the debate. This Python script also outputted a JSON file directly usable for the data visualizations.
We decided on the order of the story elements and we wrote the story itself right after the debate.
What was the hardest part of this project?
The pressure was very high for publishing this project. The debate was scheduled to start at 8:00 pm and to end at 10:00 pm. Our story had to be published before 6:00 am the next morning. We had roughly 10 hours to make an entire transcript of the debate, to identify each topic, each attack, to write the story and to code complete data visualizations.
Since the leader debates is a critical part of Canadian elections, we did a general rehearsal a few days before the French debate to test our plan and our code. We actually failed that night, unable to publish anything before the deadline! We made many changes in the way we worked, which made us faster and more efficient in our tasks.
This whole planning and rehearsal process was key to our success.
What can others learn from this project?
I hope other data reporters will learn that it could be very interesting to plan a data-driven visualizations coverage for some scheduled events (other than elections). We know in advance that some events are critical and will be talked about. It’s an opportunity to push forward an innovative work that will have great visibility.
To do that, planning is key. From the editorial, design and coding aspects, everything has to be flexible for last-minute changes. You need to automate as much as you can to keep your focus on what is really important in the end: content. It’s a lesson that we try to apply to all of our projects from now on.