Bloomberg Vaccine Tracker
Country/area: United States
Organisation size: Big
Publication date: 4 Dec 2020
Credit: By: Tom Randall, Cedric Sam, Andre Tartar, Christopher Cannon and Paul Murray; Editors: Drew Armstrong and Yue Qiu
The Bloomberg Vaccine Tracker is the most up-to-date and comprehensive tally of vaccinations in the U.S. and around the globe, powered by a network of Bloomberg reporters in more than 50 countries who collect data from local sources unavailable in any other government or public dashboard.
Bloomberg’s tracker has been cited repeatedly by state governments (including Illinois, West Virginia, Connecticut and California), elected officials, international governments and news organizations (including MSNBC, PBS, Politico and Axios).
California, after falling behind most other states in its vaccination campaign on Bloomberg’s rankings, launched a statewide data review and found that health-care providers were not properly submitting vaccination records. Maryland also moved to address what it said were problems with slow reporting because of its position on Bloomberg’s ranking. In other states, citizens have used Bloomberg’s dashboard and rankings to ask state officials why vaccine rollouts are moving ahead slowly.
In South Korea — lauded for its pandemic response — local media used Bloomberg’s database of vaccine contracts to ask the government why it had not announced deals to acquire the shots. The deals were announced days later.
The data entry and fact-checking are done through a massive and well-designed Google Sheets system, where reporters and editors take shifts working around the clock, entering data from government sources, and check each others’ work. The data then goes through a processing and cleaning pipeline to eliminate questionable entries. This data pipeline is built in Node.js. Historical data is reconstructed through daily snapshots of the data and takes into consideration government data revises.
What was the hardest part of this project?
The hardest part is to built and maintain this comprehensive vaccination database that doesn’t exist anywhere. Because of the lack of international and even national data disclosure standards, it’s challenging to deal with the data inconsistency issues. It is also challenging to build a clean and well maintained data pipeline, while coming up with the most relevant graphics that captures what people care about the most in the news
What can others learn from this project?
The lesson that other journalists can learn from the project is that automation cannot replace reporting. At the very beginning, we considered making the process fully automated, scraping states’ and countries’ vaccine dashboards. We soon realized that it would highly limit the number of states and countries we are able to include in the tracker. The half-automated process we finally decided on allowed us to include as many states and countries as possible, and at the same time to have built-in measures to prevent human errors.