As coronavirus restrictions began to lift in some U.S. states in May, Reuters analyzed data from millions of mobile phones to show how and where people began emerging from the initial lockdowns.
This project was widely viewed on Reuters.com and shared across social media. With the lack of clear national guidelines, regional parsed data like this became a valuable resource to understand the spectrum of risk the country was taking as a whole, and allow a reader to assess their personal risk in that context.
Reade Levinson used the R Stats Package to summarize and group data at the state and city level over a period of weeks. From there, we created SVG maps using D3.js. On our first pass, we built an animation for each map by category. It was slow to load and complicated to code. So we baked out the SVG and created a layout in Adobe Illustrator. We used the Ai2HTML library to load the static maps on the page as HTML and image assets.
The city-level heat strips were built in D3.js.
What was the hardest part of this project?
We built this on a very tight deadline. We knew the data provider was working with other news organizations, so we wanted to get this out quickly. We were able to turn this around in about 4 days thanks to Reade’s quick data analysis, good editing and some repurposed code from a recent mapping project.
What can others learn from this project?
Small multiples! We often have the urge to aggregate up, to make a single bigger and more complex display of data, when the opposite is really what’s required to make meaning. By repeating the maps across time and category, more vectors become clear to the end user. Regional differences, categorical differences, and time are all much easier to parse as patterns emerge across the sets of maps.