In early March, Aatish developed a Python code notebook to gather & analyze daily global COVID case numbers. He noticed that cases in different regions followed a similar pattern of exponential growth. This led him to collaborate with Henry Reich, who creates the science YouTube channel Minute Physics, to create the interactive graph Covid Trends , and also create an explainer video introducing this graph . Covid Trends received many millions of visitors in 2020, and our companion video (How To Tell If We’re Beating COVID-19) was viewed 6 million times.
Covid Trends visualized trends at a time when most venues were reporting daily numbers in isolation. It was one of the first COVID graphs to employ a 7-day average and a logarithmic scale, and it used a novel set of axes (recent cases vs cumulative cases) to emphasize deviations from exponential growth. This work inspired similar graphs at NPR, The Financial Times, Our World in Data, the Pennsylvania Department of Public Health , and the Indian Statistical Institute . It was featured in IndiaSpend (a data journalism non-profit)  and Popular Mechanics’ list of the best COVID graphs , was cited in over a dozen academic publications , incorporated input from 25 contributors , and was forked on Github over 100 times.
What was the hardest part of this project?
The most challenging aspect of this project was finding a representation of the data that clearly highlighted trends at a glance. After a few weeks of experimentation, we hit upon a few useful strategies that visually emphasized when cases in a country grew along an exponential growth trajectory (and also when cases deviated from this trajectory):
1. We employed a 7-day window to smooth out the data (while this is now standard, this was rare in early March 2020).
2. We employed a logarithmic scale and a novel choice of axes to help people easily identify exponential growth trends as straight lines in the graph. Because of our background in physics, we were familiar with the technique of representing data in ‘phase space’ and employed this idea here.
3. As time is not one of the axes in our graph, we instead used an animation to represent time. By representing time with motion, this animated view allowed viewers to more easily grasp the growth of cases over time at a glance.
What can others learn from this project?
We open-sourced our codebase under the permissive MIT license. This project has been forked on GitHub over 100 times and incorporated input from over a dozen contributors . Other journalists can learn about how to create an open-source customizable interactive tool that continues to be useful to many people, incorporates volunteer contributions, and represents data through a lens that emphasizes trends rather than individual data points in isolation.