Flattening the curve: How well is your county doing coronavirus social distancing?

The coronavirus is spreading quickly through the United States, and many elected officials across the country are reacting with policies designed to slow the transmission of the pandemic, otherwise known as “flattening the curve.”

The coronavirus is spreading quickly through the United States, and many elected officials across the country are reacting with policies designed to slow the transmission of the pandemic, otherwise known as “flattening the curve.” One of the key methods they’re encouraging is social distancing—the practice of people limiting their movements and contact with others to limit the possibility for the virus to jump from person to person. Since it was used effectively during the flu outbreak of 1918, social distancing has been widely agreed upon as one of the strongest collective actions we can take against the virus. But how do you know if the people around you are practicing it?

That was the question that the team at Unacast, a location-data company, set out to answer. In a span of just a few days, their team of data scientists and experts pulled their data, which comes from apps like games and shopping platforms that millions of Americans have on their phones, into Tableau to assess how people’s movements are changing—or not changing—in relation to local policies and the number of COVID-19 cases. You can explore the visualization for yourself on Tableau Public (and see what grade your county receives for social distancing). For more context, we chatted with two of Unacast’s data scientists—Ekaterina Kuzmina and Mathias Schläffer—to learn about the data they worked with to create this viz, and what decisions they made to represent social distancing in a way that everyone can understand and visualize.

Tableau: The data you’re working with is very sensitive—you’re looking at people’s movements and how they change in response to a crisis situation. How were you able to present it in such a way that preserves people’s privacy and security while still relaying the necessary information?

Mathias Schläffer: When you work with Unacast, we make it clear from the very beginning that we have to be fully transparent. We have a feeling of great responsibility because we have such sensitive data. For this visualization, it was very clear that we had to aggregate the data to a level where privacy is protected as much as possible, while still providing detailed insight to policymakers on a relatively small spatial level. We aggregated the data to the county level so we would not expose any small-scale individual travel behaviors, but were still able to show the differences within states that were important and quite significant.

Tableau: What got you interested in visualizing social distancing in particular?

Ekaterina Kuzmina: It was a huge change for the whole world when the virus struck. We knew that we had resources, we had data, we had our people, and we had passion to do something to help. Based on the data we have, we could help to better understand the problem, and inform some decisions in the future and in the present. Basically, we took the data that we had, applied the capacity and resources that we had, and created this visualization in just a few days.

We wanted to focus on social distancing because it’s one of the rare ways we can overcome this pandemic. There is not much we can do, but everyone can help change the situation if they obey the rules of social distancing. But we noticed that there was no clear way to actually detect if people were actually doing it, if they were obeying the rules or not. We had the data to at least do a quick analysis of that, and show it to others.

Tableau: You have so much data—how did you identify which data could help visualize social distancing? What metric did you use to convey how people are doing social distancing initially?

Mathias Schläffer: Our objective was to react very quickly. It was just around when Italy went into lockdown, and where certain policy measures in the U.S. were introduced. We wanted to show very quickly how people were reacting to those policy measures in a way that would be rather simple to understand for anyone looking at our dashboard, and that we could apply to every county. So we went for a metric that did not require any strong assumptions in terms of where people live or where people work—all it measures is the distance between signals that we get from them day over day. We then compared the same region on the same day pre-coronavirus versus post-coronavirus. That showed to be relatively robust in terms of supply fluctuations and different signal attributes that we receive. And it showed a very strong correlation with the number of coronavirus cases, and with the news headlines that came out.

Ekaterina Kuzmina: We were quite happy when we saw that our metric correlates well with the number of confirmed new cases. We felt that the metric was validated, and what we were doing made sense.

Tableau: That was the first metric you used to convey social distancing—how are you adding more detail to the visualization to even more clearly understand how people’s movements are changing?

Mathias Schläffer: A lot of policymakers have reacted with the advice to avoid all visitations to retailer services that are deemed non-essential for survival. Grocery stores, pharmacies, or any healthcare services are essential, and non-essential would be bars, cinemas, the gym—all the things that you can easily avoid in the state of an emergency. We want to look at where and when those recommendations are happening, and see if we can observe a drop in visits to places by category. It adds another layer to understanding this behavioral change.

Tableau: What other changes are you hoping to make to the visualization going forward?

Mathias Schläffer: We’ve gotten a lot of feedback that so far, we've looked at behavioral change in a way that favors those regions where behavioral change was the most necessary. So think about the difference between New York City and Wyoming. In the future, we want to incorporate a metric that measures the natural social distance in a place. So if people in a densely populated area isolate themselves, then we have a better way to understand their movements in relation to people in Wyoming who might be more naturally isolated.

Tableau: How did you decide how to present this data in the visualization? What inspired you to use a scorecard to show different counties’ performance on social distancing?

Mathias Schläffer: The early aim was certainly to make the visualization something that everybody can understand just by having a quick glimpse at it—it has to resonate with everyone. That's why we went for the school grades—everybody has some sort of connection to school grades and can immediately differentiate between an A and a C. And then of course, we also wanted to provide a bit more detail for those who are interested in the methodology and the metrics themselves. But it should be something that not only the experts, but really everyone can use to look up their community and see how they're performing, and then think about if that is sufficient, or if they see any need to change something in their own behavior.

To see more information and visualizations on data and the coronavirus outbreak, visit Tableau’s COVID-19 Data Hub. This resource is constantly developing along with the situation, and we’ll be highlighting more stories the weeks to come.