Skip to main content
Feature

Leading through COVID: Repurposing with a Purpose

In this series, we talk with Yale SOM alumni about their professional and personal lives during the global pandemic. David Browning ’99, CEO of Enveritas, explains how the nonprofit is providing public health data using tools created to verify the sustainability of coffee.

Adapted from phone interviews on May 13, 2020, and July 29, 2020.

Enveritas is a young nonprofit organization, launched in 2016. It was founded to help small-holder coffee farmers access sustainability verification. Consumers want to know their coffee was not produced using child labor, but verification can be expensive for farmers. We were able to make it free for them by re-engineering the way it is done. 

Using a combination of machine learning and boots on the ground, we have an efficient accurate system where multinationals pay us to verify global supply chains for environmental, social, and economic factors such as child labor, slavery, and deforestation. We operate in Africa, Asia, and Latin America. 

We’ve been highly virtual from day one, but as COVID started to build, we had people scattered across 15 countries. We worked to get people back to their home healthcare systems because, for many of the countries that we’re operating in, it was uncertain to what degree our team could get healthcare if they needed it, and how overloaded health systems might become. 

We paused much of our in-person verification work. Country by country, we resume based on guidance from local governments. The pause has created capacity to check in with farmers and their industries. The public health measures imposed have really varied. As have the economic supports offered to farmers. In some cases, farmers are not allowed any activity on the farm. In others, farmers have delayed work because of the restrictions that are in place. Some farmers are having trouble accessing inputs. Or they have changed how they pay workers because banks aren’t open. One of the most challenging issues has been figuring out protocols for carrying out harvest while keeping workers socially distanced. 

At the same time, the world’s coffee prices have become very volatile as markets try to divine whether harvests—which happen at different times around the world depending on the latitude and altitude of farms—will disrupt the supply side. And obviously, there has also been demand disruption where coffee shops, restaurants, and office buildings have been closed.

With our traditional work potentially sidelined due to restrictions, we saw we could contribute by using our capabilities to conduct randomized COVID testing in the countries in which we work. 

We had already expanded our AI work into the cocoa industry, and we knew our platform could be applied to other areas. We’ve been working with the government in Cote d’Ivoire to support their education goals. With machine learning, we were able to identify all of their schools using satellite imagery, understand how school locations compare with population density, and combine this information with rapid, low-cost literacy and numeracy assessments across thousands of schools to support policy making. This “big data” can help answer questions such as, “With limited funds, should we invest in textbooks, new school buildings, or teacher training, and where?”

“Leaders will still face very tough decisions, but an accurate sense of how many people have had the disease, and at what speed it is progressing, can help them make informed policy choices.”

Policy makers responding to the pandemic could also benefit from large randomized population surveys. Door-to-door surveys largely faded out decades ago, as telephones and then later the internet offered far more efficient means of conducting surveys. But for the work we do, checking for issues such as deforestation and child labor, we need to physically verify using door-to-door surveys. 

With the pandemic, randomized door-to-door testing offered the potential to reduce bias while conducting testing. Testing only in hospitals would mean missing large numbers of asymptomatic cases (those who had contracted COVID but had displayed no symptoms). Conducting testing by asking for volunteers could create a biased sample. Setting up a testing station at a supermarket could also introduce bias, as many segments of population, such as the elderly, may have been avoiding supermarkets during the pandemic. A randomized door-to-door survey could help address this issue of biased samples.

When COVID arrived, we realized that our platform would be useful for a particular gap in our knowledge. PCR testing, which looks for the presence of the virus and accounts for most of the testing being done, is extremely valuable for a doctor in an emergency ward asking if a patient has COVID or not. But PCR tests are less effective for answering the public health question of how many people have contracted the virus because PCR tests will yield a positive result only during a brief window of time while infection is occurring. 

Randomized antibody testing can determine what percentage of the population has had COVID, including asymptomatic cases. And it can be done rapidly, at a low cost, across large populations. There has been a great deal of negative press surrounding the use of antibody testing and seroprevalence surveys, but when conducted for population surveys, as opposed to individual diagnostics, and when appropriate statistical guardrails are put in place, they have a valuable role to play.

Right now, at the end of July, there are over 4.5 million confirmed or probable cases in the U.S. But based on antibody tests, the CDC estimated that perhaps 10 times as many people may have had COVID. Asymptomatic carriers may not be aware they have the virus but they can still transmit it on to others. If we knew how many people have had it and we knew how much that is changing week by week, that would be helpful for policymakers trying to decide whether to shut down economic and social movement or loosen restrictions. Leaders will still face very tough decisions, but an accurate sense of how many people have had the disease, and at what speed it is progressing, can help them make informed policy choices. 

Our platform was developed for effective randomization in places where street numbers don’t exist, and where street names don’t exist. On top of that, we have proprietary GPS tools that allow us to verify the in-person door-to-door work. 

We’re supporting governments in both Brazil and Ethiopia to cover communities of over 4 million people. With generous funding from Nespresso, Peets, and the J. M. Smucker Company, weve been able to conduct the work as a public service to each country.

Department: Feature