Home | Alumni Magazine

Making the Global Food Supply Chain Safer

Retsef Levi Retsef Levi

Three years ago, Retsef Levi knocked on Tauhid Zaman’s office door.

Zaman, KDD Career Development Professor in Communications and Technology and assistant professor of operations management, studies social network data—how to catch ISIS operatives using Twitter, for example. Levi, J. Spencer Standish (1945) Professor of Operations Management, was working on a project for the U.S. Food & Drug Administration (FDA)—a risk model for stopping adulterated food imports at the border. Could Zaman join the team? “I had no idea how I’d fit in,” Zaman says. But Levi and co-principal investigator Yasheng Huang, International Program Professor in Chinese Economy and Business, were right. Zaman was a piece of the puzzle—as were MIT Sloan’s Y. Karen Zheng; the Center for Biomedical Innovation’s Stacy Springs and Anthony Sinskey; and others from MIT’s Electrical Engineering and Computer Science departments. Last fall, they delivered the predictive analytics tools they developed to the FDA.

Yasheng HuangYasheng Huang

Now, with millions of dollars in funding from the Walmart Foundation and the HNA Group, the interdisciplinary team is digging deeper into food supply chains in China, the world’s third-largest food exporter with more than a billion of its own consumers. The researchers are collaborating with universities, industry partners, and government institutions there to build far greater transparency into food supply chains, and better understanding of risk drivers. Ultimately, they hope that their predictive models will prevent the purposeful adulteration—of food.

The seed for this project was planted back in 2011, when the federal government passed the Food Safety Modernization Act. The United States imports about 40 million individual shipments of food every year, and many of those shipments are coming from countries that lack the basic quality controls in place in the United States. It would be simply impossible to check every box—and yet, farmers and manufacturers have adulterated food for a variety of reasons. “In response to avian flu, for example, there was an increase in extensive use of antibiotics, antivirals, and herbal medicines” in poultry farming throughout Asia, Levi says. In 2008, the Chinese government started requiring a level of protein in milk and at the same time capped its price, which prompted some in the dairy trade to add melamine—a poison—to fake their way through the tests. And during outbreaks of early mortality syndrome in 2009 and 2013, farmers increased their use of antibiotics and other additives in shrimp. “All of these cases indicate that you might be able to understand what kind of socioeconomic and environmental drivers could increase the level of risk” that a particular food might have been tampered with, Levi says. So when the FDA needed some new tools, they approached MIT for help. The Abdul Latif Jameel World Water and Food Security Lab (J-WAFS), MIT’s food and water security lab, also provided a grant of funding.

The researchers sat together with their students and brainstormed. “We had to approach it like detectives,” Zaman says. “What would make a company look fishy?” One month in, the researchers found a website with copies of bills of lading—the start-to-finish shipping records for every container that moves. They paired that data with FDA inspection records and started to look for patterns.

Sloan School Career Development Professor Y. Karen Zheng studies human behavior in supply chains. She targeted one thousand Chinese manufacturers of honey, poultry, pork, seafood, and eggs. She mapped their supply chains back to individual farmers, recording the volume each farmer provided, the distance from the farm to the factory, and more. To supplement the FDA sampling data, Zheng’s students gathered data online from other countries that import from China. “About halfway through the project, we were actually quite lucky,” Zheng says. “The Chinese government frequently samples products from retail shelves. In 2015, they started to publish those sampling results online.”

Stacy Springs, director of the Biomanufacturing Research Program at MIT's Center for Biomedical Innovation (CBI), focused on case studies of companies that had been caught adulterating food—both in the past and in some ongoing cases that aren’t yet fully understood. She zeroed in on which chemical or biological contaminants were popping up. “We tried to approach it very systematically by asking ourselves, ‘What are the potential adulterants?,’ and then characterizing the possibility that those have actually infiltrated the food supply,” she says.

Everyone had to deal with data that was dirty or slow in coming. The same company’s name could be spelled three different ways, for example. And government bureaucracies don’t disseminate data via firehose. But the team plugged away. Once or twice a week, they came together in a conference room at MIT Sloan or CBI to touch base, share data, and plan their next steps.

Here’s what they discovered. “The coolest thing we found is a way to predict if a company is going to be adulterating food. The company’s product network gives it away,” Zaman says. He’d theorized that companies with a more diverse mix of products would fail inspections more often. If Company A imports shrimp, fish, and squid, it’s usually okay, Zaman says: “It’s all seafood; they know their suppliers.” But if Company B imports shrimp and paint, “that’s kind of weird,” Zaman says. “To them, they’re just looking for another cheap product.” Zaman graphed the modularity of each company and confirmed that the diversity of a company’s product mix does, in fact, correlate with failed inspections. “I was surprised by how well our predictive model did—just that single feature,” he says.

Zheng and Huang were able to correlate regional regulatory oversight and even tax evasion with food shenanigans. They hypothesized that of 320 cities in China, those with weak governance were more likely to be the origin of an adulterated shipment. “One measure that seemed to stand out was the rank of officials engaged in misconduct. Manufacturers in cities where the mayor had engaged in misconduct had more problems,” Zheng says. Also, she says, “cities in which a large fraction of the companies engage in tax evasion are significantly correlated with risk. Food companies in those regions also turn out to have more problems.”

A recent gift from the Syngenta Foundation will fund Zheng’s newest research, which will focus on discovering when and how consumers value traceability of food products and how various sourcing strategies impact the resulting safety and quality of agricultural products.

Other risk factors the team identified include the number of farmers supplying a manufacturer, the number of shipments a company sends overseas, and whether an exporter ships through a middleman country on the way to the United States.

Meanwhile, the Walmart Foundation was already working on the food safety situation in China. Impressed with MIT's risk modeling, the foundation provided funding to take the project to the next level, including the development of new tests. Current food-testing methods require testers to know which contaminant they’re looking for. Biology professor Anthony Sinskey and new team member Michael Strano of Chemical Engineering are developing sensor platforms capable of measuring not just one but hundreds of potential adulterants simultaneously, in real time, on a hand-held device.

And the team is working with their Chinese counterparts as part of a food safety collaboration center. “We’ll be able to track stock prices and social media feeds and see hot spots lighting up on maps,” says Zaman. “It will be proactive. It will allow intervention.”

The collaboration is a prime example of a cross-campus project that might only happen at MIT. “It’s a rare opportunity to bring so much expertise together to tackle a problem like this,” says Springs. “Bringing this team together allows us to simultaneously view where the vulnerabilities in the supply chains are and then develop the technology to make them more secure.

”HNA Group, inspired by this collaboration, and the work in progress, is now helping MIT scale its potential impact. Huang, who presented the team’s findings to MIT alumni in Hong Kong at the Campaign for a Better World Tour last December, agrees. “It will be a game changer if we succeed,” he says. But the work is only beginning. “The problem is not going away,” Huang says. “We need to raise the resources that will give permanence to what we’re doing.”