In 2024, Northeastern professor Shawn Bhimani helped launch Supply Trace, which uses machine learning to comb through millions of data points including news clippings, container labels and shipping documents.
When Shawn Bhimani and his colleagues launched Supply Trace one year ago, the goal of the interactive database was to shed light on the obscured global origins of certain products that make their way to the United States. To start, the project focused on forced labor and human rights violations in the clothing industry — a problem as rampant as it is difficult to track.
“Something we’ve been doing the last couple of months is trying to understand which cities illicit goods are floating through,” says Bhimani, an assistant professor of supply chain management at Northeastern University. “If you think about China to the U.S., what ports are forced labor goods entering most often?”
Supply Trace, which now has about 1,200 subscribers from the retail industry, allows its users to answer such questions, charting the path of apparel goods around the world. Bhimani and his collaborators can connect hundreds of millions of data points on imports in purchasing countries with in-person reports detailing the locations and practices of overseas facilities that use forced labor.
To date, the platform has cataloged nearly 4 million businesses globally — a scale that wouldn’t be possible without artificial intelligence.
“Before AI became mainstream, working manually, we could analyze a few dozen companies in a semester,” Bhimani says. “Now we can do that in a day. It’s sped up by orders of magnitude.”
As Bhimani explains, there are two sides to how Supply Trace uses AI. “Machine learning is used to be able to read import records that come through Customs and Border Protection in the United States and parse that data to understand what was in the shipment,” he explains. “There are certain codes that CBP uses to denote, for example, clothing versus electronics, so we can draw conclusions at a scale that we wouldn’t be able to do as human beings.”
Machine learning also helps with what the researchers call “fuzzy matching,” or identifying a certain business across different international transactions.
“You may have a production facility in China that matches another that the U.S. buys from, but with a slightly different name,” Bhimani says. “We can do some matching to figure the probability that these 10 different companies are buying from the same factory, but calling it by a slightly different name. And we have to do a lot of cleaning of the data to ensure that that is done correctly.”
In addition, AI can bulk-analyze data like product descriptions, revealing patterns of which types of goods are coming from which shipments. With all of this information, the researchers can uncover supply chain routes and their respective risk levels for U.S. buyers who want to avoid companies that use forced labor.
Supply Trace is growing, both in size and scope: the platform recently received a $1.7 million grant to build out climate risk data on top of its forced labor research.
“It helps users make better business decisions and companies to be more sustainable,” Bhimani says.