How tech startup Descartes Labs is helping Cargill unlock the future of data
September 24, 2018
What brought together a 153-year-old grain trader and a tech startup in an office full of bean bag chairs? Corn, naturally.
It was 2016 when James Weed, the global trading and analytics lead for Cargill’s agricultural supply chain enterprise, saw a demonstration from Descartes Labs, a New Mexico-based tech company adept at turning jumbles of data into usable information.
Descartes had built a digital model that used historical data to forecast corn production in the United States. It had a few rough edges, but it ended up with a 1 percent margin of error by the end of the growing season. Weed immediately recognized the potential in combining Descartes’ capabilities with Cargill’s deep and far-reaching data sets.
“James had a vision of how data would affect the business,” said Mark Johnson, Descartes Labs’ CEO and co-founder. “He knew that all the data Cargill was collecting had enormous potential if they found the right partner to help them understand it.”
Put to the test
Descartes Labs’ specialty — satellite data — has been around for more than half a century. But in spite of significant investments in satellite hardware, there had been relatively little progress in turning terabytes upon terabytes of raw information collected into commercial insights.
The tech company saw an opportunity to be the world’s data refinery. Its goal: to clean the “dirty” data produced by the myriad sensors collecting them into something ready for analysis.
The agriculture industry in particular intrigued Descartes Labs as a big, complex challenge.
"It’s what we call a wider research problem,” Johnson said. “The earth is very large at 150 million square kilometers. You don’t know where all the fields are. You don’t know what’s growing in them. So, it’s a hard problem. It’s the kind of problem that you could really sink your teeth into.”
Weed’s first sit-down with Descartes Labs impressed him, even if the company’s office space at the time— bean bag chairs, video games, a folding table — was a little different than his typical corporate meeting.
Against that backdrop, “they also had this room full of, quite frankly, some of the brightest Ph.D.s working in satellite observation and astrophysics, and they were all focused on corn,” Weed said.
He decided to give Descartes Labs a challenge: over an eight-week period, try to make sense of publicly available data sets and Cargill’s data sets in a way that would generate insights that Cargill wasn’t getting through its own analysis.
“We didn’t tell them what the data was. We changed the names. We made it a bit of a mess intentionally to see if they could handle that and use it in forecasting crops,” said Weed.
What they came up with, according to Weed, was a model that was markedly better than what either company had previously used. “They proved themselves and the technology,” he said.
The art of the possible
For Descartes Labs, a partnership with Cargill offers the young company a chance to explore complex global problems while providing access to a vast and unique private data set that can better inform their work. Johnson notes that the partnership is also teaching him about the kind of business he wants to run.
“I have so much respect for Cargill as a company, its history, its values, its people,” he said. “I think about Cargill as a model. How are we going to be around 150 years from now?”
Weed says that the collaboration has helped Cargill think more broadly about how it uses data. “It all comes back to creating unique data sets and using those to run our supply chains better, trade better and create more value for our customers.”
The partnership is one way that Cargill is quickly moving to become a technology leader in food and agriculture.
“The only thing that limits us is our imagination,” said Cargill Chief Technology Officer Keith Narr. “Further integrating technology into our operations has opened our eyes to the art of the possible.”