How satellites, drones, and planes are making hedge funds money
Every morning, hours before the opening bell signals the start of another trading day on Wall Street, an undisclosed group of hedge fund managers receive something they’ve spent good money on: the latest estimate of the world’s crude-oil supply. Oil happens to be the most closely watched commodity in the global economy, and many countries are tight-lipped about how much they’ve stowed away. So traders pay to eliminate the guesswork — to find out the number of barrels in oil-rich states like Louisiana and Texas, or the entire stockpile of Saudi Arabia.
Three thousand miles west in Mountain View, California, lies the source of that oil data, a company called Orbital Insight, which, according to its mission statement, finds “truth and transparency” in the world’s rhythms. What that means in practice is that roughly 30 engineers and scientists spend their days sifting through satellite images for information their customers — not just hedge funds but also asset managers, insurance companies, and government agencies — want. The number of ships leaving China’s ports. The total cars parked outside every Lowe’s in the United States. The income distribution of a district in Sri Lanka. In the case of oil volumes, the key is in the shadows.
Jason Lohn, a mild-mannered 52-year-old Orbital engineer, once worked on a NASA mission that examined the moon’s surface. Now he studies bird’s-eye images of the Earth from his standing desk. On an overcast day in June, Lohn opened his laptop to pull up a picture of an oil tank. From above, it resembled a flat gray disk. Lohn pointed to a moon-shaped shadow cast by the walls onto the lid of the tank. The lid floats on top of the oil, bobbing up and down as the volume inside changes. “The bigger the crescent, the farther down the lid; the farther down the lid, the less oil there is,” he said. Every night, while Orbital’s engineers sleep, an algorithm spends hours converting shadows into volumes.
Before satellite technology, if some enterprising hedge funder wanted to get a competitive edge on oil quantities, he might have hired a person to watch oil tankers move in and out of a port through binoculars. In the ’80s and ’90s, some analysts tried to track the departure and landing schedules of corporate planes in the hope that they would hint at a merger or deal. In the 1950s, Sam Walton flew over Walmart parking lots in a helicopter so he could count cars and assess his real estate investments. (An 0ft-repeated precedent involves Baron Rothschild, who as legend has it received early word of the victory at the Battle of Waterloo via carrier pigeon and made a fortune by buying up British government bonds.) What Orbital sells, some argue, are the tools to make these calculations at scale: Instead of counting one parking lot or one plane at a time, you can have a machine track all of them. “Waiting a quarter to see how a company is performing is nonsense,” says Matei Zatreanu, a former hedge fund analyst who consults for financial firms.
Orbital Insight was founded in 2013 by James “Jimi” Crawford, an engineer who led a project at Google to convert the world’s books into searchable text. It was around the time of Orbital’s launch that private companies began sending fleets of satellites into orbit, which yielded a glut of cheap, constantly refreshed images of the Earth’s surface. Data was also pouring in from the digitized world in other forms: from the geolocation coordinates tracking our movements to the email receipts tracking our purchases. Many of the companies that produced this raw data, however, lacked the tools to process it, which created an opening for “alternative data” companies. In the past five years, hundreds of them, Orbital included, have cropped up around the world, and they wade through text or pixels in search of trends.
At Orbital, a typical project begins with a broad question that a client is puzzling over: Are chain retailers dying? How is China’s trade shifting? For each of these questions, engineers have to make the abstract concrete. The leap can be an intuitive one — tabulating, for example, the cars in every Macau casino parking lot to predict how well its gambling sector will fare next quarter. When a counting algorithm gets perfected, when the car counter can accurately pick up the cars and leave out the port-a-potties and dumpsters, it can help answer even bigger questions about a given place: the degree of urbanization, the level of gas demand, the fluctuations in population. “You’re building these Lego bricks, and you can put bricks together in new ways,” says Boris Babenko, an engineer who trains Orbital’s software to recognize those cars and planes and ships. Last year, the World Bank approached Orbital to find a way to measure economic inequality. “You could think about it and say, Well, maybe we can look at the number of houses, or the shape of houses, or the presence of pools,” Babenko says. The team ultimately settled on four indicators of poverty using the algorithms it had on hand: the number of cars (more suggests higher incomes), the height of buildings (urban areas have higher buildings, and in the developing world, those areas tend to be wealthier), the construction activity (a proxy for population growth and migration), and the type of vegetation (in rural areas, thick greenery signals poverty; in urban areas, lushness signals wealth).
Since its founding, Orbital has attracted nearly $79 million in funding (one of its investors is In-Q-Tel, the venture capital arm of the CIA). And the industry at large is doing well: A recent report by J.P. Morgan estimates that investors currently spend $2 billion to $3 billion on big data, a number that’s expected to climb. With alternative data being churned out at such a pace, however, regulation still has to catch up. Mike Gantcher is the head of sales at RS Metrics, which uses planes, drones, and satellites to measure traffic outside retail stores. When clients approach RS Metrics, he says, they often question the legality of using data in this way. “That’s the first thing hedge funds ask, so they don’t go to jail,” Gantcher says. If a person were to make a trade based on nonpublic knowledge — such as a tip from a shipping executive about the contents of a cargo container — he could be accused of insider trading.
The Securities and Exchange Commission has yet to weigh in, so, for now, firms are careful to stick with data that’s publicly available: Twitter feeds, radio broadcasts from cargo ships, photographs from space. Then there’s the issue of privacy. Placed, a company that uses smartphone location data to estimate how online ads impact retail foot traffic, masks any trace of identifying information (Placed was acquired by Snap in June for $200 million). And firms like Orbital rely on lower-resolution images in which people aren’t distinguishable. Besides, a person’s identity is a rare bit of data that to hedge funders isn’t worth a whole lot. “We’re not marketers,” says Zatreanu. “We want to understand not just you but what people like you are doing.”