Advan Research Adds GPS-Based Business Closure, Road Traffic Data

New York-based geolocation provider Advan Research is preparing to roll out new datasets based on the location of mobile devices in the coming weeks, covering business closures and reopenings under measures designed to combat the spread of the Covid-19 disease, and detailed information on road traffic and shipping across the US.

Advan plans to launch the weekly dataset of business closures this week, followed around a month later by the road traffic dataset.

The Weekly Closures dataset uses an algorithm to determine whether a location is open or closed by comparing current visitor numbers to historical numbers. Users can then check information from companies and local and state government on when businesses are supposed to reopen, and compare that to the Advan data showing when a business actually reopened.

“This is a critical indicator of what will happen in the short term,” says Advan CEO and founder Yiannis Tsiounis.

Created in response to demand from two fundamental hedge fund clients, the dataset covers all industries in which the 2,250 companies it monitors operate, including “pretty much” every factory in the US, as well as warehouses, offices, and retail locations. The vendor has recently added 170 new companies to its oversight roster.

“We’ve published a baseline of typical traffic for a location, and if it shows less than 10% of its typical traffic—to account for owners or cleaning crews visiting the location—then we assume it is closed. And that’s accurate more than 95% of the time. We didn’t want to make it too complicated, which would have taken longer, because people need this data right now,” Tsiounis says.

Advan decided to make the dataset weekly to ensure the data would be accurate, to account for any day-to-day anomalies. “We do have daily data, but it’s more accurate on a weekly basis. For example, some stores are always closed on specific days,” Tsiounis adds.

The road traffic dataset measures freight traffic on every portion—the distance between each intersection or junction—of every road in the US, totaling 20 million sections of the 8.5 million miles of highway, rural roads, and city streets in the US.

Tsiounis says the dataset will be useful for firms trying to monitor the supply of goods by tracking the movement of trucks between locations. Currently it only monitors the movement of large trucks—which it can discern from individual vehicles based on their movement and the locations where they spend significant time parked, even though all vehicles are tracked the same way, using data from mobile devices—but the vendor plans to extend it to cover smaller delivery vans in the future.

Getting in on the Ground Floor

The new datasets follow the addition of elevation data—i.e., tracking a mobile device’s altitude as well as its map coordinates—which allows Advan to filter out traffic that might be incorrectly associated with a location. For example, in a tall office building, getting accurate numbers of customers for a retail business on the ground floor would require excluding people working in offices on the floors above it—otherwise, it might appear that all those people were visiting the coffee shop, and provide a skewed estimate of revenues for the shop’s parent company.

Tsiounis says that at some point, Advan may choose to map all those devices to specific companies with office space in buildings, by obtaining floor plans and details of who leases each office space, but for now is only focusing on businesses located on a building’s ground floor.

One challenge that makes confirming location across different floors challenging is that altitude data is harder to work with than two-dimensional GPS location data. Tsiounis says that the first time Advan looked at altitude data, it showed half of all cell phones in Manhattan as being 50 meters underground—and not because their owners were in basements or riding the subway, but because different phones’ operating systems report elevation differently, which the vendor must make compatible to produce a standardized and accurate dataset.