Follow the Smart Phones

Mining cell phone data could help target ads and rank local services.

Apr 14, 2010

A service launched last week by Skyhook Wireless will make it possible for other businesses to predict, with new accuracy, which local bars will be hot at 8 p.m. on Monday night, or how many people will walk past a particular billboard poster at noon on Friday.

What’s hot: SpotRank generates “heat maps” showing the density of cell-phone users for a given time and place. This image shows the southwest corner of Manhattan’s Central Park on Monday, March 29, at 6 p.m..

Skyhook Wireless’s pool of anonymized location data, gathered from cell phones that have used its services over the past 24 months, shows user behavior in every major city in North America, for every hour of every day of the week at a resolution of 100 meters. This is enabled by the 300 million check-ins received daily from every iPhone, iPad, Snow Leopard-powered laptop, as well as Dell devices and a growing number of Android-powered smart phones.

Several other companies are using similar technologies to map human activity across time and space–an activity first referred to as “reality mining.” However, no other company has made available a comparable amount of data to independent developers.

Skyhook Wireless’s new service, called SpotRank, is available to developers through an application programming interface (API) from its partner SimpleGEO–a cloud-based service for managing large quantities of geolocation data. The data resembles a heat map of population density in a given city at any point in time. The data can be strung into time sequences to show the changes in human activity as a city cycles through the workday, the commute home, and nightlife.

Internally, Skyhook Wireless has begun developing applications for SpotRank data–including new ways to inform buyers of outdoor advertising. “We can tell [advertisers] where the best spot in Manhattan is to put a sign on a side of a building,” says Skyhook CEO Ted Morgan.

“It’s very valuable data,” says David Fono, a developer at Atmosphere Industries who has begun working with the SpotRank API for games that play out across a city. “The level of data they have is staggering,” Fono says. “This is a pretty significant addition to the tool set. Just the fact that it’s a fairly reliable metric for human traffic in an area, I don’t know of anything else like that at the moment.”

There is growing interest among technology companies in mining the physical movements of users, but privacy promises to be a hot-button issue. “We are keen to do something similar [to SpotRank], but we want to make sure we maintain user privacy,” says Sharon Biggar, chief operating officer of U.K.-based Path Intelligence, which uses passive receivers to track the cell-phone traffic of shoppers and concertgoers as they visit public places. Path Intelligence can determine the location of a device to within a meter or two, and individuals can be tracked continuously as they move through an area, allowing engineers to tell business customers which stores in a mall tend to be visited together, for example. In contrast, SpotRank data only shows an aggregate number of people in any one area at any given time.

Sense Networks, a company cofounded by Sandy Pentland, a professor of computer science at MIT, also tracks population density in cities. It offers this data to companies through an application called MacroSense. An application offered by the company, called CitySense, shows data similar to that offered by SpotRank, but only for the city of San Francisco. This summer, Sense Networks also plans to offer developers an API. Sense Networks says its offering will be even more detailed, and will combine demographic data with geospatial tracking data.

A few developers are beginning to explore the potential applications of the kind of location data provided by SpotRank. Utkarsh Shrivastava, a master’s student in information security at the Georgia Institute of Technology, has combined SpotRank data with Yahoo’s database of companies to create a search tool that ranks local businesses according to how busy they are at any time. “It could also expand to a route based model,” says Shrivastava. This would allow users to choose the least-crowded means of transport between two points–an application feature that has already been implemented on roadways by some GPS devices.

Joe Stump, cofounder of SimpleGEO, says developers are working on a wide array of ideas based on using SpotRank’s data. “There are lots of ideas around potential variable pricing for advertising,” he says. The data could also be used to make location-based social networks like FourSquare more interesting and useful, Stump says.

The next step, say developers working with SpotRank, is real-time data. This would let them check the “social weather” in a town to determine, say, if there’s a particularly popular event going on.

“You could see where activity is almost in real time and overlay geo-tagged images and tweets and watch as it happens,” says Skyhook’s Morgan. “We’re also building out a social compass, which looks like a regular compass, but it can direct you based on the volume of activity [in an area].”

However, even SpotRank’s existing offering, which is strictly historical data, is useful because people are predictable. “We see that in our data, which is 24 months worth of history, we can tell you with 90 percent confidence what will happen at 2:15 on a Tuesday,” says Morgan. “Ninety-nine percent of the time we know what will happen on any given spot. We can do that for almost every street corner in the world right now.”