Tracking the mobile Web

Omniture uses DotMobi’s DeviceAtlas index to identify the make and model of the phone used to access one of its customer’s Web pages and then attempts to uniquely identify the owner of that phone through a combination of technologies. While that’s a fairly common practice on the Web, where customers do most of their surfing from a single or relatively small number of PCs, it’s a much more difficult proposition over a wireless network.

Determining one device—and its owner—from another is still an inexact science, said Andy Bovingdon, vice president of marketing for Bango, a third-party mobile content distributor that developed a mobile analytics business for the use of its customers.
Most devices do not identify IP addresses of their own, relying on instead on a proxy server in the network. The end result is that multiple users can be sharing the same IP address at any given moment and an individual user could be accessing the same Website from multiple IP addresses throughout a day.

Alternating IP addresses are fairly common in the wired Web world also, but the problem is usually solved by embedding cookies in a Web browser. Many mobile browsers, however, do not accept cookies, making that form of tracking another dead end. Many of the mobile analytics companies have resorted to Web beacons, a tiny image the size of a pixel embedded in the browser that uniquely identifies a device. Even still, not every device accepts such beacons. Bango sometimes has to resort to identifying device by a profile, which requires detecting the unique settings on a particular device and associating it with a particular carrier and geography, Bovingdon said. This is akin to identifying a particular car by its color, accessories and the town it is located in rather than a unique identifier such as a license plate. Consequently Bovingdon said that kind of profiling only works with the least popular devices and for customers of smaller carriers.

Despite those obstacles, Bango has developed unique identifiers for 25 million different users, Bovingdon said, and can peg 99% of the devices traversing its customers’ sites as unique users. Once those identifiers are in place, Bango can use them to develop specific customer profiles. Some of that information is collected by tracking a customer across the sites that Bango powers, other information is collected through random surveys and through registration on specific sites. But Bango can tell a lot about a customer just by what carrier he or she uses.

“One of things we pride ourselves on is the ability to identify the MVNOs within a parent carrier,” Bovingdon said. “Sprint has 25 different MVNOs, which tell you a lot about the user. Other people can tell you it’s the Sprint network. We can tell you if it's a Boost customer on the Sprint network.”

From that data, Bango extrapolates not just the number of unique visitors that come to a site—which is critical in determining advertising costs and placement—but demographic and interest information that can be used to target specific ads or marketing campaigns.