The intuitive notion that "you are who you call" is proving true for AT&T, as Labs-Research uses its Community of Interest (COI) methodology to help AT&T Consumer Services (ACS) identify consumers who fail to pay their bills and are shut off, and then try to sign up for new service.

Historically, when a consumer applies for service, their background information (e.g., name, address, Social Security Number) is submitted to a credit bureau and compared against an industry-sponsored third-party database containing people who have defaulted on their phone bills. This database is disproportionately funded by AT&T and a decision was made this year to terminate our relationship with this effort, explains Marget Johnson, a manager in the AT&T Consumer Services. But would the savings come at the expense of an associated increase in uncollectibles?

Pete Coulter, also in AT&T Consumer, challenged Labs-Research to come up with a solution. Daryl Pregibon of the Information and Software Systems Research Lab discussed the problem with Corinna Cortes and Chris Volinsky, also of the Information and Software Systems Research Lab. They had previously used COI technology to identify AT&T Wireless customers who canceled their service contracts and took out new contracts simply to get new phones (the so-called "phantom-churn problem").

Daryl Pregibon	Corinna Cortes

As Corinna explains, the core of the COI methodology is a data structure that approximates the calling circle of a (TN) seen on AT&T's long distance network. The COI calling pattern being searched is typically limited to regular toll calls, and excludes less informative calls, like 800 and 900 service calls, that many customers may make. If the overlap in calling circle exceeds a certain percentage, it is deemed that the old debtor and the new service applicant is likely the same person.

The COIs of customers that default on their bills are saved at the time they are restricted from the AT&T network and stored in a library. The COIs for new customers can now be matched against this library to find repeat offenders. A trial using the Labs-Research COI matching system recently was completed . The trial was extremely successful, with a sustain rate of greater than 90 percent (i.e., over 90 percent of the TNs that were blocked never called in to get reinstated).

Chris Volinsky

Currently CARM restricts thousands of new debtors every day. For each of these, the COI-based matching methodology searches among millions of new AT&T accounts within a 45-day window of the date the debtor was restricted from using the AT&T network. The results seem remarkably consistent, Daryl reports: Approximately one percent of all restricted accounts try to come back to the AT&T network as new customers.

"Using our COI technology, we're finding a much higher rate of matching with repetitive debtors and a lower 'falsing' of incorrect blocks than with the credit bureau approach," Marget says. "So not only are we having greater success, but we're irritating fewer legitimate customers who get tagged for review." Additionally, because there aren't as many accounts being flagged, fewer staff people are required to follow up intercepting calls.

"This new COI approach has two main benefits," Marget states. "First, we will be collecting on past-due accounts that we weren't collecting on before because the person simply changed their name or moved across the country. And second, our net debt exposure gets smaller, since if they didn't pay us before, there's a good likelihood they won't pay us again. It's a real win-win situation."

Back to News
Back to the Front Page