Sears, Roebuck & Co. recently implemented software from SignalTree Solutions to automate credit decisioning for customers applying for its Sears-branded Mastercard.
Lending is a core competency for Sears, which has annual revenues exceeding $40 billion. "The credit division within Sears, even though it's a small part of their revenue, is a very profitable division for them," said Dan Metsa, VP and GM of SignalTree Solutions in Chicago. "They have a pretty sophisticated credit analytics group."
Although Sears might have partnered with a bank to offer credit products to its customers, the company chose instead to draw upon its internal resources. "One of the reasons they didn't want to partner with a bank is that they're competing with a lot of banks that might be offering their cards under their own names," said Metsa.
The solution developed by SignalTree allows credit analysts, rather than IT staff, to establish rules for determining whether to extend credit to a particular customer and at what level. These rules use information from credit applications and from credit bureaus to decide a score for a customer on a 100-point scale.
While the current release of the software bases scoring decisions on a set of fixed rules established by the credit analytics group, the second phase of development will allow for more flexible scoring criteria. Metsa compares the new functionality to Excel formulas that change dynamically based on values in a spreadsheet; similarly, threshholds for granting credit will automatically adjust over time as more information becomes available about actual performance of the credit portfolio. "That's part of what Sears feels is unique in the industry, because a lot of products out there didn't offer that kind of flexibility," said Metsa.
The software resides on a UNIX machine that interacts with Sears' mainframe system used by its 860 full-line department stores and 2,100 specialized retail locations. "It's a centralized model," said Metsa. "All of the credit applications come through the central unit to be processed, so we didn't have to distribute it to all the stores."
As such, there were strict limits on how much processing time the analytics software could take. "With all the other steps that the credit application had to go through before it go to our system, we had a three to five second window," said Metsa. "But a lot of that time was already taken up by the interaction to the credit bureau or the actual capture of the application from the store."
"We were given 50 milliseconds to come up with a score," he said. "The software we produced actually does it in about 20 milliseconds."