Using data to fine-tune collections
WSECU builds model to score loans, determine best course of action.
Most members know when their loans are past due and have a plan to pay. So why frustrate them with unnecessary collection calls?
That’s the idea behind WSECU’s model that scores delinquent loans to prioritize collection outreach, says Mark Baumgartner, director of data science for the $4.4 billion asset credit union in Olympia, Wash.
He addressed the CUNA Lending Council’s People, Process and Technology Virtual Series, noting how WSECU’s model examines loans that are past due beyond a certain point and scores them based on how likely they are to reach a further level of delinquency.
“We use a mix of information about the loan, the member, and their relationship with the credit union to get these scores,” Baumgartner says. “Every morning we generate the files, they get shipped over to our collections system, and, based on that score and the type of loan we’re dealing with, we assign their collections treatment.”
The process stemmed from the desire to know the likelihood that loans will go a certain number of days past due. Addressing this issue limits the frustrations of members who don’t need collection calls and saves money.
In the first year using their model, WSECU saved about $20,000 per month “without any noticeable difference in delinquencies or collections efforts,” Baumgartner says.
“The vast majority of delinquent loans don’t reach 30 days past due, so most delinquent loans just need a nudge or no contact at all,” he says.
Previously, WSECU didn’t segment delinquent loans and referred them to a third party for collection.
“They charge for that service, so every account sent there costs us money,” Baumgartner says. “And any time you have a third party calling a member who doesn’t need it, that’s likely to be a negative experience. We want to keep that number as small as possible.”
To determine which past-due loans need contact, WSECU considers several factors: credit data; relationship data such as deposit history, transaction history, and prior past-due loans; and loan data, including the original term and how many payments remain.
WSECU takes this information and uses several models simultaneously to score each past-due loan and determine if the member needs a follow-up call from staff, an automated call, or no follow up.
“You want to foster loyalty,” Baumgartner says, “not subtract from it.”