Predictive Analytics Gaining Credence with Servicers

Similar to credit card and auto lending operations, mortgage lenders are now incorporating predictive analytics to better manage risk, said James Caldwell, a partner with Deloitte & Touche, at the MBA's National Mortgage Servicing Conference & Expo here.

"There are several stages of increasing sophistication in collections, ranging from traditional to enhanced and then to risk-based," Mr. Caldwell told conference attendees during a session entitled "Leveraging Predictive Analytics to Drive Profit in Mortgage Collections."

Deloitte conducted an online survey with the assistance of the Mortgage Bankers Association in which it conducted in-depth interviews with senior executives and examined collections operations of more than 20 mortgage companies, including six of the 10 largest services with more than $600 billion held in combined portfolios.

With rising rates, concerns over housing prices and strained economic outlook, the survey found that lenders are implementing a variety of techniques to lower costs and boost revenue. "Mortgage lending has always looked for trends in customer data. However, sophisticated analytic tools can recognize critical patterns beyond trends. The market for decision analytics software in the U.S. was $2.3 billion in 2004 and is expected to grow by 30% in the next two years."

Trends driving mortgage collections practices include concerns about credit quality and the rising rate in bankruptcies. Mortgage debt has increased 42% since 2001 to $8.8 trillion and home equity at $715 billion in 2004, according to the survey. MBA data show that interest-only and adjustable-rate mortgages made up 65% of new mortgages in 2004, up from 18% in 2003. "Industry trends will drive more lenders to adopt predictive techniques and to re-examine their collections organization. Lenders are seeing new needs and new opportunity for those in the subprime market, which accounted for 20% of all new mortgages between 2004-2005," said Mr. Caldwell. "And the impact of bankruptcy law changes in 2005 has not realized. Personal bankruptcies have doubled every decade since 1960."

Major lenders have been taking steps to improve collections by streamlining the workflow, measuring performance more accurately and building the technology infrastructure, he said. One study participant used predictive analytics to target loans for foreclosure using LTV and the number of months the account was delinquent. "By foreclosing on accounts four months delinquent and with an LTV greater than 50%, they refer 18% more homes for foreclosure and decreased charge-offs by 10%," Mr. Coldwell told the audience.

More than 80% of the top 25 mortgage lenders deploy automated collections workflow technology, according to the survey, yet only 57% of respondents said their operation has automated as much of the treatment of customers as the technology allowed.

Some additional techniques include automated account queuing and automated correspondence. Lenders route accounts to collectors for appropriate treatment considering factors like high-balance accounts, preferred customers and customers with multiple accounts in collections. Fully automated correspondence employs rules based upon risk level and strategy for the account as well as well as compliance with federal and state regulations. Real-time data on account aging provides the ability to view both dollars past due in each aging bucket and the frequency each customer has been past due over the last 12 months.

"Improving collects performance requires accurate metrics, yet the cost to collect $1 has not been widely adopted," Mr. Caldwell said. "Only 29% of respondents said they used this measure."

The leading collections practices depend on robust technology, the survey acknowledged, noting that decision engines, data mining tools and workflow management tools help facilitate the process. A portfolio management dashboard provides detailed, flexible reporting on portfolio performance with early warning indicators of risk exposure.

"Most collections operations do not fall into a single stage but combine capabilities at different levels of sophistication," he said. The enhanced collections operations had made significant strides to improve efficiency beyond the basic levels. This stage includes behavioral scorecards. Collectors are assigned accounts by risk level, and targeted call campaigns are designed for predictive dialers.

In the end, many lenders are deciding to combine predictive techniques with data warehouse and mining systems, with sophisticated behavioral scoring, profitability metrics, an enterprise-wide collections utility and seamless vendor management in order to see results in lower operating costs, higher rate of promises kept and more dollars at risk collected.

SNAPSHOT: Fast Growth for Servicers

Mortgage Debt Outstanding in '04 $8.8 Trill.

MDO Growth Since 2001 42%

Home Equity Debt in '04 $715 Bill.


(c) 2006 Mortgage Servicing News and SourceMedia, Inc. All Rights Reserved. http://www.mortgageservicingnews.com http://www.sourcemedia.com