ScoreText Captures 'Unstructured' Data
LoanPerformance here has teamed up with Intelligent Results, Bellevue, Wash., to create ScoreText, a software application that integrates structured data with "unstructured data" that might be gleaned from a lender's portfolio.
Harnessing unstructured data, often gleaned from customer service phone calls, will help lenders better understand and more accurately predict customer behavior related to servicing and collection efforts.
Historically, servicers have identified and prioritized potential problem loans using "structured numeric data," the companies said. Until now, they have not been able to systematically access information and observations generated by other forms of interaction with the borrower.
ScoreText allows mixed-data analytics. This combines free-form text from lenders' and collectors' notes, call centers, customer relationship management systems and borrower e-mails, with structured data from traditional sources such as credit scores, payment histories, loan balances, customer demographics, property record information, collections systems and account master files.
"Today's best predictive models only incorporate about 20% of the available data," said Richard Harmon, senior vice president, scoring analytics and services at LoanPerformance, in a statement released by the companies. "By exploiting the available unstructured data that makes up the other 80%, mortgage servicers can not only improve their ability to predict customer behavior, but can have a much better understanding of the key factors differentiating behavior. We believe many servicers will be able to improve their current loss mitigation performance by as much as 50%."
Extracting textual data into predictive loss mitigation models give servicers a more powerful tool to determine which loans will cure themselves and which should be handled using forbearance, short sales, foreclosure and other methods, he said.
The ScoreText solution was jointly developed by LoanPerformance, which currently tracks the monthly prepayment and delinquency performance of 46 million loans, and Intelligent Results, a company that has already successfully utilized a mixed-data approach to improve the servicing and collection functions in other consumer lending sectors such as credit cards and auto loans.
The companies say that the power of mixed-data analytics lies in its ability to identify the relationship between seemingly innocuous words like "motorcycle" and "roommate" to delinquencies and defaults. Mixed-data analytics involves a four-step process where text is cleaned, categorized, extracted and then modeled.
In addition to being used for collections and loss mitigation, the analytics can be used in customer-complaint management, customer retention, cross-sell and other functions, the companies said.
Copyright 2005 Thomson Media Inc. All Rights Reserved. http://www.thomsonmedia.com http://www.mortgageservicingnews.com