KXEN Analytics Key To US Mortgage Information Provider

Part of the country's largest property information provider and its most trusted source of mortgage data, LoanPerformance owes its dominant market position to the increasingly sophisticated information it offers, and the range of tools it has built to help users exploit that information. This capability has now taken another significant step forward with the company's adoption of the KXEN Analytic Framework.

LoanPerformance uses KXEN to create analytical models for determining prepayment and collateral risk - both serious concerns for mortgage holders and investors. Returns from mortgages depend on making these determinations accurately and quickly.

Multiple Criteria
The predictive models that LoanPerformance supplies to its customers have to work with multiple variables including interest rates, property information and demographics together with numerous mortgage product specifics. The sheer number of variables, combined with target mortgage portfolios ranging from a few hundred loans to several million, means a large number of different models is required.

Underlining this is LoanPerformance’s mortgage prepayment scoring system, PreTell. This will ultimately require more than 75 different predictive models covering a wide variety of mortgage products. In addition the system also produces predictive scores for different types of prepayment schemes including home purchase, refinance and cash-out refinance.

To develop, deploy and maintain so many models, and to meet the standards of accuracy and consistency required by its clients, LoanPerformance knew that it would take a very special kind of analytics solution to do the job.

Previous experience of KXEN's Analytic Framework, together with confidence in its vendor and in its underlying methodology, suggested that KXEN's solution would be ideal for this new application too. Key strengths were KXEN's robust modeling methodology, the accuracy and consistency of its models, its tools for rapid model development and deployment, and its component architecture that enables outputs to be easily integrated into various software platforms.

Other pluses for KXEN were the way it encodes and transforms input data, and its highly scalable regression engine. With all these factors in mind the prepayment and collateral risk modeling teams at LoanPerformance had no hesitation in selecting the product.

"The KXEN Analytic Framework provides LoanPerformance with a very robust estimation methodology that also facilitates rapid model development and testing," said the company's director of collateral risk analytics, Damien Weldon. "The predictive models generated are of high quality, as proven by a number of different client studies, and have been a key component of the success of our analytical products in these risk segments."

Dual Role
At LoanPerformance the KXEN solution works in two ways. First it produces model reports and charts which the company uses for its own internal analysis purposes. More important though it is used to create predictive models for LoanPerformance clients.

These are delivered as scoring engines on a range of platforms and deployed in several different languages including program code (C, Java, etc.), predictive modeling markup language (PMML), SQL, and SAS code.

LoanPerformance also has good things to say about KXEN's support. "If needed, the KXEN support team is prompt about answering questions about technical problems and new features," said Weldon. "The best part of working with KXEN - other than the product itself, the results we achieve and the benefit to our clients - is the outstanding KXEN team, key members of which we have known for many years."
www.kxen.com/cxo