Scoring Solutions™ Models Alternative Data to Help Lenders Approve More Loans

Scoring Solutions recently developed a case study with LexisNexis regarding underbanked/unbanked, subprime and borderline risky—consumers who can represent growth opportunities for many lenders. Lenders aim to make accurate decisions informed by greater analytic insight into loan applicant creditworthiness. The objective is not to just make more deals, but to make more deals with more precise estimates of risk.

While many lenders share this aim, the challenges to achieving it may be quite different even within the same market. Two Scoring Solutions clients in the auto lending industry illustrate this variation:

• A subprime lender wanted to improve its ability to serve populations with poor credit histories while more accurately assessing and pricing for risk. They needed an analytic model that would help them differentiate relative risk levels within these populations.

• A captive auto finance company lending mostly to consumers with good credit histories wanted to further improve portfolio performance while reducing delinquencies and losses during economic downturns. Its managers were sometimes overly cautious, turning away some applicants who probably would have been profitable.

Both of these lenders wanted to more accurately assess the creditworthiness of so-called underbanked and unbanked populations. These are individuals with “no file” or “thin file” traditional credit histories. There are many creditworthy consumers within these populations, but traditional data sources provide little or no data to identify them. Contact Scoring Solutions at score.info@scoringsolutions.com or 678-672-5450, ext. 362 for the entire article which provides the solutions provided to each of these lenders.

Scoring Solutions Evaluates Bootstrap Aggregating Modeling Methodology
Periodically Scoring Solutions researches alternative model development techniques in the interest of improving upon the models we develop. Scoring Solutions has evaluated the performance of bootstrap aggregating (aka bagging) technique used in conjunction with logistic regression models.  Specifically, acquisition risk models were developed using our well-established methodology and also employing the bagging approach. We examined the predictive lift and found that, given a sufficient sample size, both methodologies provided comparable results. We believe that Scoring Solutions continues to provide our customers with one of the most effective model development methodologies.  Scoring Solutions is committed to exploring new approaches to improve our models and the corresponding results of our customers.
Partnership with Analyse Professional Services Continues to Grow as
Scoring Solutions Provides Score360® and Probability of Default Validation Study for Mexican Marketplace

A captive auto finance company chose Scoring Solutions to implement their scoring models and strategies using Score360, a hosted scoring and decision engine solution for rapid implementation and deployment of attributes, scorecards, and strategies.  As part of this service, Scoring Solutions will receive account information on a monthly basis and return attributes, scores, and decision strategies to the company within one business day.  In addition, monthly reporting and consultation will be provided to ensure proper use and performance of the system.  Initial strategies will be programmed by Scoring Solutions, but company personnel will be trained to use the decision engine so that they can revise and devise strategies on-going and upload them to Scoring Solutions for execution each month.

Scoring Solutions recently performed a validation study for a client’s Probability of Default (“PD”) models.  The analysis included the testing of several portfolios along with an assessment of economic variables currently used in the models. PD models are important contributors to the calculation of capital reserve requirements for banks. The validation study compared the predicted default rates with the historically observed default rates – at the overall model level and at the score interval level within a PD model. It is critical that organizations using PD models validate these models regularly and re-calibrate odds when needed.

Visit www.scoringsolutions.com for white papers published by Scoring Solutions.
To learn more about how we can bring value to your organization, call us at 678-672-5450 ext. 362 or e-mail us at score.info@scoringsolutions.com.


Connect With Us:     FacebookFacebook   LinkedIn LinkedIn   Twitter   Twitter