PointPredictive leverages machine learning for income verification tool


Through advancements in machine learning, auto finance companies now have another tool to answer one of the most important questions during the underwriting process.

PointPredictive recently released its newest product — Income Validation Alert — a solution that can offer a real-time predictive assessment of an applicant’s stated income. If that income appears to be overstated by 15 percent or more, finance companies are notified so they can perform additional verification.

For applicant incomes below the 15-percent threshold, the finance company can streamline the underwriting process to ensure good contracts are not impacted.

With sub-second response times and high accuracy rates seen in the laboratory, PointPredictive expects the solution to have wide-reaching market usability, including: 

—Pre-screening online mortgage and automotive applications

—Streamlining automotive application underwriting stipulations at dealers

—Enabling faster credit decisions for consumers

—Reducing the finance company cost of utilizing income-verification solutions that might be more time consuming, less accurate and more expensive

“Validating an applicant’s stated income in real-time has been an issue for lenders for many years. Traditional tools haven’t really addressed this issue well,” said Tim Grace, chief executive officer of PointPredictive.

“Employer-based database verification checks can typically verify income on only 30 percent of the applicant submissions, leaving 70 percent of stated incomes to be checked by manual processes,” Grace continued. “Verifying income directly with the borrower by requesting paystubs is often unreliable — some lenders report more than 20 percent of paystubs received are forged or altered.

“Lenders are looking for something accurate, fast and cost effective. Income Validation Alert should help immensely in this effort,” Grace went on to say.

Income Validation Alert can analyze a borrower’s stated income against millions of reported incomes and salaries from seven diverse sources. Then, using the applicant’s employer, occupation, job title, residence and estimated years of experience, a sophisticated machine-learning model can predict the borrower’s likely income.

When the borrower’s stated income exceeds what the model predicts by 15 percent or more, the finance company is alerted to the discrepancy and can further scrutinize the borrower’s income.

PointPredictive highlighted that what makes Income Validation Alert different than other approaches is the breadth and depth of salaries and reported incomes available to the validation process. Comparing a borrower’s stated income against seven different sources simultaneously and then using machine learning to cascade through those sources to determine the most reliable income for that borrower is a groundbreaking achievement.

The company added that early results illustrate just how powerful this approach is.

“In laboratory testing with live auto lending data, we were able to successfully clear the stated income data on more than 85 percent of applications while successfully identifying more than 80 percent of the applications with known, overstated incomes,” said Greg Gancarz, data scientist with PointPredictive.

“This level of detection is clearly a breakthrough in helping lenders identify one of their biggest pain points — knowing which applications require detailed income verification,” Gancarz went on to say.

Income Validation Alert is available to banks, finance companies and card issuers to verify stated income on applications. The service is available today for real-time integration into loan origination and underwriting workflows.

For more information on Income Validation Alert from PointPredictive, send a message to info@pointpredictive.com.