Ai+Ni: Powering the way for top-notch fraud detection and prevention

The past several decades have dramatically transformed how we do things, from taking pictures on a smartphone versus a camera to streaming movies instead of popping a DVD in a player. So it’s only natural to expect significant changes in how fraud is perpetrated—and the most effective ways to fight it. 

Today, fraud is powered by technological innovation—and just like everything else around us, it’s evolving at a rapid pace. At Point Predictive, we’re trailblazing the way for lenders to effectively respond with cutting-edge machine learning models coupled with human intelligence. 

Point Predictive’s revolutionary approach to detecting and combatting fraud

Before extending credit to a borrower, lenders typically ask applicants to complete extensive paperwork requiring key information, such as their birthdate, social security number, home address, desired loan amount, interest rate, loan terms, etc.  

Lenders typically have hundreds of thousands of these records on file but won’t—and generally cannot— share them with other lenders due to privacy reasons and competing interests. This means each lender only maintains access to historical data for consumers whom they’ve previously attempted to serve.

That’s where Point Predictive’s proprietary data repository – currently containing over 130 million application records from 67 million unique consumers – is a game changer.   

How our fraud analysts use the proprietary data repository

Leveraging our proprietary data repository, our fraud analysts investigate high-risk applications to alert lenders about components of borrower-supplied information to which they should apply more scrutiny. The analysts also identify patterns that are used on a forward-looking basis to inform the foundational components of our risk detection models. 

These models then help predict the chances of future applications containing fraudulent information that leads to early payment default.

How natural intelligence (NI) factors in

Lenders receive many thousands of applications a month—and Point Predictive has a finite number of fraud analysts. These experts prioritize which applications receive a deeper review through our scoring mechanism. The higher the score our model generates, the higher the likelihood of fraud or fraud-related default. The applications that receive the highest scores are prioritized for review by our analysts, supplementing the automated risk scoring with human intelligence, or what we refer to as natural intelligence (NI). This hands-on analysis by our subject matter experts complements our artificial intelligence (AI) capabilities. 

How artificial intelligence (AI) factors in

Point Predictive’s solutions use machine learning and artificial intelligence to determine how risk attributes influence the overall likelihood that an applicant is committing fraud that is likely to lead to default on a loan.  Our binary classification models allow each incoming application to be designated as “fraud likely” or “fraud unlikely.”

Our proprietary data repository also compares the self-reported application information against past information the applicant has submitted on other loan applications. For instance, if an applicant claims to be a software engineer making $200,000, we can match them by name, SSN, birthdate, etc., to identify discrepancies in employment history or stated income. A significant difference will trigger an alert to the lender.

Benefits of using AI over traditional methods:

• Better statistical results: compared to manual methods, AI uses advanced and precise measurements (such as Area Under the Curve) to enable us to measure how effective Point Predictive is in solving binary classification for fraud detection. 
• Superior to human analysis at scale: can immediately detect nuanced patterns likely to be missed by human analysis and can synthesize a larger number of fields, data points, and volume of information at faster speed.  
• Capacity for rapid change: AI processes, recognizes and commits to memory evolving patterns and information and can easily scrap previous knowledge to relearn and adapt at speed to emerging information.

Benefits of Point Predictive’s proprietary data repository:

• A large volume of quality data: our proprietary data repository is the largest and most differentiated in the industry.
• Expert technology analysts and data scientists: our technology professionals are constantly analyzing and driving improvements to our solutions, and our data scientists know how to most effectively apply our data to prevent high-risk loans.
• Industry-best fraud analysts: our fraud analysts work with the rest of our experts to deliver an industry-revolutionizing fraud solution for lenders. 

An example of how Ai+Ni helps in the battle against fraud

A customer was successfully using our scores to detect fraud when one of our fraud analysts noticed that a certain segment of their portfolio was exposing them to higher risk. The fraud analyst worked in conjunction with another analyst to generate eight highly targeted alerts to pinpoint the problematic segment of the population for the client without spiking review rates and while minimizing false positives. Once the alerts were implemented, they then applied an accelerator to the model features that drove those alerts. As a result, the client was able to quickly and accurately isolate those applications and manage the specific risk types accordingly. 

Point Predictive solutions that use our proprietary data repository include: 

• Auto Fraud Manager: a solution that equips auto lenders with real-time access to data, scores, and alerts that lenders to identify first- and third-party fraud and optimize their origination and decision strategies.  
• BorrowerCheck^TM: generates a report that guides a dealer on whether they should gather more information on a prospective borrower’s income, employment, or identity prior to submitting their loan application to lenders. If additional scrutiny is warranted, the report also provides tips for performing additional verification.
• DealerCheck^TM: Shows lenders the quantity and quality of deals coming from a dealer compared to that dealer’s activity with other lenders, also indicating any changes in a dealer’s risk level.
• EmployerCheck^TM: a solution that uses AI to predict the legitimacy of a borrower’s employer and source of income.
• IncomePass^TM: an AI-powered alternative to paystubs, IncomePass enables lenders to easily validate income for borrowers. 

Data security measures 

We understand data security is integral in today’s world. All data transmitted to and from Point Predictive’s secure, virtual production cloud environments at Amazon Web Services (AWS) uses Transport Level Security that provides end-to-end encryption.  Data is decrypted in memory and processed to generate results.  

All data retained by Point Predictive is encrypted at rest using AES-256 compliant methods.  Point Predictive conducts continuous sweeps of its AWS environments to monitor for security vulnerabilities. A third-party penetration and vulnerability test is conducted annually and a SOC2 Type 2 report is also prepared annually by an independent auditor.

We’ve made great headway into the fight against fraud, but our work is far from over. To learn more about Point Predictive’s proprietary data repository or explore the right mix of solutions for your lending needs, request a free demo.