PointPredictive Enhances DealerTrace

SAN DIEGO, Calif., April 13, 2017 – PointPredictive Inc. today announced the general availability of an enhanced version of DealerTrace™, a comprehensive consortium-based analytic solution designed to provide auto lenders with a dealer-level, holistic view of fraud and early payment default risk.  This solution allows most lenders to detect dealer fraud up to six months earlier than currently possible with existing tools.

“Dealer fraud is an industry-level problem that affects all auto lenders,” said Frank McKenna, Chief Fraud Strategist at PointPredictive. “The enhanced DealerTrace service uses the power of the Auto Fraud Consortium to provide all participating lenders with unique cross-industry insights that allow them to identify their riskiest dealers and then take proactive and appropriate action to improve their overall loan portfolio quality.”

Participating DealerTrace lenders contribute their loan-level application information to the Auto Fraud Consortium on a monthly basis and provide information about fraudulent or early payment default loans as it becomes available.  PointPredictive analyzes and combines information on a dealer-by-dealer basis across the consortium and provides each participating DealerTrace lender with a rank-ordered list of its highest risk dealers every month.  A lender may also register for enhanced alerting to be notified when information about one of its dealers is provided by another DealerTrace lender.

DealerTrace leverages the underlying pattern-recognition technology of the fraud and early payment default detection analytics of Auto Fraud Manager to create dealer-centric risk assessments based on lender-reported application and outcome data.  Fuzzy-matching techniques are used to recognize the same dealer – perhaps operating under different names or with slightly different demographic characteristics – across multiple participating lenders.  This allows DealerTrace to form a cross-lender consortium-level view of each dealer’s risk characteristics that is beyond what any individual dealer could do on its own.  It also enables participating lenders to benchmark their book of loans with a dealer against the consortium’s view of that dealer – this is very useful to detect instances where a dealer may be routing its riskier loans to certain lenders.

“We are excited to release this updated and improved version of DealerTrace as the second solution offering for our Auto Fraud Consortium members,” stated Tim Grace, Chief Executive Officer of PointPredictive. “By pooling data at an industry level, PointPredictive helps lenders aggregate their fraud knowledge, identify new fraud patterns more quickly, and collaborate to reduce fraud losses throughout the auto lending lifecycle.”

To learn more about DealerTrace or the Auto Fraud Consortium, contact [email protected].


San Diego Business Journal – Company Uses AI To Stay Ahead of Loan Fraud

San Diego Business Journal Article — Outstanding auto loans in the United States last year topped a record $1 trillion.

San Diego startup PointPredictive Inc. aims to help auto lenders keep more of the revenue they expect to earn from those loans by identifying potentially fraudulent applications before automobiles leave dealers’ lots.

The predictive analytics company, founded in 2014 by CEO Tim Grace, chief fraud strategist Frank McKenna and head of strategic alliances Joe Jackson, sorts through vast amounts of data from lenders to determine which applications have signs of fraud. Using machine learning techniques — a branch of artificial intelligence — the company’s models adapt over time as more data is gathered and analyzed.

Today, the company says it works with five of the top 10 and 17 of the top 30 U.S. auto lenders, as ranked by the number of applications they receive. The companies share historical data with PointPredictive, which uses the information to teach its software to recognize signs of potential misdoings by loan applicants or dealers.

“You hit one spot and they pop up somewhere else. If you cover one gap, they’re going to cheat a new way,” McKenna said. “That’s why they need models that can learn the new patterns.”

Neural Networks

San Diego is a hotbed for predictive analytics talent thanks to the company to which PointPredictive traces its roots. Hecht-Nielsen Neurocomputer Corp., later HNC Software, was launched in 1986 by University of California, San Diego professor Robert Hecht-Nielsen.

HNC developed software that used neural networks — computer systems inspired by the human brain — to help spot fraudulent credit card transactions. Grace and McKenna were among its employees. Another HNC alum, Greg Gancarz, is PointPredictive’s head of analytics and scoring.

In 2002, HNC was snapped up by Fair Isaac Co. (the company behind FICO credit scores) for $810 million.



PointPredictive Launches Auto Fraud Manager

PointPredictive Inc. announced today the launch of Auto Fraud Manager – a real-time predictive pattern recognition scoring solution providing a bundled suite of predictive scores that can instantly assess the risk of First and Early Payment Default, Fraud Misrepresentation and Dealer Risk on each automotive loan application so lenders can stop fraud before underwriting and funding a loan that will likely lead to a loss.

With the launch of Auto Fraud Manager, PointPredictive is addressing the growing problem of auto lending fraud which they currently estimate will reach nearly $6 billion in 2017.    PointPredictive research indicates that most of that fraud risk is hidden in losses categorized as First Payment Default or Early Payment Default – terms that lenders use to indicate loans where the borrowers fail to make the first payment or stop making payments within the first six months of the loan, respectively.

“Our research indicates that as many as 70% of auto loans that default without a single payment have some material misrepresentation in the application,” says Frank McKenna, PointPredictive’s Chief Strategy Officer.  “The overwhelming majority of those defaulted loans with material misrepresentation are submitted by a small fraction of the auto dealers a lender works with.  As few as 3% of a lender’s dealers represent 100% of the risk on some portfolios we have examined.”

Auto Fraud Manager is unique in that it identifies three distinct risks – Fraud Risk, First/Early Payment Default Risk and Dealer Risk – in a single solution so that lenders can make an informed lending decision in real-time while the loan application is being processed and before the car leaves the dealership.  Current solutions for verifying employment or verifying income are often not available in real-time, are not able to provide lenders with instant assessments, and often result in high false positives, poor customer experiences, and costly underwriting times.  Auto Fraud Manager provides an instant (real-time) full application assessment that shows lenders the probable material misrepresentations on loan applications that will likely lead to a loss.  The lender determines their risk tolerance levels and can accept, reject, or order verification solutions depending on their specific loan and dealer risk levels.

As lenders receive applications from a dealer, Auto Fraud Manager scans the application instantly for patterns of fraud risk such as income manipulation, employment fabrication, evidence of straw borrower, collateral inflation, fraud ring activity, and dealer fraud as well as likelihood of first or early payment default.   The lender can use that information to take proactive steps to stop the fraud.   Recent tests indicate that Auto Fraud Manager can reduce risk losses at lenders by 50% or more and improve their dealer performance by more accurately targeting their riskiest loan providers early, before losses are incurred.

Auto Fraud Manager scoring models leverage a consortium approach to learn patterns of fraud and risk across the automotive industry – spanning millions of loans and a variety of lenders.  PointPredictive uses specialized machine learning and pattern recognition algorithms that have been proven successful in fraud and risk solutions for the financial services industry.  Payment cards, mortgage lending, telecommunication, and insurance are already using similar solutions to reduce their fraud losses by 50% or more. For auto lending, we can show that this approach detects the highest percentage of risky loans at the lowest false positives to provide the highest dollar loss reductions and the best customer experience.

“We’re seeing great results from Auto Fraud Manager,” adds Tim Grace, CEO of PointPredictive. “In retrospective tests on lender-provided historical loan data, we have shown that if Auto Fraud Manager had been used at application time, the lenders could have realized 50% reductions in fraud losses at very low false positive rates.  In addition, many lenders are taking advantage of our ability to score large numbers of loans quickly and sending us the last two years of their auto loan applications.  We score these loans and help them identify which loans lead to losses that we would have identified before funding and which loans are classified as credit losses in their portfolio but are actually instances of fraud or misrepresentation.”

For further information on Auto Fraud Manager (including retrospective portfolio scoring or production pilots), contact Kathleen Waid at [email protected]


Auto Fraud Will Reach $6 Billion in 2017

SAN DIEGO, Feb. 02, 2017 (GLOBE NEWSWIRE) — PointPredictive announced today the publication of a white paper detailing new analysis confirming that auto lending fraud risk has been rising for several years, but remains hidden in credit losses. With 2016 auto lending originations soaring to historically high levels, the downstream impacts are now revealing themselves in higher fraud losses. PointPredictive estimates the annual value of auto loan originations that contain some element of misrepresentation may be as high as $6 billion in 2017, which is twice as much as 2016 estimates.

The white paper, based on analysis of historical loan performance across several different portfolios, reveals that auto lending fraud can be broken into three separate categories: known fraud that lenders have been able to identify, hidden fraud that ends up misclassified as early or first payment default, and systemic fraud perpetrated by unscrupulous car dealers or representatives at car dealers.  Each of these categories is a significant contributor to the increasing fraud losses that we are forecasting for the auto lending industry in 2017.

“Our analysis revealed that hidden fraud is the largest category of auto fraud risk as it is often mistakenly categorized with all of the other credit losses,” says Frank McKenna, Chief Strategy Officer of PointPredictive.  “Early Payment Defaults range between 1 percent and 3 percent of originated loans for a typical auto lender.  We are finding that up to 70 percent of those loans that default on the first payment or within the first six months after funding have fraudulent misrepresentation in the original application.  This is a primary contributor to the increase in auto lending fraud risk we are forecasting for 2017.”

Misrepresentation on the application of a borrower’s identity, income, or employment, as well as other key factors such as the price or condition of the vehicle, has a material impact on the performance of a loan.

“Credit scores or credit policies cannot be relied on to identify or prevent these losses, and identity scores only identify a small percentage of these types of misrepresentation losses.  A predictive, full application fraud score is necessary to allow the lender to prevent a significant percentage these losses,” adds Tim Grace, CEO of PointPredictive.

The white paper also provides insight into the role that car dealers play in various ‘systemic fraud’ schemes.  For most lenders, less than 3 percent of dealers are providing loans that are responsible for 100 percent of their known fraud and early payment default risk.  Often, the frauds at these dealers can be traced to a rogue finance manager or other key employee embedded in the finance office that works with fraud rings or identity thieves to facilitate the delivery of fraudulent applications to lenders.  On the positive side, more than 97 percent of car dealers represent very low risk to lenders as they have never been associated with a single instance of fraud.

To receive a copy of the white paper contact Kathleen Waid at [email protected].


PointPredictive Releases Mortgage Fraud Score

PointPredictive Inc. announced today the launch of a new mortgage fraud score that provides an enhanced, real-time assessment of misrepresentation risk allowing lenders to reduce the number of applications currently selected for detailed fraud reviews by investigators.

This score should be requested in parallel with, and used in combination with, credit scores to ensure that the lender’s underwriting policies are informed by the likelihood of misrepresentation early in the loan process.  PointPredictive research confirms that in the absence of a fraud and misrepresentation score, credit scores alone cannot accurately predict repayment of all loans.

Most mortgage lenders have processes that are bogged down by reviewing inaccurate and unnecessary fraud alerts after much of the credit underwriting has been completed.  This slows down the lending approval process and ultimately does not deliver the consumer experience or the loss mitigation lenders want.  The service launched today helps to solve that problem by reducing fraud-related false positives by at least 40% while providing enhanced real-time targeting of misrepresentation risk that would lead to future financial losses due to default and repurchase.

“When we started PointPredictive, our vision was to leverage new predictive modeling techniques to help industries evolve,” says Joe Jackson, Head of Partner Relations.  “Mortgage lending is one of our key markets because while there is so much data available during the lending process, it is not used optimally to identify fraud risk instantly so it can be resolved.”

This new cloud-based, predictive scoring service relies on information readily available at the time of the application and can score more than three thousand (3,000) applications per second.  The mortgage lender receives real-time notification of any potential misrepresentation along with an indication of what types of information will need to be validated with the borrower.

These features allow lenders to insert this service directly in their time-sensitive processing flows, streamlining most loans for faster approval while only flagging the most risky loans for costly and slower fraud prevention efforts.

“Most mortgage applications do not have a high risk of default or repurchase due to applicant misrepresentation and their processing path can be streamlined.  Our new model provides an objective and statistically sound way to help lenders do that,” says Tim Grace, CEO of PointPredictive.   “We use sophisticated machine learning models that provide lenders a “Go / No-Go” decision about misrepresentation review in milliseconds.  With the launch of programs by many leading lenders designed to simplify and expedite the customer’s application processing experience, this new scoring service is now critical for the mortgage industry.”

PointPredictive is offering a proof of concept evaluations to lenders to demonstrate the new model’s effectiveness.  As part of the proof of concept, a lender can score historical applications (with known outcomes) to understand the model’s accuracy in identifying fraud risks that led to a financial loss.  Given the speed of the scoring service, we can review one to two years of a lender’s full application volumes very quickly. PointPredictive will also provide participating lenders with a detailed analysis of the statistical soundness of the model including review rates, detection rates and false positive rates of the model.

If you would like more information about this new predictive service or the proof of concept program, please contact [email protected]


Joel Bock, Highly Accomplished Fraud Data Scientist, Joins PointPredictive

When it comes to fighting fraud with data science, Joel Bock is one of the best.   And this week, Joel made the decision to join our team of Ph.D. educated fraud scientist working in our offices here in beautiful San Diego.

Most recently Joel spent the last several years tackling difficult fraud problems like Tax Return Fraud which skyrocketed to over $8 billion annually starting in 2013.

By leveraging neural network modeling and gradient boosting techniques on hundreds of millions of tax returns, Joel was able to create sophisticated detection models which were able to pick off fraudulent tax returns like a sniper.    With over 250 million tax returns being filed each and every year, that’s the type of thing that can have a big impact real quick.

He must have done something right as tax return fraud is down a whopping 50% in the last two years.

Working With Business

Joel chose PointPredictive because of the unique working environment where business consultants work hand in hand with fraud data scientist to build better models.

Fraud Models that are built as a tight collaboration between business and science always deliver better results with higher fraud detection and lower false positives.


Getting to Work

Joel is hitting the ground running and will be initially applying his fraud fighting skills by boosting our auto lending model performance and helping us fine tune new models that will be announced in the next month.

It’s great to have you on board Joel!


PointPredictive Launches Fraud Model Validation Service for Mortgage

SAN DIEGO, CA–(Marketwired – January 06, 2017) – PointPredictive, Inc. today launched its new Independent Model Validation and Risk Review service to help lenders comply with the Office of the Comptroller of the Currency (OCC) SR 11-7 Guidance on Model Risk Management, a guidance requiring banking organizations to evaluate the soundness of their risk models and monitor models over time to ensure they are performing as intended.

The new service delivers independent validation of current risk models on an annual basis and provides lenders a service to test multiple vendor models side-by-side to determine which model or combination of models works best on their lending portfolio to mitigate losses. The service focuses primarily on model-based Fraud Solutions and Automated Valuation Models (AVMs) that are widely used by lenders today.

“Our experience in modeling is extensive and spans the mortgage, automotive, and banking industries. Our scientists and fraud experts have built sophisticated machine learning models and garnered leading market share in detecting mortgage application fraud, mortgage backed securities fraud, and credit/debit card fraud as well as automotive application fraud, early payment default risk and dealer fraud. This puts us in a unique position to help lenders validate their current models,” says Tim Grace, CEO of PointPredictive. “It can be a daunting task for lenders to validate their models annually, and many do not have the experience in fraud modeling to understand how well a model is working. Now, they can cost-effectively outsource that task to our fraud experts and PhD scientists.”

The service provides lenders with three components:

1. Annual Validation of Mortgage Fraud Models and AVMs — Assist lenders in confirming that the models are appropriately implemented and are being used and performing as intended by using historical data and measuring the statistical soundness of the models.

2. Benchmark Model Testing — Assist lenders during vendor evaluations by analyzing model results, benchmarking those results and providing an independent assessment as to which model(s) or combination of models perform best and are statistically sound.

3. Gap Validation — A business review by PointPredictive fraud and business consultants to validate that the models are being used optimally within their operations area(s).

“The OCC guidance indicates that model validation should be conducted independently by someone not having a vested interest in whether a model is determined to be valid,” explains Joe Jackson, Head of Partner Relations for PointPredictive. “As a proven trusted provider, we are being approached by lenders to do this validation because we are unbiased and our cross-industry visibility enables us to recommend approaches to operationalize the outcomes.”

The Model Validation Service is available immediately to the mortgage industry and inquiries can be made to Kathleen Waid at [email protected].


PointPredictive Finds Fraud in Auto and Mortgage Lending Follow Similar Risk Patterns

PointPredictive, Inc. a leading provider of machine learning fraud solutions, today announced the results of new study that determined high levels of fraud on early payment auto loan defaults. The study found scoring applications for auto loans with models that were built to detect fraud resulted in lenders finding 50% more early default than traditional credit scores.

In 2007, BasePoint Analytics™ found that between 30% and 70% of mortgage loans that defaulted within the first six months contained serious misrepresentations on the original application. These misrepresentations on borrowers income, employment, collateral or even intent to occupy had a material impact on the performance of the loan but were often considered “hidden fraud” since they were never detected in the application process.

“Our analysis and experience suggest that many auto loans that default within the first six months have fraud misrepresentation on the loan application, indicates Tim Grace, CEO at PointPredictive, “when we ran our fraud pattern recognition models on the application information provided on loans that defaulted early, the models were finding strong evidence of fraud. This is the same type of behavior we saw in mortgage prior to the mortgage meltdown”

PointPredictive Auto Fraud Models analyze each application and alert lenders when it appears that there might be misrepresentation on the application related to the income, employment, collateral, borrower or dealer. While built to detect fraud, scientist were surprised to find that it did extraordinarily well in the detection of early payment default (auto loans that default within the first 6 months). In lender tests covering 1.4 million applications and 22,500 deals, the PointPredictive Fraud Score was able to detect 14x more of the total fraud experienced by the lenders than previously used tools and processes. The early payment default (EPD) score identified 4x more first pay defaults and the Dealer Score identified 2x more suspicious dealers.

Auto lending fraud, like mortgage fraud occurs when information on an auto loan application is intentionally misrepresented either by the borrower themselves, a sophisticated fraud ring, or in some cases an unscrupulous dealer. When information is manipulated and the lender does not know about it

they may underwrite the application assuming the information is valid. Intentional fraud presents a problem to auto lenders since loans that have misrepresentation are more likely to result in – early payment default – a term lenders use to indicate when no payments are ever made on the loan.

PointPredictive Auto Fraud Manager uses pattern recognition; a technique that scientist have perfected to detect fraud based on historical data mining. The solution works by analyzing historical patterns of fraud, early payment default and risky dealer activity and then scores each application as it comes in from a dealer. Lenders are automatically alerted when an individual application has a significant number of application anomalies or fraud patterns. The lender can review the application and take action before it is approved. Over time, if a particular dealer submits many applications with similar fraud patterns, the solution will alert them to that as well so they can take the appropriate action.

As part of the study, PointPredictive has published a whitepaper on the subject titled, “ You Can’t Fight Fraud with Credit Risk Tools”. The whitepaper is available by emailing [email protected]


Making it Smart and Making Customers Happy in the Process


Having just completed another successful project in record time we were thrilled when our client described our work, exactly as we would have hoped they would when we started the project. – “Point Predictive more than met our expectations..”,  “their team was available, responsive and really did an exceptional job of bringing their significant expertise to bear on our business”.  These kind words were literally music to our ears and that’s why we started PointPredictive.

Our business model has always been to partner with our customers. We bring them our expertise in data, risk and modeling and help them solve business problems quickly in a way that they can understand and use.  Through our most recent project we did just that by focusing on the business issues and working on a tactical and practical solution using smarter science.

There were many things that made this engagement successful but I felt there were 5 reasons that made it really work.

1. We started by understanding their business before we gathered the data.

There is intelligence you can derive from data but it first requires a deep understanding of the business issues to put that data into context.  That’s why we spent an entire week, working hand in hand with our customers onsite with them.  We needed to understand their business problems that they faced everyday before we could help them solve them.  After the week was over, we were able to gather years of historic data and make sense of it when we got back to the PointPredictive lab to get to work.

 2. We didn’t lock ourselves into a single analytic technique.

We realized pretty quickly on that to solve this particular issue, we were going to need to be creative and think outside of the box.  So, we decided very early on to be analytic-agnostic.  That is to say we would use any analytic technique that delivered the best predictive value while being simple enough to put into production quickly and be useable by the business.  We tried scorecards, logistic regression and even experimented with deep learning techniques.  In the end we found that a combination of techniques worked the best.


3.  We scheduled regular meetings to discuss progress and raise issues.

We are big fans of iterating a solution until we get it right.  So we scheduled weekly calls with our customers to report on results of the analytics and to get their feedback through the entire process as quickly as possible.  We didn’t develop the model in a vacuum, we created the model in an iterative process hand in hand with our customer.

4. We measured model performance for them in many different ways and focused on performance.

We took great effort to deliver a full set of performance charts for our customer.  Not only did we measure the models ability to detect risk overall, but as well its ability to reduce financial risk over time.   A model that does not help reduce financial risk has a very hard time of adoption since their is no financial incentive.  We made sure that our performance charts clearly demonstrated the benefit to improve adoption of the model.

5.  We helped to operationalize the model.

The model was extremely predictive.  We were able to show that with our performance charts. That only solves half the problem however and we needed to help our customers understand how to use that model in various aspects of their business.  That’s were our business consulting backgrounds really came into play.  Having worked with over 100 lenders, banks and financial companies on analytics and risk we were able to bring our significant expertise to bear.  Operationalizing the model helped bring it to life in a way that made sense for the business to use.

It’s so great to be able to help our customers and enable them to achieve their business goals.  That makes all the hard work we do everyday worth the effort.

For questions, please contact [email protected]