The Next Step in FICO Credit Modeling

Since its introduction by Fair, Isaac & Co. more than three decades ago, FICO has long been considered the No. 1 standalone credit decisioning model. Is there a way the banking industry can build on that foundation to create a more future-forward way of predicting lending outcomes?

The way the financial services industry analyzes data has evolved since FICO’s inception in 1989. New and exciting technology has led to innovative algorithms that give bankers a more defined look at an even greater data set. An all-encompassing view of a borrower’s story can bring a new realization: these new methods of analyzing credit, combined with the FICO mainstay, can lead to even better outcomes for everyone.  

Many facets make up an individual’s credit story — beyond payment history and amounts owed. There is data that, once analyzed, can give lenders critical insights into borrower characteristics that can’t be categorized by a single number. People are more dynamic than their credit scores.  

Imagine a traditional consumer credit scoring model as a printed picture: a one-dimensional take on a person’s whole life in credit. In that static picture, there are balances on debt obligations, utilization of revolving types of credit like credit cards, delinquency and statuses, among others. This information comes from the three major credit bureaus — Equifax, TransUnion and Experian — and represents a vast cross-section of loans originated by banks, credit unions, finance companies and other lenders across the credit industry. This information adds up to that single definitive score.

In contrast, non-traditional models that build on the foundation of FICO can incorporate additional predictive information. Think of it as the motion picture version that creates a more dynamic view of consumer creditworthiness. This model gives lenders an ability to assess point-in-time information and the momentum of trended credit data factors, which may help predict the future credit conditions for a potential borrower and allow a lender to make more informed decisions. Bankers have greater visibility into the depth of a borrower’s story, like balances or utilization increasing or decreasing, and can capture that relationship with risk outcomes.

Alternative data sources can complement static and trended credit history by introducing consumers’ checking history, property ownership and alternative finance activity into credit scoring models. Consumers with comparable credit files can have vastly different repayment history and patterns; incremental information related to creditworthiness equips lenders to optimize risk differentiation when the credit file alone doesn’t capture the full story.

Creating a new model to calculate and predict high-performing loans is no small feat. BHG Financial, a leader in unsecured business and personal loans and creator of one of the country’s largest community bank loan networks, once relied on the traditional credit scoring model to help with their decision-making. The company decided to evolve their credit model to identify miscategorized but high-quality borrowers that most lenders were missing.

BHG Financial data scientists partnered with TransUnion to analyze over 2 million consumer loans; each loan was over $20,000, had at least 36-month terms and originated between 2015 and 2017. This amounted to more than a billion pieces of data points to analyze and assess, resulting in their proprietary credit model, the rScore.  An updated credit model resulted in faster approvals, and the identification of subprime borrowers that perform well along with prime borrowers with high default rates.

Evolving the already successful and established FICO score, the chances are lower that good-paying borrowers will be labeled as high risk. This enables some lenders to approve pockets of creditworthy consumers that others might decline. At the same time, the chance of labeling risky borrowers as low risk also declines — allowing lenders to protect the credit quality of their portfolio.

Lenders unable to dedicate time and money to develop their own evolved credit scoring model can collaborate with companies that have created updated credit models, skipping the extensive research and the costly origination process. This gives them immediate access to purchasing top-quality loans with low risk, which can quickly strengthen their loan portfolio to meet their bank’s criteria. This solution is possibly the best answer to finding a more future-forward way of predicting lending outcomes.    

Should Banks Use Facebook to Offer Credit?

data-source-2-26-16.pngIs it time for banks to start using Facebook profiles to offer a loan? Or a person’s Gmail account to verify an identity? A growing number of fintech start-up companies say so.

Banks have traditionally relied on credit bureaus to supply information not only about a person’s FICO score, but also to verify identity with data such as addresses. But sometimes, these data sources come up short. Tommy Nicholas, one of the founders of New York-based startup Alloy, says he has a few banks trying out his company’s platform, which basically allows a la carte access to a variety of traditional and nontraditional data sources to verify a customer’s identity, including credit bureaus as well as services that scan social media profiles or email accounts.

The idea is to make it easier for a bank to verify someone’s identity when traditional sources fall short, for instance, the person moved recently and the new address isn’t showing up on the credit report. “You end up asking half your customers to go find a phone bill and send it to you, not to mention you have all this manual work to do,’’ Nicholas says. “It adds a lot of friction.”

Alloy is trying to get traditional commercial banks interested in the technology for verification purposes, although its use to approve loans may be a ways off. Nonbank lenders already are using a host of online and offline data to make credit decisions, especially micro-finance lenders in developing countries that lack functioning credit bureaus. Some lenders are going so far as to analyze behavioral data to make credit decisions. Smartphone data, for example, can tell a lender that you regularly use a gambling app, which could be a black mark on your alternative credit score. Customers have to agree to provide lenders with the smartphone and social media data before they can be approved for credit. Facebook recently applied for a patent to use data on its users for loan underwriting—if your friends’ average credit score met a minimum, that could be a sign that you were a good credit as well, because you associate with people who have good credit.

Lenddo, which typically works with banks in emerging markets but recently began offering its service to U.S. financial institutions, has an algorithm that assigns a non-traditional credit score based on a variety of data to predict your willingness to pay back your loan. It also verifies identity using non-traditional data, such as Facebook profiles and email accounts. Socure uses what it calls social biometrics, where it pulls data from sources such as email, phone and social media accounts to create a risk score for fraud detection. Customers have to opt in to share their data. Other companies, such as Puddle, allow people to build a trust network online to give small dollar loans to each other. Everyone contributes something to the pool, and the more people you add to your trust network, the more you are able to borrow. Paying back your loan on time increases your trustworthiness.

Advocates of the use of alternative data, like Daniel Castro, the director of the Center for Data Innovation in Washington, D.C., say the plethora of online and offline data on each person is actually making it easier to detect fraud because it’s very difficult to sustain a fake identity online that will pass scrutiny. For example, if you worked somewhere for years, you probably have LinkedIn contacts who worked at that same employer. According to Castro, it’s extremely hard to fake your connections to multiple legitimate people. “More data can be useful,’’ he says. “Long term, every bank will be integrating more data sources. It will just be malpractice on their part not to, because it will reduce risk. It will just take awhile before they figure out the best way to do that.”

John ReVeal, an attorney at Bryan Cave, says the new technologies raise questions about complying with existing banking laws. For example, the Fair Credit Reporting Act applies both to credit and deposit accounts, and consumers have a right to know why they were rejected for either type of account. That could turn some of the new data providers into de-facto credit reporting agencies, he says. Additionally, banks may have to answer questions from their regulators about how they use alternative data to make credit decisions, and ensure such decision-making doesn’t violate anti-discrimination laws.

Will the new technologies provide a better way to analyze credit and approve accounts? That remains to be seen. For now, banks and alternative providers are experimenting with the possibility of augmenting traditional sources, rather than replacing them.