Why eCommerce Needs a New Way to Assess Credit
Through Christian Mangold, CEO at Fraugster
Key points to remember:
- Traditional credit scoring models focus on a customer’s credit history and repayment history. But this approach excludes millions of potentially profitable customers and leads to unnecessary costs for companies offering BNPL services.
- Companies offering BNPL services should leverage additional data sources such as utilities, mobile phones, and rental payments, as well as device attributes and positive account balances (to name a few- uns) to improve the accuracy and reduce the cost of credit risk assessments.
- We describe how e-commerce can make alternative credit decisions by using machine learning to enrich data and network intelligence to benefit from insights into repeat buyer patterns and networks.
Credit checks are designed to answer a simple question: “Can I trust this person to repay the amount they borrow?” “. Before extending credit, a lender will perform a credit check to assess the borrower’s risk level. This check will then be performed by one of the big three credit bureaus, Equifax, Experian and Transunion. These credit checks help lenders obtain information about your financial history; including prior credit repayment, types of credit taken and length of customer’s credit history.
The problem is that these forms of credit scoring were designed for high-value, long-term financial products like personal loans and mortgages, not low-value online purchases. This means that traditional credit assessments, which continue to focus heavily on credit history, lead to false positives, denials of service and increased costs for companies offering BNPL services. So let’s take a closer look at the limitations of existing credit risk models and highlight opportunities for improvement.
Understand how credit scores are calculated
Payment history accounts for 35% of the score and focuses on past credit repayments (revolving/temporary), whether payment terms were made on time, and any past bankruptcy records. Late payments of more than 30 days negatively impact a customer’s credit rating and remain credited to them for up to 7 years. On the contrary, one-time payments can help significantly increase a customer’s score.
Amount of debt represents 30% of the score and measures the proportion of money the customer currently owes compared to the total credit available to the consumer. This is called the Credit utilization ratio. If, for example, the customer has $500 on their credit card and their limit is $1,000, their credit utilization rate is 50%. In general, having a credit utilization rate above 30% has a negative impact on one’s credit rating.
Length of credit history represents 15% of the score and measures the average age of a customer’s credit file. Having a long credit history with open and active accounts has a positive impact on credit rating. The old adage that “past behavior is the best predictor of future behavior” applies here, as well as to payment history.
New credit accounts for 10% of the score and focuses on the number of credit accounts opened within 3-6 months and the number of difficult demands lenders do this to assess a borrower’s credit report when applying for new credit. Informal requestsusually consisting of a borrower checking their own score, pre-approved loan checks or account reviews do not impact customers’ credit scores.
Composition of credit accounts for 10% of the mark. Lenders prefer to see a healthy credit mix that demonstrates the ability to handle different types of credit. Which means having a combination of credit cards, installment loans, and home loans helps boost a customer’s credit rating.
Although such a credit scoring mechanism is used by the three major credit bureaus, namely Experian, TransUnion and Equifax, stick to such a credit scoring model,
Quantify the risk associated with granting credit
Based on these factors, credit reporting companies calculate a three-digit score ranging from 300 to 850. These credit scores do not predict individual defaults, but rather place individuals among a group of people expected to exhibit the same default rate. A high score implies less risk for the borrower. For example, customers with a score above 810 have an average defect rate of only 0.1%. Even a very good score above 700 leads to an average defect rate of only 2.5%, while consumers with the lowest score (below 400) have a defect rate of nearly 50%.
The limits of traditional credit ratings
The reliance of credit scores on historical data discriminates against people who have had little or no credit in the past. This is especially true for younger cohorts, who form a major consumer group for BNPL services. Traditional credit scoring models also reinforce existing inequalities and disproportionately penalize marginalized groups. Nearly 26 million Americans are considered invisible credit, the majority from black and Hispanic communities. A recent study from Stanford University points out that minority credit scores are 5% less accurate in predicting default risk than non-minority borrower scores.
Another factor to note is that credit reports and scores often differ between credit bureaus. This is because a credit report is based on information provided by various lenders and they may not report every piece of information to all credit bureaus. The material impacts of inconsistencies were highlighted by the Consumer Financial Protection Bureau (CFPB) report, which found that inaccuracies in credit reports have doubled during the pandemic due to these same reporting issues.
Finally, traditional credit scores travel poorly, meaning they are not always transferable from one country to another. An immigrant with an exceptional credit score in his country of origin can arrive in a new country and be credited invisible. This data irregularity currently prevents a fully qualified borrower from accessing credit.
Alternative data sources for measuring credit risk
Traditional credit reporting mechanisms provide limited insight into a consumer’s situation true financial situation and, more importantly, for lenders, their real risk of default (also called bad rates on loans). This highlights the need for alternative data sources that provide additional financial information and context. Some of these alternative sources include timely payment of:
- Utility bills (gas, water, electricity)
- Telco (TV, mobile, broadband)
- Rent and mortgage
Open banking also offers better access to:
- Deposit account information (payroll deposits, average balance, etc.)
- Positive transaction history
In addition to this, providers should also consider:
- Digital Footprint (Device ID), which provides insight into historical purchase behavior
- IP address, to better track and identify customers who have made purchases using guest checkouts
The use of a wider range of data sources, such as checking account activity, is noted to improve the predictability of a possible default. It further provides lenders with insight into the payment behavior of the borrower in several areas and aspects of their life. Alternative credit scoring also overcomes the recency and reporting issues prevalent in traditional risk scoring models. Missed payments in the past can impact a borrower’s credit score for up to seven years and can therefore be classified as a lagging indicator and snapshot of a consumer’s risk. Alternative data sources provide the opportunity for more up-to-date, real-time risk assessments.
So who are the winners of a move to different credit ratings?
The biggest beneficiaries of alternative credit scoring are students, immigrants, and marginalized communities who are new to the credit ecosystem. In a PERC study, the addition of alternative data increased the credit scores of 64% of borrowers with a low profile. The Corporation for Enterprise Development (CFED) further estimated that adding utility payment data alone would reduce credit decline among thin-file borrowers from 20-25 million adults to approximately 5 million adults. .
This, in turn, presents an opportunity for lenders. The World Bank projects that alternative data could help provide formal financial services to up to one hundred million more adults.