Forbes: How shopping habits can reveal credit risk to banks and insurers

Traditional credit scoring methods often exclude consumers who lack formal credit histories, leaving a significant portion of the market underserved. In the U.S., 28 million adult Americans are credit invisible and another 21 million are unscorable. Many of them are new immigrants, young people entering the economy and gig workers who have simply not built up a credit history.

This gap presents a substantial opportunity for banks to use new alternative data to assess creditworthiness and broaden their customer base through advanced AI techniques and privacy-preserving data collaboration.

Grocery shopping data: A breakthrough in credit scoring

In Africa, we’ve shown that grocery shopping data can be used to predict a customer’s financial status, backing up findings from the study “Using Grocery Data for Credit Decisions,” published by researchers at the University of Notre Dame, Rice University and Northwestern University Kellogg School of Management.

The study showed that grocery shopping habits can serve as a reliable indicator of financial responsibility and loan repayment capabilities. The types of items consumers add to their shopping cart (such as healthy items versus cigarettes), as well as the frequency and size of their purchases at the local grocery store, help to build a behavior profile that can be compared to the behavior of credit customers who have been reliable at paying back their loans.

The real-world use of one of Africa’s largest retail grocer’s loyalty shopper data has helped South African banks predict creditworthiness for applicants with little to no credit history. By overlapping their customer datasets and employing advanced cryptography and AI, the bank analysed the shopping behavior of their best credit customers to create a predictive model that can be used to predict loan repayment probability from new credit applicants with no credit history—based purely on their shopping behavior.

In this case, eight million people who might previously have been “invisible” to the bank because they are unbanked or work in the informal economy, for example, were scored using shopping behavior data, and 3.2 million of them now qualify for affordable credit. In fact, one participating South African bank reported a predicted 29% increase in credit revenue by targeting this new audience, who were previously deemed to be too high-risk due to a lack of traditional credit information.

The challenge of data access and privacy

Accessing this data for credit scoring has long been a challenge. The growing risk of data breaches from sharing data between banks and retailers, along with strict privacy regulations and rising consumer privacy concerns, has made this difficult.

However, banks are overcoming these challenges by using privacy-preserving techniques such as data anonymization and secure multiparty computation to access and analyze this data without exchanging or revealing any personal information.

So, how does it work? Advanced cryptography and artificial intelligence are applied to consumer data that is first anonymised and then shared between a grocery retailer and a bank in a secure and neutral environment. Banks - or other financial institutions - and retailers can then analyse each other’s behavioral, demographic or transactional data without compromising individual privacy or data ever exchanging hands.

AI-driven models can be built from the shared data, then tested and used to identify patterns indicative of financial responsibility. By applying these AI models, banks can predict creditworthiness based on shopping behavior with impressive accuracy—achieving a GINI lift of 41% on thin-file clients, in our experience.

Appropriate assistance to more vulnerable consumers

The Consumer Financial Protection Bureau (CFPB) emphasizes the importance of providing consumers with access to credit that is fair, affordable and in line with responsible lending practices. Similarly, the U.K.’s Financial Conduct Authority (FCA) supports “a thriving and innovative Consumer Lending market, where consumers have access to credit which is affordable for them,” while also ensuring that consumers facing financial hardship receive appropriate assistance.

By providing access to credit for individuals previously considered high-risk, banks can offer personalized and timely credit products that meet the unique needs of these consumers. This will ultimately build trust and cultivate long-term loyalty—crucial in an increasingly competitive market.

Furthermore, by leveraging accurate and inclusive credit scoring models, banks can increase their profitability. The ability to identify low-risk customers from the previously underserved population allows banks to extend credit more confidently, reducing the likelihood of defaults and enhancing overall loan portfolio performance.

Read the full text article on Forbes.com

About the Author

Jon Jacobson is CEO of privacy-preserving data collaboration platform business Omnisient. Read more about Jon's background here.