NH NongHyup Bank weaves 10,000 alternative data points into credit model

Banks are turning to alternative credit evaluation to embrace borrowers with little financial transaction history. NH NongHyup Bank on Nov. 9 introduced the All One Bridge Model, a credit scoring model based on non-financial alternative data, into its loan screening. The move lowers the lending bar to bring in financially underserved groups, while a key question has emerged: whether non-financial data can precisely gauge a borrower's actual repayment capacity.

The model was developed by analyzing more than 10,000 alternative data points obtained by combining non-financial pseudonymized information held by nine institutions. Beyond telecom bill payment records and small-amount transaction data, which had been used only to a limited extent in traditional credit scoring, the model comprehensively reflects utility bill payment history, e-commerce shopping frequency, and consumption pattern data. These are used to quantify a customer's economic activity from multiple angles for loan screening.

In the past, bank credit evaluation relied overwhelmingly on financial records such as loan history and credit card usage. As a result, thin-filers—young adults just starting their careers, homemakers, retired seniors, and small business owners—found it difficult to obtain loans from first-tier banks or had to pay high interest rates even when they had income and a genuine willingness to repay. NongHyup Bank's adoption of alternative screening is seen as a strategy to absorb these borrowers in blind spots and secure potentially prime customers first.

The alternative credit evaluation market is already led by internet-only banks such as KakaoBank, which have delivered tangible supply results. KakaoBank has provided an additional 1.2 trillion won (about $886 million) in mid-to-low credit loans through its proprietary KakaoBank Score model, which combines data from eight institutions including retail and telecom companies. Major commercial banks such as Shinhan Bank and Hana Bank are also raising the competitive stakes by applying alternative evaluation models that combine non-financial data in their operations.

From a risk management perspective, however, incorporating non-financial data carries considerable uncertainty. Telecom and utility bill payments and online consumption patterns differ fundamentally from the heavy obligation of repaying bank loan principal and interest in terms of legal binding force. In other words, frequent shopping or timely utility payments do not necessarily mean a borrower can handle multiple-debt risk or repay loan principal and interest during an economic downturn. If the model merely boosts loan approval rates, noise or distortion in the alternative data combination process could boomerang into a surge in loan-loss provisions that erodes bank profitability.

The recent upward pressure on delinquency rates for loans to sole proprietors and mid-to-low credit borrowers makes verifying the alternative model's discriminatory power even more demanding. Aware of this default risk, NongHyup Bank chose to tighten its post-loan control network instead of lowering the new screening bar. It will split the maturity extension review system for existing borrowers, reassess each customer's credit risk level to strengthen follow-up management, and simplify and automate extension procedures only for prime customers with proven repayment capacity.

Ultimately, the success of alternative credit evaluation depends not on how many loans are extended but on whether the delinquency rate of incoming thin-filer borrowers can be stably held down to the level of prime borrowers. Whether the more than 10,000 non-financial alternative data points can serve as an early warning system that preemptively detects signs of borrower distress, and whether the model can continue to prove its credit risk discrimination amid economic volatility, will determine the future of asset soundness defense.

(Dollar figures are approximate, converted automatically at 1,000 won = $0.738, the rate on Nov. 9, 2026.)

※ This article is an AI translation of the Korean original published by Korea Credit News; it was checked by automated quality guards and published by the English edition desk.