[1mm Financial Talk] Why Toss Bank Has a Particularly High Loan Ratio for Medium and Low Credit Borrowers

Toss Bank's Mid-to-Low Credit Borrowers Account for 36.3% of Credit Loans
10%p Higher Compared to Other Internet Banks

Over 200 Variables Considered in Credit Evaluation
Utilizes Toss App Data... More Borrower Information Than Other Banks

Introduces 'Machine Learning (MLOps)' for Immediate Credit Model Improvement

The third internet bank, Toss Bank, scheduled to launch on the 5th, is drawing attention from the banking sector. Even before its launch, it announced the release of groundbreaking interest rate products, including the highest deposit interest rate in the banking sector at 2% per annum and a minimum loan interest rate in the high 2% range, causing tension not only among other internet banks but also among commercial banks. Especially as financial authorities tighten regulations on household loans, leading to rising loan interest rates and restrictions on loan limits, Toss Bank's announcement of unsecured loans with a maximum limit of 270 million won is expected to attract demand from so-called "loan refugees." The photo shows the Toss Bank headquarters in Gangnam-gu, Seoul, on the 4th. Photo by Hyunmin Kim kimhyun81@

[Asia Economy Reporter Sim Nayoung] Mr. Kim Myeong-eun (36) once experienced a bank loan delinquency when he took a short break from work three months ago. He is currently repaying a card loan he took six months ago whenever he has money. Based solely on this record, the bank's internal evaluation rated Kim's credit score as grade 8, making additional loans impossible. However, at Toss Bank, Kim was able to get a 2 million won unsecured loan. This was because the bank utilized other information about the loan applicant beyond the basic history. They thoroughly examine whether there are regular deposits into the applicant's account and whether there are consistent periodic withdrawals.

Kang Shinhyung, Manager of the Credit Modeling Team at Toss Bank, said, "Although the loan applicant has some negative factors, they are one-time occurrences. By reviewing other records to see if Kim is working diligently and earning money, and whether he is paying regular expenses like monthly rent without delay, Toss Bank decides on loan approval."

According to Toss Bank on the 14th, the proportion of unsecured loans to middle- and low-credit borrowers (based on outstanding balance) was 39% as of the end of September. This is about 10 percentage points higher compared to other internet banks. The target proportion for unsecured loans to middle- and low-credit borrowers by the end of this year is 42%. What is the reason for such an outstanding performance?

Kang Shin-hyung, Manager of Credit Modeling Team at Toss Bank

Internet banks were created for middle- and low-credit borrowers who cannot get loans from first-tier banks and face high interest rates in second-tier financial institutions. When loans to middle- and low-credit borrowers increase, concerns about the bank's soundness naturally arise. To dispel such concerns, it is important to select middle- and low-credit borrowers who are likely to repay and offer reasonable loan limits and interest rates. This is why credit scoring models (CSS) are a core competitive advantage for internet banks.

Manager Kang said, "For example, when evaluating the credit of middle- and low-credit borrowers, commercial banks consider 15 to 20 variables, other internet banks consider 50 to 60 variables, but Toss Bank looks at 200 variables. When a user of the Toss app, which consolidates all their financial information, applies for a loan at Toss Bank, the bank can review their card usage history and even assess their consumption tendencies," adding, "The coverage of borrower information is much broader than that of other banks."

There is also a differentiation in IT infrastructure compared to other banks. Toss Bank can move faster than others when it needs to improve its credit scoring model. It is the only domestic bank that has introduced a 'Machine Learning Operations (MLOps)' system that allows developers to immediately deploy newly created models. To use a car analogy, normally after creating a prototype and handing the design to the factory, the factory needs preparation steps for mass production. With MLOps, production starts immediately upon receiving the design without preparation.

Thanks to this, it is possible to promptly modify the credit scoring model to suit the current rising interest rate environment. Manager Kang emphasized, "When market conditions change rapidly, we need to respond by tuning the credit scoring model. For example, when interest rates rise, we can increase the weighting on loan repayment likelihood. When the development team changes the weights, the updated credit scoring model is immediately applied at Toss Bank."

This content was produced with the assistance of AI translation services.