Raiffeisen Bank automated valuation of collateral facilities in corporate lending

2021: Automating the valuation of trucks and residential properties

Raiffeisen Bank September 14, 2021 announced that he automated the assessment of freight cars and residential real estate as collateral in the corporate crediting with the help of technology (ML machine learning). This will allow you to increase the speed of evaluation (time to collateral) and the issuance of loans to customers. According to to data experts, bank freight transport and residential real estate are collateral for corporate loans in 10% of cases.

Effective assessment of collateral objects in lending is a key element of maintaining the high quality of the bank's loan portfolio. Depending on the team's experience, this mechanical process can take more than 8 hours. Now for two types of collateral in corporate lending - residential real estate and trucks - the assessment of objects in Raiffeisen Bank is carried out using decision support systems, and in the future it will become completely automatic. At the same time, the rate of valuation of collateral will increase three times - from 8 to 2.5 hours.

We use big data and machine learning to increase the speed of credit decisions and the effectiveness of our risk models, "said Roland Wass, head of risk management at Raiffeisen Bank. - Our strategic goal is to build an organization in which decisions are made on the basis of data. Automation of valuation of collateral objects in corporate lending is a very big step forward in creating a data-driven bank. At the same time, machine learning technologies allow not only to increase the operational efficiency of our business, but also to free up team time for more interesting tasks.

In order to evaluate the pledge, the expert introduces data in the specialized system: a cadastral number in the case of an apartment and a brand, a model and mileage in the case of a truck. The ML model estimates the market value based on the data of all comparable objects. The analyst can only verify the estimate offered by the model.

We were one of the first in the market to automate the assessment of trucks as collateral. The development of the model took our team about six months, "said Mikhail Grinenko, head of the credit control and collateral department, Raiffeisen Bank. - The main challenge of this project was related to the formation of a high-quality dataset and its processing. The obtained performance indicators of the models are superior to similar projects of other subsidiary banks of the RBI group.

MAPE (Mean Average Percentage Error) was used to assess the quality of models, which was 8.1% for apartments and 9.5% for trucks. Raiffeisen Bank collateral team plans to expand the use of models to assess collateral. In particular, personal cars, production equipment, trailers and semi-trailers, as well as commercial real estate will be included in the list of facilities.