Credit quality

Sberbank not only saw an improvement in its headline NPL metrics, but also the underlying dynamics of loan book quality reflected a continuing improvement in loan book risk metrics. Based on our models tracking previous loss experiences as well as other exogenous factors driving loan book quality, we see a continuing trend of declining expected loss (EL) value as a percentage of our retail loan book. Since the launch of our Credit Factory platform for retail business in October 2008, we have continuously improved and expanded the functionality of our loan application scoring and approval practices. The main developments of Credit Factory are concentrated around three main lines:

1. the development of new credit-scoring models and their practical implementation;
2. the introduction of new data sources into the decision-making process and automation of data exchange processes; and
3. the development of new scoring and data processing technologies.

In 2012, our main accomplishments in the field were as follows:

• first, we note the development and implementation of new region-specific scoring models for consumer loans. Existing data indicate that clients’ risk profiles differ significantly across Russia’s regions; moreover, the most relevant factors also have notable regional variability. Therefore, we built regional clusters with similar risk patterns and developed individual scoring models for each cluster. As a result, rejection levels across the regions much better reflect their specifics and the overall portfolio risk level has reduced;
• the implementation of risk-based pricing for consumer loans from March 2012 enabled us to increase the yields of higher-risk portfolios and attract higher quality clients by offering them lower-yielding loans;
• the introduction of new scoring models in mortgage lending resulted in a 0.5 pp higher approval level and lower credit risk for new portfolios;
• we integrated our data with the Equifax Interbank Fraud Prevention Service platform, aimed at discovering inconsistencies in client data and preventing fraudulent activities by borrowers;
• we instituted an automated client reliability assessment based on State Pension Fund data as part of scoring consumer loan applications. This new data source allows us to evaluate and verify the stability of a prospective borrower’s employment and income, as well as receive indirect confirmation of their employment history. The system was launched in Moscow in 2012, and we are currently planning to implement it nationwide;
• we started testing automated photo processing and verification technology. This is another anti-fraud measure aimed at handling cases of identity theft by adding another dimension to the client verification process. This is scheduled for launch in 2013.

Tools:

• Abbreviations
• Glossary

• Contacts
• Corporate website