How bank clients are assessed
Posted: Mon Jan 06, 2025 5:09 am
For retail clients, the WOE binning method is mainly used, with the help of which a continuous numerical variable or a categorical variable is divided into groups, the boundaries of which are optimized according to the Information Value (IV).
The transformed variables are then selected by their degree of influence in logistic regression, using correlation analysis and recursive feature elimination (RFE) methods for this purpose. As a result, a scoring card is formed that divides clients into risk groups.
Example of a scoring model
Example of a scoring model for individuals
Thus, summing up the client’s points depending on benin whatsapp phone number their belonging to a particular group by variables, we obtain the final score, which determines the client’s rating:
Final score
Final score and customer rating
For corporate clients, a more complex model is used, including financial, qualitative and behavioral blocks, which are processed to obtain a basic client rating.
Clients are segmented depending on the type of business: large, SME, micro or individual entrepreneur. For the largest clients, including states, an international assessment from rating agencies is used.
Example of a risk scale
The bank can also use its own risk scale, which depends on the characteristics of the corporate portfolio of a particular bank.
The model itself consists of blocks, not just one scoring card, as in retail. Each of the blocks (financial, qualitative, behavioral) is a separate scoring model. For preprocessing of variables within each model, not only binning can be used, but also normalization and standardization methods.
Corporate Client Model
The resulting blocks of models are then combined into a single base model (usually through logistic regression). The base model is used to calculate the client's base rating, which is subsequently adjusted in stages based on the risk factors in the adjustment block.
Read also: What is acquiring and why does a business need it?
the banking sector.
The transformed variables are then selected by their degree of influence in logistic regression, using correlation analysis and recursive feature elimination (RFE) methods for this purpose. As a result, a scoring card is formed that divides clients into risk groups.
Example of a scoring model
Example of a scoring model for individuals
Thus, summing up the client’s points depending on benin whatsapp phone number their belonging to a particular group by variables, we obtain the final score, which determines the client’s rating:
Final score
Final score and customer rating
For corporate clients, a more complex model is used, including financial, qualitative and behavioral blocks, which are processed to obtain a basic client rating.
Clients are segmented depending on the type of business: large, SME, micro or individual entrepreneur. For the largest clients, including states, an international assessment from rating agencies is used.
Example of a risk scale
The bank can also use its own risk scale, which depends on the characteristics of the corporate portfolio of a particular bank.
The model itself consists of blocks, not just one scoring card, as in retail. Each of the blocks (financial, qualitative, behavioral) is a separate scoring model. For preprocessing of variables within each model, not only binning can be used, but also normalization and standardization methods.
Corporate Client Model
The resulting blocks of models are then combined into a single base model (usually through logistic regression). The base model is used to calculate the client's base rating, which is subsequently adjusted in stages based on the risk factors in the adjustment block.
Read also: What is acquiring and why does a business need it?
the banking sector.