Skip to main content

Credit Modelling - OSIS

OSIS offers credit modelling solutions to help banks, Fintech companies, asset managers, pension funds, and insurance companies make better credit decisions and manage their portfolios effectively. Our solutions include powerful tools for credit risk assessment, expert guidance on credit modelling, and training to help you develop the expertise needed to manage credit risk effectively.

Credit Modelling

Credit analysts have difficulties learning from their experiences. This is because the feedback loop between granting the loan and the default of a loan can be a number of years. Either the analyst can’t remember clearly what their motivations were when granting the loan, or they changed position before the loan went into default. As far as data is concerned the timespan of the feedback loop doesn’t matter if the quality of the data is correct. Compare it to driving a new car. The feedback loop between using the brakes or accelerating and the reaction of the car is a fraction of a second. It is the same with credit models. They usually learn better and can therefore outperform credit analysts because the data used is a combination of the experiences of all analysts within one institution or even from several institutions.

Even so, there are some drawbacks to credit modelling which means the credit officer still has an important role to play. Firstly, data is scarce and secondly, data does not fully represent what the future holds and finally, there is always some information known by the credit analyst which was not captured by the data. In other words, the credit analyst and the credit model need to become a team. 

There are different areas where we can help lenders set up their credit decision tooling: 

Firstly, we can advise lenders in setting up a coherent data framework that suits their lending activities and that can form the basis for future model development.

In addition, we can set up credit risk models and rating systems for making provisions and determining pricing even if sufficient data is not yet available.

We can then create the simplest and most intuitive model possible given the available data that will serve as the basis for different ways in which credit risk can be measured:

  1. Through-the-cycle measures for the calculation of regulatory capital and economic capital; 
  2. Point-in-time measure for the calculation of IFRS9 provisions and stress testing and 
  3. Multi-year through-the-cycle and point-in-time calculations for securitization transactions.

We can integrate the credit decision tooling in the data framework to secure a full audit trail for future model validation and recalibration.

Furthermore, we can train the users of the models to efficiently use the outcomes of the models in their credit decisions, their loan management and the pricing and valuation of the loans.

Ultimately, we can facilitate the process of interaction between the risk model and the credit analyst, with the collection of new data leading to better models and better use of the models.