Model Calibration & Validation


How well is your risk rating model performing? Do you trust it enough to use it effectively and reach business decisions? Even if the model was performing well when it was first designed, it should be regularly reassessed in an objective and independent manner, and eventually fine tuned.

Main Features

In this context, our role includes analysis of borrowers with different grades assigned to them by yourselves at different times in order to assess the existing model's predictive power. Other characteristics are also examined, for example possible abnormal concentrations in a limited number of grades, and the mapping of risk grades to default probabilities.

Following the above validation, we conclude using objective statistical criteria if the model needs to be re-parameterized, if some risk factors have lost their predictive power or do not contribute significantly to the model’s overall power, and if other factors that were not used in the past merit to be introduced in the model.