Validation

The Rockboro team has an extensive history of successfully assisting financial institutions to achieve IRB recognition and meet their internal validation needs for almost a decade. While Rockboro brings specific expertise to the credit risk assessment of low-data sectors our team also has a long and successful track record of validating models and methodologies in all sectors including those (such as retail credit) that sometimes benefit from the presence of high volumes of default, loss and / or behavioural data.

In reviewing existing methodologies we typically not only focus on the analysis and interpretation of the data relating to performance of models but critically also examine:

• The suitability and completeness of data for the methodology’s stated assessment objectives

• The methodology’s “fitness for purpose”

• The conceptual soundness of the methodology

• The construction process undertaken

• The extent of alignment with any third-party methodology, if appropriate

• The process in place for ongoing surveillance and monitoring of the methodology

• Completeness of documentation for complying with best practice internal risk management and preparing for IRB recognition

Rockboro aligns its validation assistance services to the particular needs and goals of each client. Services provided include:

• A comprehensive review of the data, construction, conceptual soundness and performance of a methodology
• A review of specific validation attributes (e,g, solely the conceptual soundness of a model)
• A high level “health check” on a client’s internal validation process
• Benchmarking services
• Provision of a detailed guidance note for establishing or enhancing the internal validation process and protocols
• Training on the application of the validation process.

Rockboro brings specific experience to dealing with the validation of credit risk assessment methodologies for low-data sectors (such as large corporates, financial institutions and specialised finance) where historical data is insufficient to train a model purely on the basis of data.