Cognilytics has developed a structured approach for managing the risk associated with using Models to make decisions.
Many financial services companies are now required to have their models comply with strict regulatory standards (e.g. OCC 2011-12, FRB SR 11-7, Basel II, Solvency II). Cognilytics provides end-to-end Model Risk Management solutions including:
- Data Quality and Integrity—Lineage and process verification, data element reconciliation, data quality verification
- Model Development – Methodology & Design, Modeling Dataset Creation, Model Estimation
- Model Refresh – Rapid re-estimation of model parameters
- Model Validation – Independent review of conceptual soundness, replication of model development, benchmarking against alternative models/approaches
- Model Performance Management – automated ongoing monitoring, outcomes analysis, stress testing, sensitivity analysis
- Model Governance – Documentation templates, audit trail/work flow management, change controls, model risk frameworks
We have applied these solutions to hundreds of models for industry leading clients in both banking and insurance.
Our team includes former Credit Risk and Model Risk leaders from top financial firms like Capital One, American Express, Wells Fargo, AIG and HSBC. We have partnered with top companies in both Banking and Insurance to accelerate the adoption of these tougher regulatory standards, and in many cases influenced how these firms have chosen to comply with the requirements.
Our approach to Model Validation and Model Performance Management is structured enough to ensure compliance with the regulatory guidance yet flexible enough to be applied as deeply as our clients need us to go. We leverage our expertise in Independent Review, Replication, and Benchmarking to deliver results that hold up to scrutiny from the OCC and Federal Reserve. We have tackled models of various types and levels of complexity, including models with additional regulatory requirements (Basel, CCAR, DFAST, AML), non-traditional models leveraging judgmental inputs and/or logic, third-party models, and models lacking sufficient documentation.