Model risk can be defined as the potential loss an institution may incur, as a consequence of decisions that could be principally based on the output of (internal) models, due to errors in the development, implementation or use of such models (CRD IV, Article 3.1.11).
While certain elements of model risk may be state driven (i.e., prevailing at a certain stage of the model lifecycle) - such as model initiation, development, implementation, usage, ongoing monitoring and decommissioning – the risk of model decay is ubiquitous and may materialize at any point during a model’s existence.
Although the use of models by financial institutions has brought objectivity in their decision making, it has also led to a significant increase in model risk. Mitigation of model risk requires effective and robust implementation of an MRM framework that involves defining ownership, describing roles and responsibilities and enabling various stakeholders to work together in a synchronized manner.
The MRM function acts as a key point of reference for all matters related to model risk and paves the way for setting up the guidelines on model development, validation, classification, monitoring, documentation, inventory and reporting.
Furthermore, as banks continue to double down on their investments in groundbreaking financial technology, advanced analytical tools and resources skilled in the study of data science, the coverage and sophistication of model development is only slated to increase – modern-day advances in process automation, machine learning and artificial intelligence point towards exponential growth in the ways data can be analyzed and manipulated to create a competitive advantage, from gaining a more thorough understanding of the risks pertaining to business decisions undertaken by institutions to developing strategies for customer targeting and market penetration.
While the use of predictive analytics is on this upward trajectory, investing in a strong model monitoring framework, with dedicated roles and responsibilities extending across all phases of the model lifecycle, is a pressing requirement that banks can no longer afford to circumvent. Just as a colossal skyscraper will crumble without a strong cornerstone, predictive models can only benefit an organization if they have been developed, implemented, applied and reviewed correctly. With great power comes the need for greater due diligence.