Challenges to overcome
Banks now depend on thousands of models to support business functions ranging from risk management, regulatory reporting and capital planning to AML, fraud and marketing. Modeling communities in these functions tend to have different needs, use different techniques, and build and execute models with different tools
During the model-building process, this diversity is a huge advantage. It empowers each modeling community to use the right tools for their specific business needs – and unlock the full power of modeling languages such as SAS®, Python and R.
But these models still need to be managed, executed and governed separately. Managing them using completely different toolsets is a liability. The complexity makes it difficult to establish a single model risk management framework for all model types. And as regulatory scrutiny continues to extend across a wider range of modeling communities, this will become increasingly problematic.
When downstream model management processes aren’t standardized, it can also lead to inefficiencies. For example, manual handoffs between data science, IT and line-of-business teams can cause delays and increase the risk of errors. Documenting models and managing approval flows are tedious and time consuming, distracting skilled resources from higher value work. And the resulting lack of agility can make it difficult to adapt models quickly when regulators require changes or suggest improvements.