Model risk management

Institutions increasingly rely on models for the decisions they make. These decisions can be for internal or external purposes, but in the end they will always impact their clients and other stakeholders.

As a consequence, more frequent use and increasing complexity of models (for example with machine learning models) can cause model errors. These model errors can lead to wrong decisions. This can lead to e.g. reputational or financial damages to institutions and their clients. In order to identify the challenges that institutions face regarding model risk, institutions should ask themselves:

• What defines a model and how many models do we have within our institution?
• Are we effectively measuring model risk across the entire model lifecycle?
• Do we have a governance with clear roles and responsibilities for everyone working with models in our institution?
• Have we defined and executed the right processes to mitigate the identified model risks?
• Do we have an integrated model risk management system in place to keep track of all models and the resulting model risks?
  

An effective model risk management framework can help to mitigate model risk. We design and develop these frameworks using six building blocks or components:

• Governance, with an integrated and complete model stakeholder universe, across all lines of defence and senior management. This includes the alignment of roles and responsibilities within the existing organisational structure.
• Policies and documents, around a single overarching MRM policy that governs and links to all model risk documents and policies, standards and work instructions.
• Inventory and model landscape, driving the scope of the framework with a clear model definition that should include all models within the institution.
• Processes, standardisation and automation of processes to allow for process efficiencies across models and more continuous insights into model risk.
• Workflow and lifecycle management, to set clear controls around all model lifecycle steps, but also to coordinate and prioritise resource allocations of work based on model risk monitoring.
• Monitoring and reporting, enabling a risk-based view of the model landscape including a model risk appetite statement.

We strongly believe that model risk should be included in the risk management framework of institutions for strategic reasons. With a mature model risk management framework, institutions create insights into the entire model landscape and across all steps of the model lifecycle. This in turn leads to awareness and mitigation of model risk within institutions.

We have established a leading role in model risk management framework by covering our clients’ needs with the following services:

• Design, development and implementation of model risk management frameworks or specific components.
• Maturity and ambition assessment of model risk management improvements, including a clear roadmap and follow-up plan.
• Deloitte Model Analytics Platform, our in-house developed model risk management tool.
• Model validation managed services, where we act as a flexible pool of experts to enhance internal teams with resources for model development or validation.
  
• Development and automation of model monitoring and model validation frameworks, in order to standardise and automate model monitoring and create more continuous insights into model quality and model risks.