Model Risk Management Lifecycle Overview - MATLAB & Simulink
Video Player is loading.
Current Time 0:00
Duration 5:12
Loaded: 3.15%
Stream Type LIVE
Remaining Time 5:12
 
1x
  • Chapters
  • descriptions off, selected
  • en (Main), selected
    Video length is 5:12

    Model Risk Management Lifecycle Overview

    Brian Douglas

    Watch an overview of model risk management, a critical process for financial institutions to manage the risks associated with their decision-making models, impacting areas such as pricing, credit scoring, and fraud detection. Learn about models’ potential risks, such as errors, biases, and misuse, leading to unreliable business decisions. Be introduced to Modelscape, a model risk management solution by MathWorks, to manage these risks, particularly when handling numerous models at different lifecycle stages. Model risk management can be broken into six components: development, validation, testing, deployment, monitoring, and governance, each addressing different risk aspects in a model’s lifecycle. You will gain insight into how risks are introduced at each stage and the necessity of an enterprise-level framework for managing these risks effectively.

    Published: 19 Jan 2024

    Model risk management is a process that financial institutions use to identify, measure, monitor, and control the risks associated with their models. This is important because models are used to make data-informed decisions, like pricing and valuation estimations, credit scoring, and even marketing. However, the use of models has risks. They can be prone to errors, biases, and assumptions.

    Also, a model could be developed for one purpose and then be misused for another purpose. And each of these risks can lead to incorrect or unreliable results, which means potentially making the wrong business decisions. This is where model risk management comes in. To understand why model risk management is important, let's look at the different ways that model risk occurs and what can happen if it's not properly managed.

    The first place where risk can be introduced is during the development stage of the model. At the very beginning of model development, it's important to identify the business need for the model and to document what the model is for and who is responsible for developing it. With the right documentation and communication in place, it lowers the risk that the model's purpose and intended audience is unclear. This minimizes the chance of a misalignment between the stakeholders and the development team, which would potentially result in wasted resources and time.

    During the creation of the model, there is risk that the wrong data or the wrong modeling methodology is used and therefore increases the risk of building a model with errors and biases. Minimizing risk during model development is important because finding and correcting those errors early is cheaper than finding them during the validation and testing of the completed model.

    Validation is where we check to make sure that we are building the right model. This includes making sure the model performs as expected by understanding the model's assumptions, data sources, and methodologies, and ensuring that it achieves the intended purpose by aligning with the expectations and the needs of the business. After validation, we test the model.

    Testing means that we are checking the model's performance under current scenarios and conditions to assess compliance with the requirements. Testing decreases model risk by showing that the entire operational range of the model meets the expected accuracy and reliability. With a valid and tested model, the next step involves transforming it into a production-ready model that can be used by the business to make decisions. In this step, the model transitions from the developer's environment where it has been tested and verified into a model that is ready to be deployed to different environments across the business.

    Model risk can be introduced if the development environment is different than the deployed environment. For example, delays in deploying models can be a source of risk if the model is out of date by the time it is deployed. Once a model is deployed, risk still needs to be assessed through ongoing monitoring of the model's performance and results. This includes tracking the model's accuracy, reliability, and consistency over time and identifying any changes or trends that may affect its performance.

    For example, assumptions under which a model was created may no longer be valid, or drift in the model input data can lead to unacceptable performance degradation. In this way, risk can be introduced over time that didn't exist when it was first deployed. The final component of model risk management involves model governance to ensure that every stage of a model's life cycle is continuously subject to appropriate oversight, controls, and accountability. This includes establishing policies, procedures, and standards for model risk management and ensuring that they are followed consistently across the institution.

    As we can see, managing risks throughout the lifecycle of a single model is important. However, institutions often have to manage dozens or hundreds of models. And managing individual models in isolation increases the risk of overlooking the potential impacts stemming from their interdependencies. This is what justifies the need for Modelscape, an enterprise-level framework for model risk management.

    Modelscape is a solution that enables financial institutions to reduce the complexity of managing the life cycles of financial models while improving model documentation, transparency, and compliance. With Modelscape, companies can ensure that their models are valid and accurate, and that when they're put into production, they are the same models that have been developed.

    They can also monitor those models to ensure that they continue to be valid and accurate. And they can do this for hundreds of models, each at their own stage of development. To learn more about how MathWorks Modelscape can help you manage model risk at your institution, visit mathworks.com/modelscape.

    [MUSIC PLAYING]