Practitioner course in risk management
Model Risk Management
How financial institutions control the risk of using models: what counts as a model, inventories and tiering, independent validation, conceptual soundness, benchmarking and outcomes analysis, ongoing monitoring, and the governance expected by supervisors.
Every price, risk number and credit decision a financial institution produces depends on models, and every model can be wrong or used wrongly. Model risk management is the discipline of finding and controlling those errors. Since US supervisors issued their SR 11-7 guidance in 2011, it has become a function of its own in banks and a growing concern for asset managers and trading firms.
What does a model validator actually do?
A validator checks three things independently of the people who built the model. Is the theory and design sound for its purpose? Is it implemented correctly, which often means rebuilding it? And do its outputs match reality, tested by benchmarking against alternative models and by comparing predictions with outcomes? The findings go into a report graded by severity.
How much of a model’s risk comes from how it is used?
Often more than from the model itself. A model built for one portfolio or market regime is applied to another, or its limitations are forgotten as it is passed between teams. The course treats use and governance as part of the risk, not an afterthought.
Who is this course for?
Risk analysts, quants moving into validation, and model developers who want to know what validators look for. It counts toward the Certificate in Quantitative Risk Management.
What you will be able to do
- Define model risk and identify its main sources.
- Set up a model inventory with risk tiering.
- Plan and carry out a validation covering soundness, implementation and outcomes.
- Design ongoing monitoring with thresholds and escalation.
- Write a validation report with findings graded by severity.
Syllabus, module by module
What model risk is
Fundamental errors, implementation errors and misuse; well-known model failures.
Supervisory guidance
The US SR 11-7 guidance and comparable expectations elsewhere, in outline.
Inventory and tiering
Defining a model, recording it, and tiering by materiality and complexity.
Validation
Conceptual soundness, replication, benchmarking, sensitivity and outcomes analysis.
Ongoing monitoring
Performance metrics, drift, thresholds and when to revalidate.
Machine learning models
Explainability, data drift and validating models without closed forms.
Governance and reporting
Roles, independence, findings, remediation and reporting to the board.
How the certificate is earned
An assessed project: you validate a supplied pricing or risk model, including replication, benchmarking and outcomes analysis, and submit your code and a validation report. On a pass you receive the Course certificate in Model Risk Management, issued by the Institute with a verification code. Read how certification works.
Before you start
Working Python, statistics to regression level, and familiarity with at least one pricing or risk model.
This course counts toward
Questions applicants ask
What is SR 11-7?
Supervisory guidance on model risk management issued by the US Federal Reserve and the Office of the Comptroller of the Currency in 2011. It defines a model broadly and sets expectations for development, validation and governance, and is widely used as a reference outside the US.
Does the course cover validating machine learning models?
Yes, as one unit. It covers what changes when a model has no closed form, including explainability methods and monitoring for data drift.