Regulatory expectations for parallel testing
Supervisory Guidance on Model Risk Management (SR Letter 11-7) describes the key aspects of effective model risk management and sets regulatory expectations. The guidance does not explicitly mandate parallel testing when models are implemented, but the expectation is clear. When institutions modify or install new models to reflect new data techniques or performance concerns, regulators expect meaningful evidence that the changes improve results.
The guidance highlights parallel testing, or parallel outcomes analysis, as an essential approach for identifying gaps in new models. If the new or adjusted model does not demonstrate stronger performance, the institution should recognize that further refinement may be necessary before replacing the original model. In practice, this reinforces the case for running models in parallel, supporting sound model governance and defensible decision-making. It demonstrates that financial institutions aren’t changing models for change’s sake.
The guidance also addresses documentation requirements. Clear records of testing scope, issue resolution, and management approval not only help examiners understand what changed but also provide proof of continuity and control.
Beyond compliance: Testing’s operational benefits
While compliance is often the initial driver, parallel testing delivers operational value across all pillars of the institution. It gives teams time to learn new workflows, identify training needs, and fine-tune processes before the pressure of full adoption.
It also creates space for informed decision-making. Differences between systems can reveal data quality issues, process inconsistencies, or risk assumptions that might otherwise have gone unnoticed. Addressing those findings strengthens the overall program, not just the new technology.
Most importantly, parallel testing protects customers, members, and communities. Whether the institution is monitoring transactions, underwriting loans, or managing portfolio risk, accurate systems support fair, consistent, and timely outcomes.
