top of page


Join our mailing list

Never miss an update

April 01, 2018

Artificial Intelligence risk

The increasing reliance on Artificial Intelligence across the financial services industry introduces additional risk stemming from the fact that the technology is complex and not well understood by users and risk managers alike. Adaptation of model risk management techniques and procedures to the unique challenges associated with model that are based on Artificial Intelligence is required. 

March 01, 2018

How to assess the maturity of a Model Risk Management program?

An effective, mature Model Risk Management program can contribute to the bank's bottom line by avoiding losses, reducing costs, and improving efficiency of capital planning. To assess the maturity of MRM practices within banks we evaluate the program maturity by focusing on nine characteristics of such programs. 

February 01, 2018

The third line of defense

The third line of defense includes internal and external auditors. In the model risk context, the third line of defense validates the risk assessments conducted by the second line of defense and provides senior management with independent assessment of the organization and effectiveness of the second line of defense. 

January 01, 2018

Validating the output component of a model

The output component of a model is concerned with translating model results into useful actionable, information. An assessment of the Output, or Reporting, component requires an evaluation of three key elements: model performance, model stability, and model robustness.

December 01, 2017

On the validation of expert-based models

Model developers often rely on experts for different tasks within the modeling process ranging from conceptualizing and designing the model to the determination of parameter values. Sometimes, experts are used to adjust the output of a model, or even as a replacement for a formal model, when they are used to generate a model overlay. Consequently, model validators often encounter models that incorporate expert opinions or expert judgment.

November 01, 2017

Validating the processing component of a model

The processing component of a model is concerned with transforming inputs into meaningful reports. The wide range of methodologies that are used include quantitative and qualitative techniques. However, the same key elements need to be assessed for any given methodology as part of the validation of the model.

October 01, 2017

On the validation of model inputs

Even a theoretically sound model that is computationally accurate will produce inaccurate output if fed the wrong input.  The validation of model inputs includes an examination of the definition and appropriateness of the data inputs, data extraction, data checks, data assumptions, and data manipulations.

October 01, 2017

Data preparation: the role of model developer

A key component in any model, statistical or otherwise, is the data used as input for the model. Data preparation includes three distinct tasks: extraction, cleansing, and setup. Each of these tasks is critical as it can affect the accuracy of the data and thus the reliability of the model’s results. Data preparation is the responsibility of Model Developer.

September 01, 2017

A framework for model validation

The key to mitigating model risk is model validation. Systematic model validation entails a comprehensive review of all the key components of the model vis-à-vis its business purpose: the data, the theory links key constructs, the practical implementation of the theory, and the documentation and reports that communicate the model’s processing and results.

September 01, 2017

The second line of defense

The second line of defense includes professional risk managers responsible for developing and applying assessment methodologies, risk standards and practices. In the context of model risk, a key role of the second line of defense is model validation – an independent review process aimed at limiting model risk.

August 01, 2017

Model development and the first line of defense

The model development process is inherently subjective and requires making numerous choices at multiple stages of the process at multiple levels within the institution. Therefore, a critical component of reducing model risk is implementing a disciplined, sound, and comprehensive model development process. Such a process would constitute a robust first line of defense in the governance and management of model risk.

August 01, 2017

The three lines of defense model

The most common governance model within financial institutions today is the "three lines of defense" model. This model assigns risk management responsibilities to three mutually exclusive groups within financial institutions: business units, a centralized risk management group, and internal auditors.

July 01, 2017

Risk-based tiering of models

The model inventory can potentially include a very large number of models. For governance and prioritization purposes, models are typically assigned tiers. A common practice in financial institutions is to assign tiers based on the risk the model poses. A model’s tier classification determines the scope, intensity, and frequency of all applicable model risk management activities – models with a higher model risk will receive a higher degree of oversight and governance.

July 01, 2017

Model inventory management

At the heart of any model risk management program is the model inventory. The goal of the centralized model inventory is to enable the assessment of model risk across the institution and over time vis-à-vis the institution’s risk appetite. For this purpose, it is essential that the model inventory will include relevant, up-to-date, and accessible information about all models used in the institution.

June 01, 2017

Classifying vendor models

The "black box" nature of vendor models poses substantial challenges to the assessment and mitigation of model risk. Sometimes it is difficult to determine whether they should even be considered models. 

June 01, 2017

But, is it a model?

What exactly comprises a model? This is the most fundamental question for any model risk management program. Depending on the definition of a model, any such program identifies models throughout the organization and lists them in a centralized model inventory. 

Please reload

bottom of page