How To Avoid Drowning in the Ocean of Model Risk Management

The number of models is continuously increasing, and risk management teams can’t grow fast enough. To top it off, model budgets are shrinking instead of growing with the demand.

Model governance regulators are moving their focus closer and closer to the qualitative elements of a model – the model governance process. Unfortunately, the regulator also often asks the bank to explain their internal risk mitigation process without the providing the specific guidelines necessary to match the model to the bank’s methodology. 

Banks often develop their model governance framework to comply with SR 11-7, the US regulatory model governance standard. While this is good, we think it’s time to bring model inventory back to the users and provide easier usability through better portfolio overview and agility. Modern model governance framework should be a robust, easy-to-use tool that will give your bank a competitive advantage. 

At CloseIT, we help customers address the topics of inventories, storing risk parameters, team cooperation, and automation. We’ve used our collective experience to develop and introduce our own model governance product, the Model Governance Suite (MGS).

How should you start with MGS?

Let’s first assume that the MGS’s default configuration of model types and their attributes doesn’t entirely match the bank’s methodology. Our team then helps identify all model types, not just risk credit models, that are relevant and registered under the firm’s model risk management framework. 

Following that, we adjust the list of collected attributes and adjust or define workflows and specific business rules to your specifications. The next step lets model owners register all existing and planned models into the system.

Linked models

Something that the “Excel repository” can’t provide is an instant overview of the dependencies within the entire model portfolio. Models can be connected by configurable link types in order to define dependency graphs at various levels. A typical graph represents models providing input to other models - eg. an operational scorecard providing a rating to IRB models, providing input to ALM models. 

In Model Governance Suite it takes only one click to open an ALM model and see all the input models from all input layers, including their unresolved issues, and review potential model cascade risk all on one page.

Automation - only when auditable and repeatable 

There are plenty of tasks done by model risk management experts which are rather repetitive, mechanical tasks, but still consume lot of time. For example, execution of model monitoring function on regular validation data samples.

MGS is a methodology as well as a technology agnostic tool. So, define your own model monitoring function in python or R and MGS takes care of regular execution, interpreting results, and taking corresponding actions such as assigning tasks and sending warning emails. All that is strictly audited and even reproduceable - if audit asks.

Model Risk Management - automated risk scoring

MGS lets you define cross-model attributes of “risk score” and “importance”. These attributes will be regularly adjusted by a defined algorithm considering qualitative (model attributes) or quantitative data. 

Of course, in its early implementation stages, the risk score and importance can only be updated manually, but the automation can be added step by step to unload manual work from firm’s MRM experts.

Every active model is, on a daily basis, evaluated through a customisable decision tree - send email notifications, contact other systems, or create user warnings. One example of the MRM dashboard targeted to a senior management’s iPad is this report similar to following a two-dimensional projection of the model risk including trends.

Your own business rules adjustable on the fly

Define a combination of attributes which isn’t relevant or assign a DQ check to particular validation data sample all using the MGS embedded business rules.

… the cornerstone of your model risk management framework is here.

Is it just a convenience or is it a necessity? 

Our new generation of a model governance tool can simplify or automate many of the daily tasks of the model risk management team. With an increasing number of models and requirements creating more work for same amount of people, the CROs are facing a difficult decision. Does the bank want:

• the old excel inventory,

• inventory fulfilling basic regulatory needs, OR

• new generation model governance tool that will quickly return the initial investments?