Editorial

Does the lack of a uniform approach across model monitoring and model validation increase costs and extend validation times?

In the last few years financial institutions have made good progress in managing model risk. ‍Model risk is now considered to be a key risk in the same way Financial and Operational risks are. In the last twelve years since the inaugural regulatory guidance (SR11-7) came out and yet the management of model risk varies significantly across financial institutions.

Contributor

George has more than 22 years of Financial Services experience, specialising in Trading and Risk Management.

George Petropoulos
Chief Product Officer, Pricing & Risk

In the last few years financial institutions have made good progress in managing model risk.

Model risk is now considered to be a key risk in the same way Financial and Operational risks are. In the last twelve years since the inaugural regulatory guidance (SR11-7) came out and yet the management of model risk varies significantly across financial institutions.  

The use of models to make informed decisions ranging from judging customer behaviour to pricing a derivative product or even deciding whether a transaction is fraudulent is increasing and that exerts more pressure to model risk management departments and in turn creates bottlenecks for product development.  

Among other suggestions, recent regulatory guidance (SS1/23) introduced a new model definition that will almost certainly expand model inventories. The approach taken so far by most financial institutions to rely on a model inventory with limited features plus hiring relatively expensive resources hasn’t yielded the expected results. Most financial institutions are arguably quite far from managing model risk using a holistic approach.

A holistic approach to model monitoring

In most financial institutions, certain key parts of the model lifecycle are manual and require a lot of data scraping. Model monitoring is one such area that needs improvement to achieve operational efficiency. The technology used is fragmented and requires investment as often new applications are required to be built.

Drawing examples from the derivatives pricing model’s area there is little evidence to suggest that vendors of pricing and risk management systems can offer to their customers APIs (Application Programming Interfaces) to run model monitoring at scale across all traded assets. There is also little evidence to suggest that financial institutions are ready to undertake this activity due to the design of their architecture and lack of available tooling.

In addition to the technology required to perform model monitoring, designing, and writing the code used for monitoring is time consuming. This is not only due to the variability of appropriate tests but also because the analytics often do not have such features i.e., testing at scale.

Designing a set of standard sets and deploying in an environment accessible by both model developers and model validators would increase operational efficiency instantly. As a by-product of that efficiency gain technical model documentation can be streamlined and even automated and easily reproduced.

To arrive at this level of automation and operational efficiency financial institutions will have to invest in solutions capable of handling the flows of the model lifecycle and extending their analytics so that model monitoring can be unified and automated.

Improving time to model validation and recurring validation

Model validation as an activity has been taking place long before the publication of SR11-7. The key difference was that the validators were not asked to approve the models but only to provide an independent opinion. Since this activity was formalised and moved to the second line of defence model validation is taking longer and consumes more financial resources from financial institutions.

The ongoing race that ensued to hire the brightest and most experienced model validators has caused capacity constraints within model validation departments thus creating very long validation pipelines.

In many financial institutions model validation is now viewed as a bottleneck in the model lifecycle and has increased the time to market for new products. Furthermore, there is little difference between the time to re-validated existing approved model’s vs the time to validate new models. As a result, other related processes are affected, and businesses are assessing their investments in this area.

Without a close collaboration between model owners and model validators this problem will persist, and it will only consume more financial resources as staff sourcing and retention remains a struggle. Additionally investing in a solution that can increase collaboration across the different lines of defence, promote standardisation and automation would yield tangible benefits such as:

  • Flexible model inventory and model lifecycle workflow
  • Reduced time to validation and re-validation
  • Automated model documentation


These groupings represent different approaches and stages of maturity in the technology found within financial institutions including:

To find our more, contact us today and speak to one of our experts.