Model Management in 2024: What You Need to Know to Prepare for E-23 Guidelines

July 1, 2025 marks the scheduled implementation date for the forthcoming regulatory adjustments. Since its launch in September 2017, the Office of the Superintendent of Financial Institutions (OSFI) has embarked on a commendable journey towards instating standardized protocols within the financial industry. These initiatives have, over time, contributed to a heightened degree of transparency and fortified controls over crucial models.

The updated Guideline E-23 constitutes a compendium of best practices that federally regulated deposit-taking institutions, encompassing banks, insurance companies, and pension plans, are obligated to adhere to, effective from July 2025. The comprehensive particulars of the definitive guidelines are anticipated for release in December 2023. Arguably, this timeline affords organizations a runway to position themselves for success. It is worth noting that the market currently offers robust out-of-the-box platforms that not only align with but surpass the impending regulation.

Canada has carved a global reputation for its pioneering role in shaping financial regulations – whether pertaining to BASEL, IFRS 9/17, or B-20 – Canada has consistently championed the cause of prudent risk management, a stance that has stood us in good stead over the past decades. However, notwithstanding our sound approach to risk management, the advent of Big Data, the upsurge in Digital Fraud accelerated by COVID, the proliferation of AI/ML technologies, and the growing reliance on third-party data providers have ushered in new imperatives and challenges for financial institutions.

 Reviewing your third-party data

 In the aforementioned contexts, data emerges as the linchpin. Nevertheless, the effective handling of data necessitates a substantial degree of oversight. From data lineage to data monitoring, given the sheer volume of data, modern organizations necessitate a robust array of tools to aptly harness it. For instance, these tools may be instrumental in generating lending decisions.

Speaking of lending, third-party data has become almost indispensable for any institution involved in credit issuance. While third-party data has time and again demonstrated its value, a lender may often find themselves in the dark regarding the quality, predictive power, completeness, staleness, and other attributes of the data consumed.

Adding to the complexity of data management is the fact that we have transitioned into a new economic era post-COVID. This new economic chapter has ushered in an altered landscape, characterized by macro-economic factors that affect us globally and micro-economic drivers that shape our credit behaviors. Notably, factors such as inflation and a higher-interest environment have risen to the forefront of economic considerations…all reflected in recent consumer data.

 Enhancing models

Data, being an indispensable component of any model, inevitably manifests itself as either a sound or suboptimal decision by a lender or an insurer, often through automated decisions generated by models or scorecards. Now, having in mind that we are transitioning to a new economic chapter, it becomes imperative to ascertain the continued relevance, accuracy, representativeness, and reliability of third-party data feeding into these models, which are employed daily for purposes such as adjudication, pricing, limit assignment, fraud detection, propensity analysis, and more.

How frequently do financial institutions employ scores and other 'super variables' that remain shrouded in obscurity? Conversely, how frequently do third-party providers unveil the logic of their data collection, score-building, and validation processes with full transparency?

Numerous questions arise, and reassuringly, there are corresponding answers. Although financial institutions on the receiving end may not possess complete visibility into these aspects, mechanisms have been established for automated data and model monitoring to serve as stewards and as early warning indicators, forestalling potential disruptions.

Selecting the right tools 

Clarity, oversight, accountability, monitoring, and contingency planning – are all words featured in the Proposed Revisions to Guideline E-23, as published on May 20, 2023. These elements, underpinning the forthcoming enterprise-level regulations, extend their relevance to the realms of Risk, Fraud, Marketing, and beyond, offering valuable guidance on the essential toolkit that financial institutions should be equipped with to navigate the impending challenges successfully.

The financial industry has come a long way. A typical financial institution now manages hundreds, if not thousands, of models in its repertoire. Thus, the effective management of this ever-expanding and evolving inventory stands as a key requirement. Automation, governance, native integration and ease of use are valuable allies in this endeavor.

Impact on insurance

Up until this point, there have been no specific regulations related to model risk impacting the Canadian insurance industry. To understand why, one needs to understand the role of the Actuary within the context of model risk. The result of OSFI’s E-23 — which brings the first instance of top-down model risk regulation to Canadian insurers - can thus go one of several ways. For insurers that have already applied ASOP-56 best practices to their entire inventory of models — this may simply be an exercise of adjusting nomenclature to better reflect the new E-23 regulations.

For others, current model risk programs may only need to be expanded to reflect formally out-of-scope or non-actuarial models. For insurers that have thus far only applied bare minimum model risk controls or allowed individual model owners to have inconsistent standards up until today – the new standards will represent a significant shift from the norm and will require heavy investment of time and resources to bring their model risk management programs up to formal compliance.