R&C: In what ways have financial institutions (FIs) been making greater use of models as part of their decision making in recent years? What benefits can this process yield?

Asermely: Financial organisations understand having better data is a competitive advantage. In years past, an analyst likely made a go/no go decision on a loan. Today, that decision is made by a model. Even the most skilled analyst is susceptible to bias and human error. Computational finance models, however, allow us to utilise data more effectively to make unbiased decisions that are intuitive, repeatable and transparent. A proper modelling approach puts these decisions through an intensive review process to improve the model’s quality.

R&C: What challenges face FIs when using models? Are they struggling to govern models that are increasingly complex, non-intuitive, unstructured and difficult to categorise?

Asermely: As the number and turnover of models increase exponentially, financial institutions are struggling to manage and care for their models properly. Financial instructions are also shifting to more complex ‘black box’ machine learning (ML) and artificial intelligence (AI) models. Many of these institutions are unable to answer simple questions like: Do we know who is using which models? Are we using our models for their designed use cases? Are we supporting outdated models we no longer use? Historically SharePoint, spreadsheets and generic operational risk tools were sufficient to manage models. In today’s fast-moving, increasingly complex and competitive landscape, however, these approaches are becoming increasingly costly and will not scale to meet the evolving needs of the firm. In addition, their complexity makes it difficult to understand how models are interconnected and can potentially result in snowballing risks. Financial institutions should consider models to be high-value assets that require purpose-built model risk management tools, ongoing testing and governance.

Oct-Dec 2018 Issue