Expanding modelling ops for extending datasets

• Enterprise-wide stress-testing product of the year
• Regulatory capital calculation product of the year
• IFRS 9 enterprise solution of the year
• Consumer credit modelling software of the year

The Covid-19 market disruption put considerable stress on financial institutions as the 2020 pandemic introduced additional uncertainty around regulatory timelines, with many regulations being delayed or altered.

Risk models fell short, as typical data sources became unreliable for effective decision-making amid fast-changing market conditions. Models had to be scrapped for new ones, and firms with more advanced data extensibility features, and big iron cloud computing to power the new scenarios, were well positioned to withstand the demands on systems to deliver clear insights across the spectrum of risk modelling.

This was a year like no other, and it required a deep understanding of legacy systems, under strain from added volatility in the early days of the pandemic. But it also involved an innovative use of cutting-edge technology, such as machine learning, AI components and cloud-based computing services, to deliver scale as well as to understand where existing algorithms missed the onset of developing fraud threats that poured into banking systems as millions of customers shifted to online activities.

These are some of the reasons SAS was successful across four categories of the 2021 Risk Technology Awards: consumer credit modelling software of the year; enterprise-wide stress-testing product of the year; regulatory capital calculation product of the year, and IFRS 9 – enterprise solution of the year.

The interrelated categories, and SAS’s achievements in innovation, reflect a holistic view of the end-to-end risk modelling lifecycle that financial institutions realise is necessary to manage bottlenecks and inefficiencies. 

In particular, SAS was able to respond quickly to the pandemic with the launch of its Covid-19 Incident Command System for identifying practical use cases to
help customers mitigate health and economic impacts of Covid-19. Chief among the services was a free AI-driven environment to mine Covid-19 research, such as the Covid-19 (Coronavirus) Data Analytics Resource Hub.

Amid the market turmoil and lockdowns, the hub enabled customers to exchange innovation and better understand evolving credit-impairment scenarios. In the process, customers were able to assess their time horizons to modernise data infrastructure.

After all, the newer approaches to modelling require large, unbiased datasets to start model development and for ongoing model support to deliver more real-time engagement. This is a critical factor for financial institutions, which must manage IT budgets wisely to remain competitive in an economic environment where interest margins are at historic lows. 

Today, the innovations support complex mathematical models that can generate thousands of modelling scenarios of potential returns, says Martim Rocha, director and global head of risk banking solutions, and risk research and quantitative solutions for the SAS Institute. Interest rate scenarios provide key data points that reflect market trends to highlight potential equity returns, or the impact of inflation. 

The results stem from many years of building out a footprint in analytics, Rocha adds. “What we’re learning about AI and machine learning is that the technology is increasingly adaptable and provable in the way the business requires.”

But like SAS’s embrace of open source technologies, the key is how financial institutions govern those models. “AI and machine learning are new, and successful in areas,” says Rocha. “But all of these need to accommodate workflows before automation, which creates opportunities to better manage the models and change them over time.”

Risk platform transformation

SAS pioneered its Risk Stratum platform to host applications such as its enterprise stress-testing, regulatory capital and IFRS 9/CECL products, with the latter proving its value at a time when risk-weighted asset models were hard at work on reconciliations to reflect changed circumstances. For example, analysts can run potential stress tests and then look at distribution curves on modelling scenarios to see what long-term investments might yield under the new conditions.

The offerings address major pain points for banking customers, such as reducing backlog, adding standardisation, and common models and assumptions to drive analytics. Given the unique nature of the stimulus that accompanied the disruption to business post-pandemic, and the mess it made on model scenarios, flexibility and real-time analysis become even more critical.  

Firms found they got what they paid for. If they invested in data infrastructure to make it more extensible – reusable from a common data environment – they got through the volatility with better visibility as a result of fast, heavy-duty recalculations. 

This is a key tenet of the SAS strategy: a focus on the quality of architectural design of content, platform and platform components that leverage long-term assets and a roadmap for data modernisation.

SAS achieved enterprise-wide stress-testing product of the year award at a time when market disruptions put considerable stress on customers as pressure mounted on their cashflows and market volatility made a hash of their assumptions. 

Data transformation

The awards reflect the best of innovation, as well as offer a snapshot of financial institutions’ time horizon for upgrading legacy risk management systems. For example, data management is, in many areas, still deficient, such as with regulatory taxonomies, data mapping and data transformation – at a time when new data sources are flowing into systems to inform sentiment analysis that feeds machine-learning tools.

Any data mismatch is a problem when risk data needs to be reconciled with financial data before it is reported to regulators, and all necessary reports and schedules need to be updated and submitted appropriately. 

Among its responses to the challenges from the Covid-19 pandemic, SAS expanded its IP Solution Packages targeting Covid-19 risks across financial services and government, including a new monitoring IP to guard against account takeovers, business e-mail compromise and stimulus fraud. 

In addition, the risk team mobilised to create two use cases to help financial institutions navigate disruption and uncertainty. They include, for example, a risk modelling accelerator to help customers quickly run analyses and test models to gain a full picture of how market disruptions were impacting their business. 

Second, SAS built a Scenario Impact Simulator to build out forecasting outcomes of various stress scenarios, with interactive analysis delivered in risk dashboards. 

The decisions banks and lenders made with these tools helped determine how well they survived the crisis and how strongly they recovered. 

New dynamics for consumer credit modelling 

Already undergoing a step change due to factors such as digitalisation and the adoption of next-generation technologies, as well as extra regulatory demands, banks must be even more agile in providing customers with a consistent and contextual experience across the credit lifecycle. It involves larger datasets that must be wrangled to a much greater degree, on the fly, using dynamic risk modelling and predictive analytics that rely on more sources of data to complete a robust view of the credit profile.

One example involves a digital bank that specialises in consumer credit and financing of small and mid-sized businesses throughout Italy. The bank relies on predictive analytics and a cloud-first approach to mitigate risk, better serve clients and plan for the future, all of which came under pressure as millions of workers around the globe shifted to online work and more digital transactions.

SAS’s integrated environment supports the end-to-end credit lifecycle while meeting all model governance requirements: from regulatory capital and provisioning models, such as through-the-cycle or forward-looking probability of default, loss-given default and exposure-at-default estimation. It includes application and behavioural scoring, to models fulfilling auxiliary functions, such as income estimation, prepayment and propensity models. 

Again, data management tools in this realm must be capable of handling enterprise volumes in a number of formats. The models must be presented to decision-makers in a variety of methods, including programmatic and graphical interfaces. This is behind SAS’s approach to a single unified platform, providing integrated governance of data and models, auditability and lineage.

Open and flexible platform

The SAS Risk Solutions platform accommodates a wide range of analytics skills and techniques, graphical or programmatic user interfaces, SAS and open source. In addition, the solutions are white box: fully open to be tailored to unique customer requirements. The distribution and dissemination of results can be done via APIs, business rules, user interfaces and in-memory, in-stream, in-database, docker containers or mobile. 

The techniques include a mix of automated machine learning, workflow management to automate model validation, and automated hyper parameter tuning to deliver an expansion of ModelOps to operationalise a larger range of models. 

Cloud-based offerings are built to deliver dynamic on-demand scalability and flexibility, delivered in partnership with Microsoft Azure cloud services and deployed with Kubernetes. It offers connectors for cloud data sources, a conduit for customers leveraging their risk management architecture.

Today, the products are delivered in extensive implementations in multinational banks around the world with myriad SAS user communities sharing knowledge. So, while clients can take advantage of pre-configured systems out of the box to comply with regulations around the globe, the flexible platform – its investment in open-source for development, for example – provides the customer with options based on their own configuration needs and risk management objectives.

New horizons

In the past year, SAS has started the transition of its risk management solutions to its state-of-the-art Viya platform, on which much of its credit risk modelling capability relies.

New features include scorecard development capability, plus interactive and optimal binning, scorecard development, and reject inferencing. Also, new tools address back-testing and benchmarking with regulatory compliant model performance reporting. It also enables embedding rules and analytics in decision flows to drive decisions at scale. 

The judges said:

Enterprise-wide stress-testing product of the year:

• “Well-designed architecture and full orchestration of the production cycle. Brings together a lot of rich functionality and capabilities.”  
• “Proactive and continuous innovation.” 
• “List of supported applications is very comprehensive – and clients like it.” 

Regulatory capital calculation product of the year:

• “Way ahead on next round of Basel regulations and frameworks with innovative approach for the big four in clients’ needs: compliance, performance, governance and flexibility.” 
• “Differentiator with SAS is quality of architectural design of content, platform and platform components.”

IFRS 9 enterprise solution of the year:

• “Very comprehensive, multi-component, multi-disciplinary, multi-department solution. Convincing integration of risk, finance, data and technology. Feature and benefit rich. No mean achievement.”  
• “Impressively flexible, without becoming incoherent. Lots of dev work, with appropriate focus on reporting and visualisation.”  

Consumer credit modelling software of the year:

• “Excellent product description and client testimonial. Good proactive innovation.”  
• “Long list of additions/refinements. Genuine focus on customer needs.”  
• “SAS continues to innovate and expand its product suite. Impressive.”  

Sponsored content