By Ajay Katara, Domain Consultant in Banking Risk Management area at Tata Consultancy Services

Digital Transformation continues to lead transformation across all facets of Banking, advancement in Data and Analytics have opened new doors of enhancement for Credit Decisioning Models as well. Credit Decisioning Models form a core function within the lending process, however during the pandemic it was observed that new credit decision models performed well over the traditional credit decision models in terms providing a sharper focus in identifying credit worthy customers, quicker automated decisioning, increased revenue and reduced loss rates for the Banks.

Hence many Banks aspire to move towards advanced Credit Decision Models but are saddled with Legacy challenges like access to limited data sources, minimalistic Analytical engines, technology limitations, subjective assessments from underwriters, transitionary period to advanced models and regulatory acceptance of models. Such challenges make the banks ponder over the efforts involved in moving towards advanced models. However, the benefits of transitioning cannot be weighed down too. Many Banks who have advanced to Newer or Advanced Credit Decision models have realized the following benefits

1. Reduced Loss Rates – Likelihood of Default is a key parameter based on which the Loan is priced and provided, Advanced and Automated Credit Decision Models tap can process data from multiple sources. Apart from the traditional data like customer financials and historical behavior, advanced models also tap into data from secondary sources like social media. Machine learning based models can self-learn over a period based on the historical data and provide forward looking assessments on customers likelihood of default which can help Banks to arrive at a more accurate decisioning thereby bringing down the loss rates involved and also the regulatory capital provisions required.
2. Operational Efficiency gains – Automated Credit Decisioning systems significantly reduces turnaround times in the approval process of Loan applications by automating the manual tasks involved like gathering customer information and financials required for decisioning and makes the process more objective, traceable, and transparent. Technology can help automating decisioning and processing for different types of loans for e.g., for High quality loans it may help by automating the decisioning and medium to low quality loans can provide scoring details quickly which can enable the Loan Sanctioning team to consider those inputs as starting point for further analysis.
3. Consistent Underwriting standards – The Loan policy of the Bank is not always very specific with respect to loan decision factors and is subject to interpretation by underwriters who make operate from different regions and loan segments .Automated and Digitized Credit Decision models imbibe the organization policy and ensure that consistent underwriting standards are followed ,a customized loan-decision model ensures factors like loan type or size, credit scores are consistently applied, providing an automated approval, rejection or further underwriting analysis.
4. Increases Revenue – Automated Intelligent Decision Models help Bank to quickly assess and identify Credit worthy loans and can customize their acceptance rates and pricing. An automated decisioning process also reduces the cost of acquisition of customers and bypass the manual processes involved.

Banks will look to invest and uplift into the new Age models to bring in the required scalability and efficiencies. Some of the key factors that they should consider while investing in futuristic models are outlined below

• Modular Design is the way ahead – Modular design makes it flexible for Banks to include or exclude models which provide different dimensions before arriving at a decision .With a Modular design Banks can have more flexibility to integrate newer data sets .Typically each data set feeds in to a sub model which is then aggregated in to a final score .In adverse scenarios like pandemic or Recession newer sub models can be introduced which can provide more contemporary inputs to the final decision score. A potent modular design architecture depends on active participation from Business, Data and Modelling team.
• Data fuels the Model – Banks need to expand data sources which fuel its decision models, apart from looking at data from Internal sources like customer financials, historical behaviors etc., external data sets like credit bureau data, social media behavior, Fraud detection services should also be added into the fray to arrive at decision scores. Some Machine learning based models also tap into previous decisions which are taken by Relationship managers to understand their impact on decision scores for current prospects.
• Data Mining for Signals – Advent of Machine Learning and Artificial Intelligence have made it possible to draw inferences from the existing data sets, these algorithms can create synthetic cashflows for customers and provide a forecast on customers income and expenses. These models can also identify region specific factors that can impact customers credit worthiness. Missed or delayed payments on obligations like credit card payments etc. also feed into the models and can provide signals around future payment patterns of the customer. Some banks even create challenger models where additional information and scenarios are fed in to validate the model performance.
• Business Insights – Though technology forms a critical lever in executing the expectations around model development, Business insights also are very critical while designing decisioning models .A prudent approach involves taking insights from existing underwriters and relationship managers who can provide instances from real time experiences which provide an additional dimension to the Credit decision models

The past decade has seen multiple events from economic meltdown to the pandemic much recently and it is very clear that going forward the focus for the Banks will be investing in credit decision models that are quick to adopt and leverage digital technologies to scale as per the current dynamics. Most Banks are adopting Machine learning based models, which fare better than the traditional models but come saddled with some shortcomings in terms of sophistication and stability, these models are beginning of a journey which banks need to undertake to reach a more responsible and a reliant Digitized automated decision solution. As they say future is uncertain and change is the only constant which can help Global Banks and Financial institution cope better through the changing times.

The views and opinions expressed in this article belong solely to the authors and do not represent those of the authors’ employer organization.

About the Author

Ajay Katara is a Domain Consultant in Banking Risk Management area at Tata Consultancy Services (TCS). He has extensive experience of more than 15 years in Consulting & Solution design space cutting across CCAR Consulting, AML, Basel II implementation and Credit risk, and has worked with several financial enterprises across geographies. He has significantly contributed to the conceptualization of strategic offerings in the risk management space and has been instrumental in successfully driving various consulting engagements. He has also authored many editorials, details of which can be found in his linked in Profile. (https://www.linkedin.com/in/ajaykatara/)