Predictive Algorithms Can Mitigate the Risk of NPAs For Banks
Monitoring the actual cash flow of the business through prudent practices is the key to move to cash flow-based lending
The Indian banking system has long been burdened with the problem of bad loans. Since 2014-15, Indian banks have written off INR 5.7 trillion worth of bad loans. According to data released by the finance ministry, India's 42 scheduled commercial banks (SCBs) collectively wrote off bad loans worth INR 2.12 trillion in 2018-19, up from INR 1.5 trillion in the previous year. This accounted for 20 per cent of all their non-performing assets (NPAs). When it comes to the country's public-sector banks (PSBs), they wrote off bad loans worth INR 1.9 trillion in 2018-19. The exponential rise in write-offs has occurred simultaneously with an increase in NPAs. Government statistics reveal that the NPA of PSBs stood at INR 7.27 trillion as of September 30, 2019.
While the problem of bad debt from large corporates has mainly arisen out of fraudulent activities (for instance IL&FS, Jaypee Infratech, Jaiprakash Associates, etc.), the NPAs at MSMEs and smaller businesses have been more driven by poor monitoring and understanding of ground realities for such entities. The COVID-19 pandemic has further worsened this crisis despite a series of regulatory measures. Rating agency Fitch said Indian banks will continue to face significant asset-quality challenges for the next couple of years. Against this backdrop, banks, NFBCs, and even peer-to-peer lending platforms are moving away from archaic practices and poor due-diligence methodologies to technology-enabled, data-driven credit rating systems. Data intelligence firms are playing a crucial role in bringing this shift by offering their solutions to banks and other financial institutes.
Predictive algorithms for risk monitoring
Monitoring the actual cash flow of the business through prudent practices is the key to move to cash flow-based lending. It is, therefore, imperative to capture a business's digital footprint, especially at a time like this when meeting the business owners in person and physically visiting their areas of operation is not a feasible option. Banks need to bring "singularity' to all datasets pertaining to a business, including financial, non-financial, and alternate datasets. With predictive algorithms, this data can supplement and add value to the credit decision-making process. However, traditional banks have not managed to marry financial data with non-financial data to build a robust predictive algorithm model. This is where technology firms come into the picture. They harness the power of machine learning (ML) to generate early warning signals, thereby enabling financial institutes to lower the risk of bad loans.
Removing human bias from scorecards
There have been several instances where lenders bypassed scorecards and credit assessment models to disburse loans because they "knew' the business well. Human bias has always been a deterrent to prudent underwriting practices, which eventually adds to the bad loan crisis. However, this practice is now changing. Automated data-driven scorecards built on ML models that are customized for Indian businesses have been deployed at many public and private sector banks for speedier and prudent underwriting. While credit rating agencies claim to provide such credit reports for all businesses, they still rely on manual processes. This again defeats the purpose of a data-driven underwriting process. Data intelligence can only come in when there is an automated approach to build the scorecard, eliminating any scope for human bias. New-age data analytics firms have not only recognized this gap but also introduced digitals scorecards for businesses that are underpinned by ML algorithms to help lenders make accurate credit decisions.
Automated UBO mapping
In developed economies such as the US and the UK, where traditional businesses have evolved over the decades led by a few business owners, UBO analysis has always been an area of focus. A high tax regime and need to reduce the owner's liability in each entity has allowed several related entities to be formed across the network. In India, such tools have never been deployed except only recently. Leading public sector banks are now utilizing UBO network charts to identify related parties, ultimate beneficiaries, and ownership matrices. The need of the hour is to do this on a proactive basis for large key accounts with automated BPO mapping tools.
As the Indian economy attempts to get back on to its feet, financial institutes are evaluating how they can underwrite and onboard new customers quickly. For businesses, the need to pre-populate public data during the onboarding and KYC process itself is being looked into. However, waiting for the customer to submit financial statements and audited reports, which can be pulled directly from regulatory sources, is a sheer waste of time and resources. Building such capabilities directly into the KYC and on-boarding journey through advanced technologies can facilitate faster onboarding.
The GST regime in India in the last two-three years has brought about transparency, accountability, and more importantly, an additional financial dataset for the lender to evaluate. With user consent, this data can present a much more insightful look into the business at the time of the underwriting or refinancing process. While there is a need to make certain sections of this data public, the source in itself is a goldmine for credit assessment.
Capitalizing on business data
There are large volumes of transaction and business operations data sitting on any firm's ERP systems. The question is—how does one tap into such a data source to build genuine trust in good businesses? The solution is data lake deployments and public-private data partnership models, which can help unlock the true potential of such resources, allowing for a more detailed and accurate credit assessment process. This, in turn, will help lenders make informed credit decisions and avoid the trap of bad loans.