Hi
I was reading Alex Johnson’s (fellow fintech nerd and ex-director of fintech research at Cornerstone Advisors) newsletter, where he hypothesises a world of continuous underwriting in lending.
It got me thinking about a few things -
What utopia looks like for lenders and borrowers?
What’s the ideal world where both lenders and borrowers could meet halfway?
Whether this ideal world has room for continuous underwriting?
What does this hypothetical world of continuous underwriting mean for risk-based pricing?
What public digital goods would we need to make it happen?
What does this mean for borrowers?
What would a borrower’s utopia look like?
Interest rates are almost non-existent, with lenders even paying you to take out a loan because you're awesome. The lending process is as easy as ordering pizza and requires no paperwork or credit checks. Lenders act as fairy godmothers, making your dreams come true with funds for your business or a castle in the south of France. Borrowing and lending are fun social activities, like a party, and are built on trust and mutual support.
What would a lender’s utopia look like?
Borrowers pay up on time, every time - no ifs, ands, or buts. Credit scores are flawless, never a gap or mistake. Loan applications breeze through instantly, with no human touch required. Interest rates are sky-high, and profits flow like a river. The lender has bottomless pockets and no capital or liquidity worries. No competition in sight; the lender rules the roost. Borrowers have nowhere else to turn, fully dependent on the lender. Fraud and default are non-existent, and all loans are repaid in full. The lender has all the intel, making decisions with perfect knowledge.
What’s the ideal world where borrowers and lenders can meet?
In an ideal world of credit, lenders and borrowers would have access to a real-time credit bureau that constantly updates information about their financial behaviour. This system would enable continuous underwriting of consumers for credit based on real-time changes in their cash flow and repayment behaviour. Borrowers would no longer have to go through a lengthy and cumbersome application process to get approved for credit. Instead, they would be continuously and passively underwritten, making the borrowing process seamless and convenient. In addition, interest rates and lending terms would be fair, transparent and personalised, with borrowers having access to multiple lenders, creating competition that drives down interest rates. This would create a virtuous cycle of lending, where the benefits of borrowing are shared by all members of society. Overall, this ideal world of credit would promote financial inclusion, transparency, and fairness, benefiting both borrowers and lenders alike.
Sounds too good to be true? Definitely.
We’ve already postulated what the ideal world would look like. Clearly, the borrower’s ideal world is as different from that of a lender’s as chalk and cheese. Notwithstanding, neither of these utopian visions will ever come to fruition in this messy, complex world ruled by multi-dimensional game theory.
But it’s still worth evaluating whether continuous underwriting can make at least some of it happen - perhaps not by putting lenders and borrowers at loggerheads but by harmonising their respective goals into a more efficient framework.
What is continuous underwriting?
The evolution towards continuous underwriting is made possible by increased data and device connectivity. That means more opportunities for personalisation. Currently, traditional underwriting models suffer from three primary data gaps. First, it is constrained to a single moment in time - it measures parameters as on date with a limited view into the past or the future from that point in time. Usually, credit scores ( updated every 45 days) and maybe some behavioural data are the source of data. Second, some data points are subjectively derived - often at an industry or geographic level - and that may not be precise enough for underwriting. Third, it fails to capture any large lifestyle, social or financial changes for the borrowers.
Think of continuous underwriting as an infinity loop -an unending flow of information to lenders that feeds into crafting personalised offers for borrowers, which helps fine-tune risk models.
It’s a perpetual motion of one component fuelling the next - a multiplayer shooter game played over high-speed connections until eternity if you will. There’s action, and then there’s entropy.
What kind of data will continuous underwriting need?
Transactional data, particularly open banking data, is becoming increasingly important for leaders looking to improve credit analytics. Interoperable data-sharing frameworks, such as Account Aggregator, provide a more comprehensive view of customers and can include transaction details from multiple banks. This can add colour and texture to a credit profile and help lenders understand spending patterns, behavioural cues, saving propensity, and other non-traditional parameters that can aid credit decisions.
Additionally, non-traditional external data, such as telecom data and information from social and professional networks, can also supplement regular data.
More
What data infrastructure rails do we already have?
The Account Aggregator (AA) ecosystem: The Account Aggregator ecosystem introduced a consolidation of financial information beyond just credit behaviour insights from credit bureaus. It facilitates the aggregation of banking, spending, insurance, investment and other financial information.
The ecosystem, despite its astronomical growth, is still a work-in-progress
While the framework caters to retail lending use cases fairly comprehensively, it still has a long way to go regarding MSME lending. According to Sahamati, individual credit underwriting will comprise 38% of the bank statements shared through AAs by 2027. In the same year, only 10% of the bank statements shared through the framework will likely be used for MSME lending purposes.
Open Credit Enablement Network (OCEN): OCEN is the fourth part of the India Stack framework, following Aadhaar (identity), UPI (payment), and AA (data sharing). OCEN essentially builds a standardised data-sharing layer between marketplaces or platforms (referred to as LSPs, i.e. Loan Service Providers) that generate data and lenders who can capitalise on this data to offer cash-flow-based credit products. The core problem that OCEN solves is the creation of protocols that eliminate multiple 1X1 integrations between LSPs and lenders.
What infrastructure rails do we need?
A Public Credit Registry - The RBI’s mammoth project, still underway, intends to be the full database of data on all credit relationships in the nation, including payments, restructuring, defaults, and resolutions, from the point of credit creation to its termination
Integration of the TReDS platform with GSTN e-invoicing is a significant roadblock. Integration with GST will ensure the automatic uploading of all GST invoices approved by buyers onto TReDS. Subject to consent from MSMEs, this could solve the problem of insufficient verified transaction data of businesses that are available to lenders.
Digitised land records. And not just digitising agricultural land that’s already underway as part of the Agristack, but a single, nationwide window to determine property ownership.
Utility payments data pipeline: TRAI, Central Electricity Regulatory Commission (CERC), and Credit Information Companies should develop a framework for sharing data between telecom companies, electrical utilities, and credit bureaus.
How would a bank’s risk-based pricing decisions change?
Machine Learning for evolved risk-based pricing
Risk is typically calculated based on age, income, employment status, past repayments, and loan amount. That’s about five variables with many sub-indicators, parameters and thresholds. But, the point is that a finite set of indicators can determine whether a borrower gets credit when they need it. Not very inclusive.
Imagine having to make a lending decision for
A 28-year-old man in Mumbai who earns Rs 60,000 a month, pays his rent on time, has a great pay-later approved limit on almost every food/cab delivery app, and has no history of default on loans.
A 28-year-old man in Mumbai who earns Rs 60,000 a month, hasn’t been able to pay his rent for the last couple of months, is eligible for one pay later offer on one app, and has no history of default on loans.
Most likely, traditional risk models would put them both in the same category based on age, income and repayment history - ignoring the part where the second person hasn’t been able to make rent over the last four-five months.
Enter non-traditional data gathering and machine learning. Machine learning identifies complex, nonlinear patterns in large data sets and adds precision to risk modelling. These models learn with every bit of new information they acquire, improving their predictive power over time.
Source: Mckinsey
Portfolio monitoring shouldn’t just be a post-disbursal activity for risk assessment models to improve over time. Currently, once a loan is approved, existing borrowers are monitored for early warning signals of credit risk. Continuous underwriting will mean that borrowers are constantly evaluated for any changes in their financial or behavioural indicators - thus, it doesn’t just raise alarm bells on the possibility of default but also helps lenders upsell or improve the current engagement with the customer.
In some ways, this is how Google does products; for example - the more people use it for search, the more data there is to train its algorithms, and the better its search results become, stealing more users from other search products, a self-reinforcing virtuous cycle.
······ ······
Vast amounts of data have been aggregated by data platforms to manage the flow better.
What’s the business imperative for lenders?
The biggest function of continuous underwriting is creating dynamic, high-performing credit risk models. McKinsey posits that adapting high-performing credit models based on multiple data sources results in -
5-15% increase in revenue
20-40% reduction in credit loss rates
20-40% efficiency gains
How does it help borrowers?
We’ve established utopias for borrowers and lenders - they won’t happen. We’ve also established an ideal world - for lenders so far, but what does it mean for borrowers?
Better borrowing experience: Continuous underwriting will result in new pricing models with better offers, terms or repayment conditions for various borrowers. The best part is that it can evolve in real-time and borrowers can see their experience/offers improve with potentially every single repayment on time, for instance.
How will this be implemented?
According to McKinsey, the migration to continuous underwriting will involve four phases, each forcing increased personalisation and customer engagement:
Phase 1: Lenders focus on automating the underwriting process to improve efficiency and reduce inconsistencies.
Phase 2: Move to accelerated underwriting through digitally-submitted applications.
Phase 3: Micro-segmentation and personalization for individualised offers using insights from comprehensive data sets.
Phase 4: Continuous underwriting with dynamic adjustments based on customer behaviour.
Continuous underwriting in its most utopian form will culminate in the ongoing digitalisation and intelligent risk modelling process. It’ll bring to reality the pipe dream of minimal defaults, profit maximisation and impeccable customer experience, powered through highly effective risk engines and backed by informed consumer consent. It’s a world that is far away but one working towards.
What do you think? I’d love to hear your thoughts, and I’ll see you next week.
Cheers,
Rajat