India’s digital lending ecosystem is no longer an emerging story, but one of the defining pillars of the country’s digital economy. From instant personal loans and Buy Now Pay Later (BNPL) offerings to embedded finance and MSME credit, technology has fundamentally transformed how credit is accessed and delivered. According to a recent report, fintech lenders disbursed nearly 13 crore personal loans in FY26, accounting for almost three-fourths of all personal loan volumes, albeit with an average ticket size of less than Rs. 10,000. The message is clear: lending is becoming faster, smaller, and significantly more inclusive.
This growth is being powered by India’s rapidly evolving digital public infrastructure. The Account Aggregator ecosystem, for instance, facilitated over Rs. 1.6 lakh crore worth of loans in FY25, enabling consent-based access to financial information and helping lenders make more informed credit decisions. Yet, while the infrastructure enabling digital lending has evolved rapidly, the industry’s approach to measuring risk has not kept pace.
Founder & CEO
DPDzero
The models that have governed lending decisions for decades are increasingly struggling to operate in an environment characterised by instant approvals, millions of first-time borrowers, unsecured lending and constantly evolving customer behaviour.
The challenge isn’t that traditional underwriting principles are fundamentally flawed. Rather, they were built for a very different lending landscape.
Historically, credit decisions have relied on four broad pillars: repayment history, collateral, source of income and existing debt obligations. These parameters worked well when lending was slower, borrowers had relatively predictable financial profiles, and credit histories were well established.
Today’s borrower is far more dynamic.
Take repayment history, for instance. Credit bureau records continue to remain the industry’s most trusted source of historical performance, but they are, by definition, backward-looking. They tell us how a borrower behaved yesterday – not necessarily how their financial circumstances or repayment intent may have changed today.
More importantly, traditional sources of borrower data often present an incomplete picture. New-to-Credit (NTC) borrowers, self-employed professionals, gig workers, MSMEs and customers with limited or disputed credit histories are frequently underrepresented within conventional underwriting models. The absence of historical data is often interpreted as the absence of creditworthiness, when in reality it may simply reflect a lack of visibility.
Income assessment presents a similar challenge. While salaried employment offers structured and verifiable income records, a growing section of India’s workforce now earns through multiple streams – consulting assignments, online businesses, rental income, freelancing or seasonal enterprises. These diversified income sources are often difficult to capture through conventional underwriting frameworks, leading to an incomplete assessment of a borrower’s actual repayment capacity.
Debt obligations are equally nuanced. Existing liabilities captured by designated regulatory bodies provide only one part of the picture. Household expenses, family responsibilities, sudden medical emergencies, cost of living and regional economic variations all influence a borrower’s ability to repay, yet these variables rarely feature meaningfully in traditional risk models. Even revolving credit products such as credit cards introduce uncertainty, as future utilisation and repayment behaviour cannot be accurately predicted using static data alone.
These limitations become even more pronounced in unsecured lending.
Traditionally, collateral acted as an additional layer of risk mitigation. Whether it was property, gold or another secured asset, lenders had a secondary buffer against potential defaults. But as digital lending has been skewed towards unsecured personal loans and BNPL products, underwriting quality has become the primary determinant of portfolio performance.
Small-ticket digital loans have undoubtedly improved access to credit, but they have also introduced new complexities. High-frequency borrowing across multiple platforms often generates fragmented repayment histories that are difficult to interpret in isolation. The challenge is compounded by delays in credit reporting.
Loan approvals today happen within minutes, but risk information still travels much more slowly. A borrower may apply simultaneously across multiple digital lenders, receive approvals from more than one institution before those liabilities are reflected in records, and inadvertently become overleveraged. The industry refers to this as loan stacking, and it is one of the clearest examples of why static risk models are increasingly inadequate in a real-time lending ecosystem.
As digital lending scales further, underwriting accuracy must evolve at the same pace as disbursement speed.
Encouragingly, this transformation is already underway.
Lenders are increasingly supplementing traditional bureau data with richer, alternative data signals that provide a more holistic understanding of borrower behaviour. Consent-based financial information, transaction patterns, digital footprints and behavioural indicators are helping build more comprehensive risk profiles, particularly for borrowers with thin or non-existent credit histories.
Equally significant is the evolution of modelling techniques themselves. Traditional regression-based scorecards are gradually giving way to advanced machine learning approaches such as gradient boosting models that can analyse hundreds of interconnected variables simultaneously. These models are better equipped to identify subtle behavioural patterns, continuously learn from new data and improve predictive accuracy as borrower behaviour evolves.
The objective isn’t simply to approve more loans, but to approve the right ones.
However, technology alone cannot solve the industry’s challenges.
One of the biggest structural gaps remains fragmented borrower information. Financial data today exists across multiple systems, such as credit bureaus, banks, GST records, income tax filings, Account Aggregator networks and various digital touchpoints. Every lender attempts to piece together this information independently, often investing significant resources while still operating with incomplete visibility.
This fragmented approach increases underwriting costs at a time when lenders are expanding into underserved customer segments across Tier II, Tier III and rural India. The willingness to lend has grown remarkably, but the underlying information ecosystem has not evolved at the same pace.
The next phase of India’s credit evolution therefore requires greater collaboration across the ecosystem.
Credit bureaus must evolve beyond being repositories of historical repayment records. Reporting cycles need to become significantly shorter so that newly originated loans are visible almost immediately, reducing information asymmetry and curbing loan stacking before it occurs. At the same time, trusted, consent-led data-sharing frameworks can enable lenders to access richer borrower insights while preserving customer privacy and strengthening credit decision-making.
Ultimately, lending has always been a business of confidence – the confidence that capital deployed today will return tomorrow. That confidence has traditionally been built on historical data. In a high-scale digital lending environment, however, history alone is no longer enough.
India has built one of the world’s most sophisticated digital financial infrastructures. The next competitive advantage will not come from approving loans faster – it will come from making faster decisions without compromising on risk. Dynamic underwriting, real-time intelligence and a more connected credit ecosystem will be essential to achieving that balance.
Financial inclusion and prudent risk management are not competing priorities. In fact, they reinforce one another. The more accurately lenders understand borrowers, the more confidently they can extend credit to new and underserved segments while maintaining portfolio quality.
The question is no longer whether traditional risk models need to evolve. The market has already answered that.
The real question is whether the ecosystem can evolve quickly enough to support the next phase of growth for India’s digital lending journey.
–Authored by Ananth Shroff, Founder & CEO, DPDzero
