Bureau credit scores should not be the sole filter: IDfy's Ashok Hariharan

Founded in 2011, IDfy is an identity verification platform to eliminate fraud, establish trust, and fulfil the need for security and compliance. With the central bank tightening underwriting standards, evaluation of customers can no longer rely only on credit bureau scores. Ashok Hariharan, founder of IDfy, interacted with Raghu Mohan via email on issues in retail onboarding and underwriting. Edited excerpts:

 

Do you think traditional underwriting is failing new-to–credit (NTC) borrowers?

 

In my view, it does not serve NTC borrowers well. It still leans heavily on bureau scores, which only reward those who already have a credit history. For first-time applicants, this becomes an exclusion filter. We see this reflected in the data: less than 18 per cent of new loan originations in 2024 went to NTC borrowers, down from 25 per cent the previous year. In unsecured loans, the share is even lower, around 10 per cent, and drops further for higher ticket sizes. This hurts precisely the segment that needs access most: young professionals and startup employees who may have strong financial potential, but no track record. The paradox is stark: India’s most digital generation, with rich online financial footprints, is also the most excluded from formal credit.

 

If credit bureau scores are not to be taken as the sole indicator for NTC, what are the other reasonable evaluation parameters?

 

I don’t believe bureau scores should be discarded, but they must not be the sole filter. What matters more are forward-looking signals of ability and the intent to repay. A person’s banking transactions like salary credits, spending consistency, bounce rates, and savings behaviour reveal far more than a backward-looking number. For self-employed borrowers, tax and Goods and Service Tax filings are strong proxies for business health. Verified income signals, whether payroll data or gig-platform earnings, can also provide stability markers that bureau scores miss. India’s digital infrastructure already allows this shift. In FY24, UPI (Unified Payments Interface) transactions were worth ₹71 trillion, a footprint larger and more relevant than bureau coverage. With the account aggregator framework, lenders can now access verified data directly from banks, while AI (artificial intelligence) tools decode messy narrations into clear cash-flow statements. This makes it possible to bring corporate-style cash-flow analysis to individual borrowers. That, to me, is how we move from judging credit history to evaluating credit potential.

 

The Reserve Bank of India’s Financial Stability Report of June 2025 says even as unsecured retail lending has moderated, asset quality has relatively weakened with gross non-performing asset (NPA) ratio at 1.8 per cent vis-à-vis 1.2 per cent in March 2025. This even as household debt has gone up. What is your reading of this larger setting?

 

We need to dig deeper to interpret these numbers. Unsecured retail lending, which grew 12–13 per cent last year, is now expanding only in single digits. So even if household debt appears higher, it is in the context of slower incremental growth. What is really happening is that lenders are tightening underwriting criteria. That has led to slightly better NPA outcomes, but at the cost of slower book growth. It is also important to note that banks haven’t necessarily raised their bureau score cut-offs; instead, they have tightened other aspects of underwriting. This is the crux of the matter: bureau scores don’t correlate as strongly with lending quality as we often assume. Credit quality can improve through richer, contextual parameters without being punitive to borrowers. That, to me, strengthens the case for looking beyond bureau scores, because it allows lenders to grow their books responsibly without compromising on risk. 

But lenders are sticking to legacy on-boarding and underwriting approaches because of regulatory concerns on delinquencies (like in unsecured credit). And they are going for the better-rated borrowers. Surely, these aspects cannot be given the go-by.

 

To be fair, lenders are constantly reviewing their on-boarding and underwriting frameworks. But in practice, most banks tend to respond to credit quality pressures by adding more checks, in effect, more ways to reject, rather than balancing this with mechanisms to expand the funnel. Non-banking financial companies are relatively more mindful of rejection criteria, yet they too are not fully oriented towards making credit decisions based on new, non-binary data points. Ideally, the efficacy of any new variable or underwriting step should be tested in a champion–challenger mode, as we see in data-centric banks globally. In India, though, tightly coupled infrastructures between loan origination and management systems often prohibit active experimentation. This kills ideas before they are tried. A good example is banking data: most lenders agree there are logical variables to be derived, but the effort required to integrate them into experiments is so high that they are often abandoned at the idea stage itself.