A private sector bank's marketing team identifies a segment. Salaried professionals, 28–40, bureau scores above 740, steady income band. Every acquisition metric says yes. The CRO's team runs the numbers. Bureau check passes. Income verification passes. Loan gets approved.
Eighteen months later, that vintage is in the top quartile of slippages.
Nobody made a bad credit decision. The decision was made on incomplete information. The signals that would have said no. Transaction volatility, high-frequency borrowing behavior and cash flow irregularity were sitting in systems that never connected to the underwriting engine. The CRO did not fail to think. The organisation failed to connect.
Not a credit judgement problem but it is a data flow problem
The Indian banking sector is, by headline metrics, in remarkable shape. The GNPA ratio of scheduled commercial banks reached a multi-decade low of 2.1% as of September 2025. Aggregate net profit hit ₹4.01 lakh crore in FY25. Capital buffers are robust. The RBI's Financial Stability Report reads like a sector that has figured things out.
Look one level beneath the aggregate, and a different picture emerges.
In private sector unsecured loans now contribute 76% of total slippages. More than half of all retail loan slippages across the system originate from personal loans and credit cards. These are the products where the signal set is thinnest: a point-in-time bureau pull, a demographic proxy, an income estimate. Approved fast. Monitored late.
The headline portfolio looks clean even as some vintage stress is accumulating inside it.
Marketing optimizes for conversion probability. It uses the signals available to it in form of bureau scores, income bands and segment demographics because these signals predict whether a customer will say yes to the product. Risk optimizes for default probability. It uses largely the same signals, because those are what the origination system ingests.
Both teams are doing their jobs to the best of their resources but with an incomplete view of the same customer. The transaction-level behavior, the ERP data, the cash flow patterns that would distinguish a stable borrower from a stressed one, all of that information exists somewhere in the organisation, or is accessible through the Account Aggregator framework. It just never reaches the underwriting decision.
Every time a new product is launched, or a new segment is targeted, the blind spot widens. Because the signal infrastructure was not connected when the first product was built, and it is not connected now.
The CRIF Highmark MSMEx Spotlight Report from December 2025 makes the cost of this visible. MSME borrowers managed through working capital facilities (which require active transaction monitoring and frequent statement reviews) show a PAR 91-180 of 1.34%. Borrowers on term loans, originated on a one-time assessment of ITR and bureau score, show 2.4%. A 79% higher delinquency rate. Same borrower profile on paper. Different signal depth in practice.
That 79% is what the blind spot costs!
The more uncomfortable finding is what the digital transformation of the last five years actually delivered.
Under EASE 7.0 and 8.0, PSB loan turnaround times fell from 30–45 days to under 60 minutes in some cases. Every public sector bank achieved 100% digital journey implementation for retail, agriculture, and MSME lending by FY25. The industry celebrated the speed.
What the Lok Sabha records (Ministry of Finance responses from March 2026) is that PSB Loans in 59 Minutes primarily uses three data sources: GST filing data, ITR details, and six-month bank statements uploaded as PDFs for OCR processing. These are the same signals as traditional paper-based lending. They are delivered faster but not more connected.
The sector did not build signal infrastructure. It built faster pipes for the same limited signals.
Meanwhile, the Account Aggregator framework has made 2.12 billion financial accounts shareable. That is 61% of total accounts in the system. As of September 2025, 308 million consents have been fulfilled. The Finarkein AA in Action report puts it precisely: 71% of consumers will share financial data for better loan terms, up from 33% in 2023.
Despite all of this, AA-facilitated lending accounted for approximately 3% of total incremental lending penetration in India in FY25. The data is available. The decisioning layer is not built to use it.
The RBI has now made this a compliance question, not just a strategic one. The January 2026 Comprehensive Credit Risk Management Framework mandates Early Warning Systems that use advanced data analytics and technology-driven tools, explicitly including Account Aggregator and ULI frameworks , to detect financial stress continuously as against episodically. Institutions with inadequate EWS signal coverage face increased operational risk capital requirements. The EASE 8.0 agenda lists AA integration and agentic AI adoption as new priorities, which means the regulator has implicitly acknowledged that the previous cycles did not achieve signal connectivity.
The mandate does not prescribe field-level inputs. But it makes the direction of travel unambiguous. Marketing-only signal sets are now a regulatory liability.
The CRO does not have a credit judgement problem. The CRO has a data architecture problem dressed up as a credit judgement problem.
The fix is not a tighter credit policy. Instead it is a connected intelligence layer -one where the signals marketing uses to identify a segment are visible to risk, and the signals risk uses to monitor a portfolio are visible to marketing, at the moment the decision is made.
When the CMO and CRO are looking at the same data, the blind spot closes.
FINVIJ builds decision intelligence infrastructure for financial institutions. Pragna DIS connects marketing signals, bureau data, transaction history, and risk models into a single intelligence layer so credit decisions are made with the full picture, not half of it. If you wish to see how Pragna delivers, write to vijay@finvij.com or request a demo.