India · Consumer Credit Economics · 15 May 2026
Indian credit card NPAs at ~2.2% and trending. RBI risk weights raised in November 2024 on unsecured retail. The card-issuer's underwriting model is being repriced — and the cohort that gets credit next is structurally different from the cohort that got it in 2022-23.
The composite scenario
Three composite consumer-credit stakeholders — grounded in publicly-disclosed NPA data (RBI FSR, SBI Card, HDFC Bank, ICICI Bank, Axis Bank quarterly results), RBI master directions, and CIBIL CMR cohort data — surface what is measurable about the credit card announcement layer, and what is structurally opaque about its vintage-NPA and sourcing-channel layer.
Anand manages a portfolio of ~4.2 million active cards. The November 2024 RBI circular raising risk weights on consumer credit and credit cards to 125% from 100% changed the capital calculus overnight. Cost of credit — already moving on RBI repo rate signals — tightened further at the portfolio margin. New underwriting cut-offs moved CIBIL thresholds from 720 to 740+ for unsecured cards. The cohort that would have been approved in Q3 FY24 is the cohort now denied.
The composite-modelled friction is not at the aggregate NPA line — it is at the vintage-NPA curve for cards issued in the 2022-23 sourcing surge. That vintage carries a sourcing-channel mix (DSA-heavy, fintech co-brand, pre-approved SMS offers) that the quarterly NPA number averages across all vintages and obscures.
NPA Failure Mode: vintage-cohort blendingThree active cards: HDFC Millennia (limit ₹1.2L), SBI SimplyCLICK (limit ₹80K), Axis Flipkart (limit ₹60K). Total sanctioned unsecured limit: ₹2.6L. Credit utilisation at 72% blended. Two BNPL accounts (Slice, LazyPay) — both visible in the credit bureau record but not uniformly counted as "credit card outstanding" in the NPA numerator.
EMI conversion on a ₹38K appliance purchase — effective APR 28.4% post the RBI's effective-APR disclosure requirement, which is now disclosed on the statement but not in the marketing collateral. The composite-modelled friction is the gap between the disclosed utilisation ratio and the actual debt-service-obligation ratio when BNPL EMIs are added to the credit-card minimum payment requirement.
NPA Failure Mode: BNPL-credit stack invisible in bureau NPA denominatorTasked with designing macro-prudential interventions for the unsecured retail segment. The November 2024 risk-weight increase is one instrument — but the vintage-NPA curve for the 2022-23 sourcing cohort will take 18–24 months to fully surface in quarterly NPA disclosures. The tool that arrives ahead of the curve is the CIBIL CMR (Credit Market Report) — which breaks out NPA by credit score band and sourcing channel — but it is not a public disclosure.
"The 2.2% is
blended. The 2022-23
vintage is not."
Anand, composite 01 · Risk head · Major card issuer
"Three cards.
Two BNPL.
One bureau record."
Rekha, composite 02 · Multi-card revolver · Urban Tier-2
"The risk weight
moved. The vintage
is still moving."
Vikram, composite 03 · Supervision officer · RBI
The disclosure asymmetry
The NPA is real. The vintage decomposition is conditional on disclosure.
What is announced is blended. What is blended hides the sourcing-channel vintage math. The asymmetry holds across issuer size, credit-score bands, and BNPL-overlap ratios. It is the predictable output of a quarterly-disclosure architecture optimised for aggregate NPA trajectory, not cohort-level credit-quality decomposition.
Three composite stakeholders. Three NPA failure modes. One signal underneath: the credit card NPA cycle works at the blended aggregate rate the announcement layer discloses — and the next-sourcing-cohort and BNPL-stack reality does not yet appear at that resolution. The cycle is not in question. The vintage decomposition of the cohort doing the cycling is.
Disclosure-eligible accountability
These six questions are disclosure-eligible under SEBI LODR and the RBI's quarterly results framework. The answers, when they arrive, will be more useful than the next aggregate NPA percentage.
What is the vintage-cohort NPA curve for cards issued post November 2024 vs pre November 2024, segmented by issuer and by sourcing channel (DSA · fintech co-brand · bank branch · pre-approved)?
How does the underwriting CIBIL cut-off (current floor: 740+) flow through to portfolio NPA over an 18-month horizon — and what is the disclosed NPA rate for the sub-720 cohort issued pre-tightening?
For multi-card revolvers holding 3 or more active cards, what is the disclosed credit-utilisation-to-default migration matrix — and how is BNPL outstanding counted in the denominator?
What is the BNPL-to-credit-card portfolio overlap ratio in the issuer's active book, and how is it stacked in the credit bureau record — is BNPL counted as equivalent "credit" in the NPA numerator and denominator?
How is the EMI-conversion product priced post the RBI's effective-APR disclosure requirement — and what is the disclosed default rate on EMI-converted balances vs revolving balances at month-12?
For card portfolios in restructuring or under the RBI's Resolution Framework, what is the disclosed forbearance quantum and recovery rate at month-24 — broken out from the standard NPA line?
The three composite scenarios in this report are not allegations. They are the predictable shape of what happens when a sector's credit-quality story is disclosed on a quarterly blended-NPA basis and its vintage-cohort decomposition is held back to a portfolio average.
The question is not whether credit card NPAs are rising. It is whether the cohort doing the rising is the cohort the underwriting model assumed.
RAOSCAFF uses AI-modelled composite scenarios and a predict-not-recommend voice to surface what India's flagship sectors actually deliver in credit-quality terms — and to whom.
Predict-not-recommend: these are AI-modelled structural patterns, not investment, legal, or financial advice.
RAOSCAFF Intelligence — new briefs publish as investigations close.
Composite consumer-credit stakeholder profiles modelled from publicly-disclosed NPA data (RBI Financial Stability Report, RBI Supervisory Report on credit cards), listed-company quarterly results (SBI Card, HDFC Bank card segment, ICICI Bank retail credit, Axis Bank card NPA commentary), CIBIL Credit Market Report published ranges, and RBI Master Direction on Credit Card and Debit Card Issuance (2022, amended 2024).
No specific company alleged. No proprietary numbers reproduced. The analysis layer is RAOSCAFF's; the disclosure framework is public; the composite scenarios are scenario-grounded model outputs based on publicly-cited NPA ranges and credit bureau data. Vintage-NPA estimates, BNPL overlap ratios, and EMI-conversion default rates are illustrative ranges drawn from the public-disclosure envelope, not specific to any one issuer.
1. RBI Financial Stability Report (Dec 2024, Jun 2025) — sectoral NPA data, unsecured retail stress commentary, risk-weight circular reference.
2. SBI Card / HDFC Bank / ICICI Bank / Axis Bank — quarterly results FY25-26, card segment NPA disclosures, provisioning commentary.
3. RBI Master Direction on Credit Card and Debit Card Issuance (2022, updated Nov 2024 risk-weight circular) — underwriting norms, effective-APR disclosure mandate, sourcing-channel regulation.
4. CIBIL Credit Market Report — credit-score-band NPA data, utilisation migration matrices (published industry ranges).
5. BNPL provider disclosures (Slice / Uni / LazyPay) — credit-bureau reporting methodology, NPA classification approach.
6. SEBI LODR — segment-reporting and quarterly results disclosure obligations applicable to listed card issuers and banks.