The checkout screen at Instacart shows $187.43. Four clicks later, it shows four payments of $46.85. Milk, eggs, broccoli, bread — now on installment. The transaction posts to Klarna, not Visa. The credit bureau never hears about it.
Twenty-five percent of buy-now-pay-later users bought groceries on BNPL in the last year, according to a LendingTree survey of 2,000 U.S. consumers released this month. A year earlier, it was 14%. For Gen Z, one in three financed groceries this way. BNPL groceries are the fastest-growing category in the fastest-growing consumer credit product in America, and most credit unions cannot see any of it.
The number that matters is 41%
Of the BNPL users surveyed, 41% paid back a loan late in the past year. That is not late on a luxury handbag. That is late on chicken breasts. LendingTree's own framing was blunt: consumers are using BNPL "to extend their budget," not to time-shift a discretionary purchase. When a member finances food, the distinction between credit as convenience and credit as coping is gone.
Klarna's financials confirm what the survey suggests. The company swung to a $273 million net loss for fiscal year 2025, down from a $21 million profit the year before, with credit provisions rising as its active consumer base grew to 118 million — a 28% year-over-year jump. Klarna is bigger and losing more money on the same loan book. The U.S. consumer is the reason.
Why your prime members are not as prime as they look
Most credit union underwriting still runs on three inputs: a credit bureau pull, a debt-to-income calculation from pay stubs, and a look at the member's deposit history. None of those capture BNPL at scale.
Equifax built the infrastructure to ingest BNPL tradelines back in 2022. FICO launched a BNPL-aware scoring model — Score 10 BNPL — in fall 2025. In practice, pay-in-four loans from Klarna, Afterpay, and Affirm still rarely appear on a standard tri-bureau pull. Furnishing is voluntary, inconsistent, and most BNPL lenders still do not report. The member's file looks clean because the debt is categorized as trade credit and reported in patches, if at all.
So here is what happens in an auto-loan decision: a member with a 740 FICO, steady W-2, no hard inquiries, and no revolving balance walks through the door. The model approves them at tier one. The file does not show that the same member has open balances across four BNPL providers, missed two installments last quarter, and started putting groceries on four payments. The underwriting decision is made in the dark on half the member's liabilities.
What BNPL groceries actually signal
Financing food is a distress signal, not a convenience choice. The same member who will Klarna a sweatshirt for the gamification of four equal payments will not casually install-payment a gallon of milk. If they are doing it, the household cash flow math has already broken.
This is exactly the signal CU risk models were built to catch. Rising overdraft frequency. Payday-loan inquiries. Cash-advance flags on the debit card. Paycheck-to-groceries timing. The telemetry that used to say "stressed household" has a new input layer now, and most credit unions are not ingesting it.
What the regulators are not doing about it
The Consumer Financial Protection Bureau's December 2025 BNPL market report flagged the grocery trend as a structural concern, not a cycle-driven blip. But the 2024 interpretive rule that would have pulled BNPL products under Regulation Z as credit cards — forcing periodic statements, billing-dispute rights, and clearer reporting — is being withdrawn. The CFPB confirmed in a March 2025 court filing that it plans to revoke the rule. Enforcement has been deprioritized.
Meanwhile the on-ramp to financing food is being paved at every major node. Klarna replaced Affirm as Walmart's exclusive BNPL partner. DoorDash integrated Klarna for food delivery. Instacart already had it. The information asymmetry between the BNPL lender and the credit union is not closing. It is widening, with regulatory cover.
What a CU underwriting response looks like
The fixes are not exotic.
Add a BNPL disclosure question to every consumer loan application and require it in writing. Subscribe to the extended credit bureau products that include BNPL tradelines where furnished. Adopt Score 10 BNPL as a parallel decisioning input even if it is not the primary score. Build a cash-flow-based underwriting overlay that reads the member's own deposit account for the pattern: grocery store transaction, followed within hours by small-dollar debits to Klarna, Afterpay, or Affirm. That data is sitting in every credit union's core, and almost nobody is reading it.
The members who need a credit union most are the ones BNPL is intercepting first. The same household that five years ago would have asked for a $500 share-secured line of credit is now splitting groceries into four payments at checkout and never walking through the branch.
The real risk is not the Klarna loan
It is the auto loan the credit union just approved for the member whose grocery bills are on Klarna. Or the HELOC. Or the credit card line increase. The decision gets made on a file that omits the most recent, most behavioral, most predictive signal about that member's ability to pay — which is that last week, they could not cover dinner.
The underwriting model has a blind spot. BNPL groceries are where it lives.