BNPL Repayment Can Hide the Checking-Account Damage
Executive summary
Pay-in-four lending produces an appealing performance story. The loan is small, the first installment is collected at checkout, later payments are automated, and most loans are repaid. The Consumer Financial Protection Bureau reported a 1.8% charge-off rate for 2023 among the firms it studied.
New Federal Reserve evidence shows why that lender-level result is incomplete. In the 2025 Survey of Household Economics and Decisionmaking, 11% of buy now, pay later users said a BNPL payment triggered an overdraft or non-sufficient-funds fee. The rate was 18% among users with less than $100 in emergency savings and 4% among those able to cover $2,000 or more. Forty-three percent of users who financed groceries or food delivery were charged extra for paying late or incurred an overdraft or NSF fee.
The mainstream interpretation is that high repayment validates underwriting and confirms that pay-in-four is a convenient, low-cost alternative to revolving credit. The contrarian reading is that repayment success can be partly manufactured by the payment architecture. A down payment and automatic debit protect the BNPL receivable, but the debit can move the shortfall into the consumer's checking account. The lender is paid; the bank account absorbs the failure.
Operators should therefore measure BNPL as a linked flow across merchant conversion, lender repayment and household liquidity. A clean loan book does not establish a healthy customer outcome.
Market context
Pay-in-four lets a consumer split a purchase into four equal payments over six weeks, generally without interest or fees when payments are made on time. It differs from longer point-of-sale installment lending and from revolving credit. The customer receives the product immediately, pays one-quarter at checkout and schedules three biweekly installments.
The product has moved from a niche checkout feature into a meaningful credit channel. The Federal Reserve's August 2026 review, drawing on CFPB data, says the number of loans made by five large BNPL firms increased from 19.8 million in 2019 to 335.8 million in 2023. Dollar volume rose from $2.7 billion to $45.2 billion. Roughly 45 million people used a BNPL app monthly at the end of 2024, up from about 10 million in early 2021.
The Fed's main consumer evidence comes from the 2025 SHED, a nationally representative survey of nearly 13,000 adults. Sixteen percent of adults said they used BNPL in 2025, up from 10% in 2021. Users often described the product as a preference: 87% cited spreading payments and 82% cited convenience. That matters because a stress-only account would be wrong. BNPL is useful to many customers and can make cash flow more manageable without the interest charged on a revolving card balance.
But usage is not evenly distributed. Adults with less than $100 in emergency savings used BNPL at almost four times the rate of adults able to cover $2,000 or more. Growth has been concentrated among people facing credit and liquidity constraints. One in five users financed groceries or food delivery; the share was 29% among users earning less than $50,000 and 9% among those earning at least $100,000.
The product's credit visibility remains fragmented. Most pay-in-four loans are not reported to the credit bureaus, so timely repayment generally does not help a score and other lenders cannot see the full set of outstanding BNPL obligations. The provider can observe its own repayment history, but neither the credit file nor another BNPL lender necessarily shows the household's complete installment stack.
Findings
Finding 1
The repayment mechanism protects the lender before it protects household liquidity.
Pay-in-four underwriting typically combines a soft credit check with the provider's proprietary repayment history. The mechanics then add two controls. The borrower makes a down payment at origination, and the provider verifies through card networks that funds sufficient to repay the loan are available in the linked account at that moment. Later installments are commonly automated debits from a checking account.
Those controls are rational. They reduce exposure, simplify collection and make a no-interest product possible. They also influence where distress appears. A borrower can repay the BNPL installment while lacking enough money to cover another debit that arrives the same day. The BNPL account remains current, while the checking account goes negative or a different obligation is returned.
The Fed's data show a steep liquidity gradient.
| Emergency expense covered using savings | Adults using BNPL | BNPL users whose payment triggered overdraft/NSF |
|---|---|---|
| Less than $100 | 31% | 18% |
| $100-$499 | 28% | 13% |
| $500-$999 | 21% | 6% |
| $1,000-$1,999 | 16% | 6% |
| $2,000 or more | 8% | 4% |
| Overall | 16% | 11% |
Source: Federal Reserve, Consumer & Community Context, accessible tables, published August 2026; underlying period is calendar 2025. Units are percentages. BNPL use is measured among all adults; overdraft/NSF incidence is measured among BNPL users, so the two columns have different denominators. Emergency capacity is self-reported savings available at the time of survey. Figures are descriptive and do not prove that BNPL caused broader financial distress.
The 18% rate in the lowest-savings group is 4.5 times the 4% rate in the highest-savings group. That is not evidence that every low-liquidity user is harmed or that BNPL is worse than the available alternative. It is evidence that repayment performance and account health answer different questions.
The correct unit of analysis is the household payment sequence. Operators need to know whether an installment was paid, whether it created a bank fee, whether another bill failed, and whether the customer used fresh credit to restore liquidity. Public data do not provide that full sequence. The gap is precisely why a low charge-off rate should not end the analysis.
Finding 2
Essential purchases reveal a different product than discretionary checkout does.
Clothing and accessories remain the most common BNPL use case, reported by 49% of users. Electronics, furniture and appliances also fit the familiar proposition: convert a larger retail ticket by matching payment timing to the household budget.
Groceries, food delivery and medical or veterinary expenses behave differently. They are consumed quickly, cannot be repossessed in any meaningful sense and often recur before the prior installment plan has ended. Financing them does not smooth a durable asset over its useful life. It pulls a basic expense forward while preserving three claims on future cash.
| Purchase financed with BNPL | Share of BNPL users who bought category | Charged extra for paying late or incurred overdraft/NSF |
|---|---|---|
| Clothing or accessories | 49% | 25% |
| Electronics | 32% | 20% |
| Furniture or appliances | 26% | 23% |
| Groceries or food delivery | 20% | 43% |
| Travel expenses | 19% | 24% |
| Medical or veterinary procedures | 8% | 34% |
| Other | 16% | 18% |
| Overall | Not applicable | 21% |
Source: Federal Reserve, August 2026, using the 2025 SHED. Units are percentages of BNPL users. Respondents could report multiple purchase categories, so category shares do not sum to 100%. The adverse-outcome measure combines a BNPL late charge with a bank overdraft or NSF fee and does not identify which fee applied. Associations are descriptive; purchase categories may proxy for unobserved financial constraints.
The 43% adverse-fee rate among grocery or food-delivery users is not a generic BNPL statistic. It identifies a use case where the six-week structure may be too long relative to the replenishment cycle. A household that uses pay-in-four for groceries every two weeks can have multiple plans running before the first basket is fully paid.
Merchant conversion can still improve. But the incremental order may be borrowing from the customer's next several pay periods and from future category demand. A merchant that evaluates only approval and initial conversion will miss refund behavior, repeat cadence, customer-service contacts and later payment stress.
Finding 3
The market's credit record is incomplete in both directions.
Non-reporting creates a double blind spot. Consumers generally receive no credit-building benefit from successful pay-in-four repayment, while lenders lack a holistic view of concurrent obligations. Collections may eventually reach a credit file, but normal outstanding plans usually do not.
This weakens the meaning of provider-specific repayment data. A lender can learn that a customer repaid its last three plans. It may not know that the same customer has four active plans elsewhere, a low checking-account balance and a card payment due tomorrow. Proprietary behavioral data improve local underwriting without necessarily improving market-wide visibility.
| Observed metric | What it establishes | What it can miss |
|---|---|---|
| BNPL repayment rate | Provider's receivable was collected | Overdraft, NSF, displaced bills, new borrowing |
| Charge-off rate | Provider recognized a loan loss | Bank fees and hardship before charge-off |
| Checkout conversion | More orders completed with the option | Incrementality, returns, later demand pull-forward |
| Repeat BNPL use | Customer returned to the product | Reliance on rolling installment stacks |
| Credit-bureau file | Traditional reported obligations | Most active pay-in-four plans and timely BNPL payment history |
Source and method: Blackrock Research mapping based on the Federal Reserve's August 2026 product description and the CFPB market figures it cites. Period: evidence published 2025-2026, describing activity through 2025. Units are qualitative metric definitions. Limitation: this framework does not assign causality or estimate net consumer welfare; it identifies measurement boundaries.
The mainstream case remains plausible: short, small loans with no interest can be cheaper than revolving balances, payday loans or overdrafts. The missing evidence is substitution. Public surveys do not show, transaction by transaction, what the consumer would have used instead or whether the BNPL debit caused another obligation to fail.
That uncertainty cuts both ways. It prevents declaring BNPL categorically harmful. It also prevents treating a repaid loan as proof that the financing improved the household's position.
Implications for operators
BNPL providers should add liquidity outcomes to risk dashboards. Alongside approval, repayment and loss, track payment rescheduling, debit failures, repeated attempts, customer-reported overdrafts, hardship contacts and the number of concurrent plans visible within the provider. Give customers a clear view of total installments by pay date rather than presenting each purchase as an isolated plan.
Merchants should measure the option as a cohort, not a tender badge. Compare incremental conversion, average order value, returns, repeat timing, gross margin, support demand and churn with similar customers who did not use BNPL. Essential categories need separate policies because replenishment can overlap with outstanding installments.
Banks can see the downstream account effect that a BNPL provider often cannot. They should identify automated installment debits in cash-flow tools, warn when upcoming payments exceed projected available funds and make it easy to understand which merchant purchase created the debit. Interventions should inform customers without arbitrarily blocking lawful payments.
Credit bureaus and lenders need reporting designs that reflect short duration without flooding files with account openings and closures. A useful system should show concurrent exposure and reward timely repayment while avoiding a structure that mechanically harms scores because each six-week plan appears as a new loan. The design problem is real; invisibility is not a neutral solution.
Regulators and researchers should connect datasets across lenders and deposit accounts using privacy-preserving methods. The decisive question is not only whether BNPL is repaid. It is whether total fees, missed obligations and subsequent borrowing are lower than under the consumer's realistic alternative.
Risks & open questions
The SHED is a survey, so the evidence relies on recall and self-reporting. It identifies associations, not causality. Consumers with low savings are more likely both to use BNPL and to incur overdraft fees for many reasons. A BNPL installment may trigger the fee without being the underlying cause of financial fragility.
The adverse-outcome question combines fees charged by a BNPL provider for late payment with overdraft or NSF fees charged by a bank. It does not provide dollar amounts. A small late fee and a cascade of returned-payment costs therefore appear in the same measure.
The thesis would weaken if linked transaction data showed that BNPL users paid lower total financing and account fees than matched consumers using cards, overdrafts or alternative credit, while experiencing no increase in missed essential bills. It would also weaken if broader reporting materially reduced loan stacking without excluding thin-file consumers.
The thesis would strengthen if the same data showed that automated BNPL debits are routinely followed by overdrafts, returned utility or rent payments, or fresh short-term borrowing. Public evidence does not yet resolve that sequence.
Appendix / methodology notes
This report uses the Federal Reserve Board's August 2026 Consumer & Community Context and its accessible tables, which draw primarily on the 2025 Survey of Household Economics and Decisionmaking. The SHED is nationally representative and included nearly 13,000 respondents. Market-volume figures come from the CFPB's December 2025 review of five large BNPL firms and cover 2019 through 2023.
All table values are direct transcriptions except the stated 4.5-times comparison, calculated as 18 divided by 4. Percentages use the denominators specified by the Fed. No synthetic market data are presented.
A conclusive follow-up study would link, at the household level, every pay-in-four origination and installment to daily deposit balances, overdraft/NSF events, other bill payments, other BNPL plans, credit-card use and subsequent short-term borrowing. It should compare matched users and non-users by liquidity, income and purchase category, then measure total fees and missed obligations for at least 90 days after purchase. That dataset is not public.