CHI Company Investigation · Finance
Affirm
Pay-over-time loans · Pay in 4 · Affirm Card · Affirm Money
Fourteen research streams · Disposition: narrow investigation only · No numerical score assigned
A cleaner penalty structure than revolving credit — paired with growing reliance on borrower interest, more frequent use and wider access to pay-over-time credit.
This investigation asked two questions. Did Affirm keep the customer promise it was built on? And what changed around that promise as the company grew? The first answer is largely yes: no late fees, no penalty rate, no deferred interest, no compounding, and a cost fixed at checkout. The second is where the finding sits. The relationship has moved from a loan for one purchase at one merchant toward credit that is available almost everywhere, used far more often, and paid for in large part by borrower interest.
The promise What survives
No late fee, no penalty rate, no deferred interest, no compounding — in every period examined.
Affirm's 2016 Terms of Service already told borrowers they would never be charged a late fee, and every annual report since the IPO repeats that Affirm has charged $0 in late fees since its founding.A1A5 No research stream found a late fee, penalty rate, deferred interest or compounding interest in any market. No paid consumer tier has launched. Each loan has a fixed schedule, a fixed end date and a total cost shown before the borrower commits.
The relationship What changed
More of the money comes from borrower interest, the price ceiling rose, and use became far more frequent.
Interest income was about 48% of fiscal 2026 revenue, and about 70% of volume carried interest. The maximum APR rose from 30% to 36% in fiscal 2023, which management tied to higher funding costs. Transactions per active consumer rose from 2.0 to 7.0 a year, and the Affirm Card put pay-over-time credit in 5.2 million wallets.A5A7A15
Both columns describe the same product at the same time. Meaningful customer value and meaningful extraction coexist here. That is why the evidence supports a narrow investigation and not a broader hostility thesis — and why this page argues neither that Affirm became hostile nor that it is simply customer-friendly.
Executive finding
Affirm kept the promise it made about fees. What changed is the scale, frequency and price of the credit around it.
Affirm sells fixed-term instalment loans at checkout, through its app and through the Affirm Card. A partner bank or Affirm itself makes each loan; Affirm services it and either keeps it or sells it. CHI's research phase ran fourteen parallel streams over Affirm's filings, shareholder letters, consumer and merchant terms, regulatory record and independent research, and closed on 1 October 2026.
No late fee, penalty APR, deferred interest or compounding interest was found in any period or market examined.
Contractual interest paid by borrowers was 42% of FY2026 revenue; about 70% of volume was interest-bearing.
The APR ceiling rose from 30% to 36%, explicitly to offset funding costs. No later reduction is on the record — and none can be tested.
Transactions per active consumer rose from 2.0 to 7.0; repeat customers made 96% of transactions; the card reached about 20% of quarterly volume.
No public enforcement action was found, the institutional signals are not predictive, and portfolio credit held up. Household-level outcomes are unknown.
How to read this page
The research labels every material claim by what kind of evidence it is. This page keeps those labels on the claims where the distinction matters.
The original promise
The promise was certainty, not cheapness. And it was never a promise of no consequences.
Affirm's earliest consumer pitch was not a low price. It was a fixed, known cost with no penalties for falling behind — a closed-end loan with an end date, set against the open-ended balance of a credit card.
- 2013
The first product was a checkout tool, not a loan: Affirm paid the merchant at once and the customer had 30 days, interest-free, to repay, with revenue from a merchant fee. Fact (secondary source)
- June 2014
The instalment loan launched. Co-founder and chief executive Max Levchin told the press that the goal was not to make money from people going into debt. Fact (press report, paraphrased)
- 2015–2017
Public statements stressed a simple fixed-term loan with no compounding and no hidden fees, and argued that people care more about certainty of what they pay than about the price of credit. (press reports, paraphrased)
- 24 February 2016 · the earliest primary terms
A closed-end loan of 3, 6 or 12 months from Cross River Bank at 10% to 30% fixed APR, with 0% at select merchants. “You will never be charged any late fees.” The same terms allowed Affirm to report missed payments to credit bureaus and sent disputes to binding arbitration.A1 Fact
- January 2019
To the FDIC: APRs of 0–30%, no origination, prepayment, late or other consumer fees, an average loan of about $800 and terms of 3 to 48 months. The maximum term had lengthened; the ceiling had not moved.A23 Fact
- November 2020 · the IPO filing
“$0 in late fees” since 2012; simple interest that never compounds; consumers never pay more than they agreed at checkout. The same filing disclosed, as a risk factor, that Affirm's mission may conflict with the short-term interests of its stockholders.A4 Company claim
What the promise was
- No late fees — contractual since at least February 2016.
- No penalty rate and no deferred interest.
- Simple interest on a fixed schedule — no compounding.
- A total cost fixed at checkout, and no prepayment penalty.
What it never was
- A promise of low rates. The 2016 ceiling was already 30%.
- A promise of no consequences. Negative credit reporting and arbitration were in the 2016 terms.
- “No fees of any kind.” The standing phrase is “no late or hidden fees”. The card agreement reserves “optional fee-based programs” whose content is unknown, and third-party reports of processing fees on some legacy Canadian plans are unresolved. Not established
The wording has softened at the edges; the substance has not. In the business section of the annual report, “never” became “do not” in two steps (FY2023, resting on phrase search only, and FY2024), and “honest” left the new-products sentence while staying in the mission. No document explains the change and no late fee has been introduced. Not established This page reports it as a change of wording and nothing more.
How Affirm makes money
Affirm is substantially borrower-funded. The accounting makes that both clearer and murkier than it looks.
Affirm reports five revenue lines and no consumer-fee line. In fiscal 2026 — the year to 30 June 2026 — revenue was $4,261.1 million.A5
$2,047.5m · 48.0%Interest incomeMostly borrower interest; see below
$1,149.9m · 27.0%Merchant network revenueFees paid by merchants
$596.6m · 14.0%Gain on sale of loansPaid by loan buyers
$294.0m · 6.9%Card network revenueInterchange, paid by merchants
$173.1m · 4.1%Servicing incomePaid by third-party loan owners
Source: Form 10-K for fiscal 2026, income statement.A5 The five lines foot to total revenue. Fact
The accounting line is not the same as who ultimately pays
Contractual interest paid by borrowers
$1,796.8m of simple interest on unpaid principal. This is money borrowers pay, and it is the single largest source of Affirm's revenue. Fact
Paid by merchants at the point of sale
Merchant network revenue (27.0%) plus card network revenue (6.9%). Merchant-paid revenue was about half of the total in fiscal 2019 and 2020. Fact
Paid by loan buyers and investors
Gain on sale (14.0%) and servicing (4.1%). These buyers are largely paying for future borrower interest on interest-bearing loans, or for the unamortised merchant discount on 0% loans. They are a channel for the other two payers, not an independent third one; no number can be put on the split. Analytical inference
The remaining 5.9% sits inside “interest income” but is not borrower interest. When Affirm buys a 0% or below-market loan from a partner bank at full face value, it books the gap to fair value as an upfront loss and earns it back over the loan's life as “interest income” — $332.8m of discount amortisation in FY2026, less $27.4m of premium amortisation and $54.7m of interest receivable charged off. On a merchant-funded 0% loan the borrower pays no interest; the merchant's fee carries the cost. Accounting line is not ultimate economic payer.A5
The path is U-shaped, not a steady climb. Interest was already 45% of revenue in fiscal 2019. The low years, FY2020–FY2022, were shaped by Peloton's merchant-funded 0% loans (Peloton alone was 28% of FY2020 revenue), so the IPO period is a poor baseline. Measured on contractual borrower interest alone the shift is sharper: 27.1% of revenue in FY2022, 44.9% in FY2024 and 42.2% in FY2026. In dollars, contractual interest rose about 4.9 times from FY2022 to FY2026 while volume rose about 3.2 times. Ratios computed from 10-K figures.A5 Fact
“We earn a commission from businesses, and shoppers pay interest on some items.”Affirm, “Why Affirm” consumer page, retrieved 1 October 2026
Explaining the two-sided model on a consumer page counts in Affirm's favour. “Some items” understates the position: interest-bearing loans were about 70% of fiscal 2026 volume.A3 Analytical inference
Who pays for 0%? Affirm says nearly 95% of 0% monthly volume at integrated merchants is funded by the merchant, which pays a larger fee for those offers; manufacturers fund some promotions; and Affirm funds 0% itself on its own app and card.A13A28A5 Company claim The borrower on a 0% loan pays principal only. Whether merchants recover the cost through prices paid by all shoppers has not been measured for Affirm by anyone. One public 2020 merchant agreement barred the merchant from surcharging Affirm users.A26 Not established
Interest exposure
About 70 cents of every dollar financed through Affirm in fiscal 2026 carried interest.
Affirm reports volume (gross merchandise volume, or GMV) in three product buckets. In fiscal 2026 GMV was $50.2 billion.A5
~70% · Interest-bearingMonthly instalment loans, 0–36% APR based on credit, terms up to 60 months
14% · 0% monthly instalmentsMostly merchant-funded at integrated merchants
16% · Pay-in-XPay in 2, Pay in 4 and Pay in 30: short and interest-free
The mix over time
All 0% products together fell from 43% of GMV in FY2020 to 26% in FY2024, then recovered to 30% in FY2026. The FQ4'26 quarter alone was 72 / 13 / 15. Affirm renamed product buckets over time, so year-to-year comparisons are approximate.A5A15 Fact
Who receives 0%
By Affirm's account, about 80% of 0% monthly volume in one 2025 quarter came from prime and super-prime borrowers, against roughly 50% for interest-bearing products.A28 Company claim A shopper who does not qualify for a merchant's 0% offer can be shown an interest-bearing offer for the same item. That is ordinary risk-based pricing, and it means “as low as 0%” describes an offer many applicants will not receive. The share who do is not disclosed. Analytical inference
By count, 0% looks larger
Because 0% purchases tend to be smaller, they are a bigger share of transactions than of dollars: 39% of transactions in the December 2025 quarter were at 0%, and more than 60% of new customers chose a 0% product for their first purchase.A14 Company claim How many customers acquired at 0% later take interest-bearing loans is not disclosed. Not established
APR and repricing
The ceiling rose from 30% to 36%. Management said why. Nothing on the record says it came back down.
This is the clearest price change in Affirm's history, and the one documented episode in which an external cost was passed to customers in management's own words.
Fixed APR; 0% at select merchants. Terms of Service.A1
No consumer fees. Letter to the FDIC.A23
Interest-bearing range, as reported by trade press. (secondary)
“0–36% APR based on credit”; Pay in 4 at 0%. Live terms.A3
Where the ceiling comes from. The maximum rate is set by the bank that originates the loan. Cross River Bank, the principal originator through fiscal 2022, permitted up to 30%. Celtic Bank's 36% limit first appears in the FY2022 annual report, and since FY2024 Celtic and Lead Bank have originated substantially all US loans at up to 36%.A5A6 Fact The originator set moved from a 30% bank to 36% banks over the same period in which the cap rose. No document states the reason. Analytical inference
First specific written statement — the cap rise and its funding-cost rationale, in the shareholder letter
Share of interest-bearing volume offered at up to 36% APR, by date. Weighted-average interest-bearing APR at origination rose 210 basis points between September 2022 and July 2023. Shareholder letters.A7A8A9A10 Company claim
“In order to offset our higher funding costs, we are in the process of passing through interest rate increases to continue responsibly extending access to credit to some consumers.”Affirm shareholder letter, fiscal Q2 2023, 8 February 2023 · text retrieved by machine extraction and re-fetched; not proof-read by eye
Two stated reasons, one letter
The same letter gave a second rationale: a higher ceiling would let Affirm approve more consumers in a stressed economy, and many would still see rates well below 36%. Management described the measures as having a “dual purpose”. The approval effect is asserted, not measured, and Affirm publishes no approval rate.A7 Company claim
Merchants shared the burden: higher merchant fees on 0% offers at select merchants, merchant-subsidised 4–6% loans in place of some 0% programmes, shorter 0% terms and more down payments. No existing loan was repriced — Affirm's loans are fixed-rate, so the increase reached new loans only. By 2024 more than 40% of origination volume in one quarter was at 30–36% APR, up from about 30% a year earlier.A12 Fact
The language followed the action
Ten per cent of interest-bearing volume was already offered at up to 36% by 30 September 2022. The first specific written statement came on 8 February 2023, at least four months later. Management itself called the late start to price increases “a key operational misstep”.A7 Fact
The cost of the move for a borrower is concrete but modest per loan: on $1,000 over 12 months, 36% instead of 30% adds about $2.97 a month, or $35.70 in total. Fact, computed
Funding cost fell
7.7% → ~5.8%Affirm's reported average cost of funds, from the September 2024 quarter to the March and June 2026 quarters.A15 Company claim
What management said would happen
On 2024 earnings calls: that lower rates would show up first as better approval rates and as reinvestment, not margin. “Pricing has also been stable,” the August 2025 letter said.A13 Margin (revenue less transaction costs) ran at 4.0% of GMV in FY2025 and 4.16% in FY2026, above the 3–4% long-term range Affirm had stated.A5
What can be said
No statement was found that the cap or interest-bearing APRs were lowered as funding costs fell. Affirm does not disclose its average APR, so whether consumer prices came down cannot be tested. Analytical inference from silence
Asymmetric repricing — up with costs, not down with them — is therefore not established. This page does not say Affirm never lowered rates.
What the price means in dollars
The only comparator on this page is a general-purpose credit card. The comparison cuts both ways, and which way depends on how the borrower would otherwise have used the card.
| Product and repayment | Term | Cost of credit |
|---|---|---|
| Affirm-style simple-interest loan, equal monthly payments | ||
| At 15% APR — the rate in every Affirm consumer example found | 12 months | $83.10 |
| At 30% APR — the old ceiling | 12 months | $169.85 |
| At 36% APR — the current ceiling | 12 months | $205.55 |
| At 36% APR, long term | 36 months | $648.94 |
| Credit card at 22.15% APR, daily compounding | ||
| Same schedule, every payment on time | 12 months | $125.15 |
| Same schedule, one missed payment and a $32 late fee | 12 months | $183.70 |
| Minimum payments only, $35 floor | 42 months | $441 |
| Minimum payments only, $25 floor | 71 months | $787 |
Like for like, Affirm at 30–36% costs roughly 36–64% more interest than an average card repaid on the same schedule. Against a card paid at the minimum, or one that triggers late fees, it costs far less. “Simple interest, no compounding” is not where the saving comes from: on a loan paid on time it is worth about $1.36 per $1,000 a year (at a 24% card rate). The protection comes from forced amortisation and the absence of late fees. Every consumer example found uses 15% APR beside a 36% ceiling; whether 15% is representative cannot be tested. Card rate: Federal Reserve G.19, accounts assessed interest, release of 8 September 2026.R7 Arithmetic computed and rechecked by the research team. Analytical inference
Funding architecture
Affirm lends wholesale money. When that money gets dearer, the price of the next loan is where it can show up.
Affirm does not fund its loans with deposits. Every dollar it lends is financed through banks, bond markets and institutional loan buyers, so the yield a new loan has to produce is set partly outside the company.A5 Analytical inference
A credit decision per purchase
The customer picks a plan; Affirm's model underwrites that transaction.
The bank originates
Celtic Bank and Lead Bank originate substantially all US loans, at up to 36%. Affirm also lends directly under state licences.
Affirm buys the loan
At face value. In FY2026 it bought $40.2bn from banks and originated $9.5bn itself.
Four channels
Warehouse credit lines, securitisations, pass-through loan sales and forward-flow sales to institutional buyers.
Investors set a price
Pension funds, insurers and asset managers. A buyer that wants more yield at renewal is how funding pressure can reach a new loan's APR or a merchant's fee.
The FY2023 squeeze. Funding costs rose from 0.45% of GMV in FY2022 to 0.91% in FY2023 and 1.29% in FY2024. Margin fell to 2.5% of GMV in the worst quarter and Affirm lost $985.3 million in fiscal 2023. On the call that followed the February 2023 letter, management said the loans Affirm sold needed to carry more revenue, in the form of higher APRs.A5A7 Fact (spoken remark, paraphrased)
Who bears what. Neither the borrower nor the merchant bears credit loss in any funding channel: Affirm carries it on the loans it keeps, buyers on the loans they buy, with some capped risk-sharing. The borrower bears the price of funding to the extent it is in the APR, and the merchant to the extent it is in the fee. Analytical inference
Counter-evidence. No existing borrower was repriced, because the loans are fixed-rate. The pass-through was late, partial and shared with merchants, who had to agree to it. Two shocks — a bond sale pulled in March 2022 (secondary source) and the bank failures of March 2023 — produced no identifiable customer change. This is one rate cycle.
Frequency
Frequency was a stated objective, and it worked: the average active customer went from two transactions a year to seven.
Affirm counts transactions per active consumer over the trailing twelve months. It is the clearest single measure of how the relationship has changed.
Lower row: active consumers. Sources: 10-K filings and the FQ4'26 supplement.A5A15 Fact
“Frequency is the word you will hear increasingly more often from us in FY'24.”Affirm shareholder letter, fiscal Q4 2023, 24 August 2023
A declared strategy
The IPO filing named “higher frequency purchases” as a growth route; the FY2021 annual report carried a growth heading about higher-frequency purchases; a May 2023 letter called frequency “the path to the top of the consumer wallet”; and the FY2024 report's heading was “Increase Consumer Transaction Frequency and In-store Usage”.A4A8A9 Fact Affirm also tells merchants that its promotional messaging and prequalification can “increase spending”.A20
Growth per customer now comes from frequency, not bigger tickets. Average order value fell from $323 (March 2024 quarter) to $273 a year later; the latest reading is about $266 (June 2026 quarter, not verified against the primary document). Repeat customers made 96% of FY2026 transactions, up from 72% in FY2021.A5
What the series cannot show
Not every counted transaction is borrowing. Whether pay-now Affirm Card purchases are included is not established, and 39% of transactions in one recent quarter were at 0%. Rising frequency therefore overstates growth in interest-bearing borrowing by an unknown amount. Not established
Satisfaction and dependence look the same in this data. A 96% repeat share with stable delinquency is what satisfied customers returning would look like; it is also what reliance would look like. There is no Affirm-specific evidence on whether its loans replace costlier card debt or add to total borrowing. This page does not infer dependence. The evidence does not allow it either way.
Independent causal studies of pay-later checkout put the spending effect at roughly 4–20%, well below the 70–85% order-value figures in Affirm's merchant marketing, which compare Affirm orders with other orders rather than measure cause. Almost all of that research concerns interest-free pay-in-four products, not Affirm's interest-bearing loans. Analytical inference
Affirm Card
The card turned Affirm from a button at one checkout into a credit option in the wallet.
The Affirm Card is a Visa debit card issued by Evolve Bank & Trust or Stride Bank. It began as “Debit+” in 2021 and was renamed and relaunched in 2023 — not withdrawn.A27A5 Fact
Pay now
Purchases draw on an Affirm Money account or a linked bank account within one to three days. Because settlement lags, even this mode involves a short advance, and getting the card is itself a credit decision. Analytical inference
Pay later, before purchase
The customer opens the app, requests a plan for an estimated amount, and uses the card within two hours.
Pay later, after purchase
An in-app feature lets the customer apply for an instalment loan on a purchase already made. The time window and minimum amount could not be read. Not established
Underwritten per transaction
Each pay-over-time plan is applied for and approved separately. Holding the card does not guarantee the ability to pay over time.
No revolving line
No committed limit, no minimum payment, no balance that rolls. No annual fee and no late fee; plans at 0–36% APR based on credit.
“Purchasing power”
The app shows an estimate of how much a customer can spend, which can change with credit and merchant. Legally an estimate; an Affirm executive has described linking a bank account as letting Affirm “raise credit limits”.
Bar: card GMV in the quarter. Lower figure: active cardholders and, where reported, the share of active consumers who hold the card. In the June 2026 quarter the card carried about 20% of Affirm's GMV ($2,842m of $14.1bn), up from about 11–12% a year earlier; 81% of active consumers do not hold it. Shareholder letters and supplements.A9A11A13A14A15 Company metrics
Mostly a lending instrument, by value
Interest-bearing loans were about 80% of card GMV in mid-2023, about 90% at the end of 2023 and more than 80% in mid-2024. The 0% share of card volume has since risen, to about 14% in mid-2025 and “almost 20%” at the end of 2025.A9A11A14 Company claim Card users transacted three times more often than the Affirm average in 2023, by Affirm's account; Affirm's May 2026 target is 20 million cardholders spending $7,500 a year, from 4.4 million at $2,400.A9A16
The pay-now share is not established
Over 40% of card transactions by count were pay-now in mid-2023, and pay-now was nearly 10% of card GMV in early 2024. Affirm has not disclosed the figure since. An unverified secondary summary of the August 2026 call on the June 2026 quarter has the card still “north of 80%” interest-bearing, which sits awkwardly beside the December 2025 letter’s 0% share of almost 20% unless the two use different bases. Not established How far the card is used as an everyday debit card is therefore unknown.
The card makes Affirm persistently available without converting the underlying credit into revolving debt.
Each plan is still a separate, closed-end, fixed-payment loan that Affirm can decline. What changes is where and how often Affirm is available — at any Visa terminal, before or after the purchase, with a standing figure in the app. The risk that distinguishes this from the pre-2021 product is loan stacking and frequency, and no published card-level credit data lets anyone outside Affirm test it. Analytical inferenceUnderwriting
Every purchase is a new credit decision. That lets Affirm say no at any point — and leaves the customer no fixed limit to plan against.
Affirm opens no revolving line. Each purchase is assessed on its own, and the price offered depends on the transaction as well as the person.
What sets the price. Affirm's annual report says rates vary with transaction risk, the consumer's creditworthiness, the repayment term chosen, the loan amount and the arrangement with the merchant. The same person buying the same item can face a different price at a different merchant or term. Down payments can also move the APR.A5 Fact
Inquiry type. The CFPB reports that all six pay-later firms it sampled, Affirm included, ran soft credit checks for pay-in-four loans.R2 Affirm's own US page says only that the rate “is subject to an eligibility check”. The inquiry type for longer monthly loans is not established in the research, and a third-party claim of hard checks on Canadian monthly plans is unconfirmed. Not established
More data, more automation. Affirm's training set grew from 7.5 million loans at the IPO to about 553 million; the model now uses more than 1,000 data points per decision and, since January 2026, bank-account cash-flow data for consumers who link an account. Automated systems choose which offers each shopper sees, with a stated aim to “maximize sales conversion and minimize risk”. Whether APR itself is varied in those tests is not disclosed.A5A14 Company claim
Price and approval are linked. Affirm presented the 36% ceiling partly as a way to approve more consumers. In November 2024 it attributed a rise in delinquencies partly to “expanded approvals with the APR range expansion”.A12 Fact
Incentives. Affirm says that because it charges no late fees it has “zero incentive to extend credit that can't be repaid”.A24 That holds for penalty revenue. Affirm also earns merchant fees at origination and sells about half its loans, so an incentive to originate exists independent of repayment. Against that, Affirm takes first loss on the roughly half of the portfolio it keeps. Analytical inference
What is not published. No approval rate, no per-consumer exposure, no count of concurrent loans. Affirm also cannot see other lenders' pay-later loans, which the CFPB calls a “blind spot” for the whole industry.R2
| Period | Credit posture | What followed |
|---|---|---|
| FY2021–FY2022 · fastest growth | Share of held balances in Affirm's top internal score band fell from 77% (Sep 2020) to 55% (Jun 2022). Timing established; cause not. | 30+ day delinquency roughly doubled; net charge-offs 8.6% of average loans held in FY2022 |
| Calendar 2022 · tightening | Shorter terms, larger down payments, income checks in some cases | Delinquency fell from 3.2% (Sep 2022) to 2.3% (Jun 2023) |
| FY2023 · repricing | Ceiling to 36%, presented partly as a way to approve more consumers | Higher delinquency later tied by Affirm to expanded approvals |
| FY2024–FY2025 | Top score band back to 65% and 69% | Delinquency stable at 2.3–2.5% at year-end |
| FY2026 | The score-band table was not found in the supplements reviewed — to be confirmed by eye | Each quarter at or above the same quarter a year earlier; 2.5% in June 2026 |
Delinquency is 30+ days on monthly instalment loans, excluding Pay-in-X and Peloton; the ceiling and score data are from Affirm's supplements and annual reports.A15 Portfolio credit held up while frequency more than doubled: delinquency stayed in a 2.3–2.8% band from 2023 to 2026. Net charge-offs were 7.4% of average loans held in FY2026, about 1.9 times the bank-card rate, though short loan duration inflates that comparison; Affirm says recent monthly-instalment cohorts are tracking toward about 3.5% losses per dollar lent.A13 Company claim Card comparison: Federal Reserve.R8 These are outcomes for Affirm's portfolio. Outcomes for households are not published. Analytical inference
When the customer pays late
No monetary penalty is not the same as no consequence.
This is where Affirm's design is strongest, and where it is most often over-read.
No monetary penalty
What a missed payment does not cost
- No late fee — $0 charged since founding, on Affirm's account, uncontradicted.A5
- No penalty APR.
- No compounding — interest is not charged on interest.
- No deferred interest waiting to land if a promotion is missed.
- Hardship changes are capped: deferrals (up to three months in total) and re-amortisations (up to twelve extra months) cannot raise total interest above the original.A5
Real consequences
What a missed payment can lead to
- A credit-file record — every product is reported to Experian and TransUnion since 2025.A21
- Charge-off at 120 days past due.A5
- Collection after charge-off — $75.6m recovered in FY2026.
- Possible loss of access to further Affirm credit — reported by third parties only, not established in an Affirm document. Not established
- Binding individual arbitration for disputes with Affirm, unless the customer opted out by posted letter within 30 days.A2
The promise was no penalty pricing. It was never no consequences — and since 2025 the main consequence reaches the borrower's credit file.
Two open points. Whether interest keeps accruing beyond the disclosed total after a missed payment is not established: no loan agreement was read, and Affirm's filings say consumers do not pay more than they agreed at checkout. And loans under a hardship modification were $24.1 million (0.25% of loans) in June 2026, beside about $237 million of loans 30 or more days late; why the figure is small is not known. Not established
Returns, refunds and disputes
After the purchase is where the customer is weakest. One purchase creates two contracts, and the gaps between them fall on the borrower.
The sale is between the customer and the merchant. The loan is between the customer and a lender, serviced by Affirm. The merchant decides whether and when to refund; Affirm controls the loan, the payment schedule and the credit-bureau record.A2 Fact
“Affirm does not refund interest that has already been paid.”Affirm Business Hub, “Fully refunding a charge”
Affirm's stated reason is that the interest accrued while the money was borrowed.A17A19 The reasoning is coherent, and a cardholder who carries a balance does not get interest back on a return either. Two things make the Affirm case sharper: an interest-bearing Affirm loan accrues from day one with no grace period, so even a prompt return leaves the borrower out of pocket — about $25 on a $1,000 loan at 30% returned after 30 days — and the terms reach interest accrued before the goods shipped. On a 0% loan a return costs nothing. Affirm also keeps its own fee from the merchant on a refunded sale. Analytical inference
Sources: Affirm Terms of Service (28 July 2026), Business Hub refund and dispute pages, developer documentation.A2A17A18A19 Affirm's US consumer help-centre articles could not be read; consumer-facing policy is reconstructed from the terms and merchant-facing documents.
Payments pause during a dispute
Under Affirm's dispute policy the customer does not have to pay while a dispute is open, Affirm withholds the disputed amount, and a merchant that loses pays Affirm a $15 fee.A18 The CFPB said in 2022 this protection was missing across the industry.
Card-style billing rights in the terms
The 2026 terms include billing-error rights and statements of the kind a 2024 federal rule required, and carry them after that rule was withdrawn in 2025. When they were introduced is not established.A2
Affirm is the judge
Affirm decides between its borrower and a merchant that is also its paying customer, typically within 15 days of collecting evidence. It publishes no dispute outcomes. The only routes beyond it are arbitration, small claims or, for card purchases, card-network rules.
Three different deadlines
The merchant policy says customers are not limited to a post-purchase deadline. The Terms of Service say to contact Affirm within 120 days of purchase. Billing errors must be raised within 60 days of the statement. The merchant hears “no deadline”; the customer's contract reserves one.
Contractual, not statutory
A credit-card holder's dispute and billing-error rights come from federal law. Affirm's equivalents come from a contract Affirm wrote and can amend by posting changes. Its right to withhold payment for a quality dispute is limited, read literally, to purchases over $50 made in the home state or within 100 miles — conditions that fit online shopping poorly, though the working dispute policy is broader.
When a merchant fails
No Affirm document read states what happens to a borrower's loan when a merchant goes out of business before delivering. A “product not received” dispute would ordinarily apply; the outcome in an insolvency is unknown.
Credit reporting
Since 2025, every Affirm loan can reach a credit file. That helps the on-time payer and records every miss.
Longer loans only
Affirm reported only the monthly instalments of longer-term loans, to Experian. When this began is not established.
No furnishing found
No announcement of reporting to the third major bureau was found.
Affirm said the new data would not be factored into consumers' traditional credit scores in the near term, but may be in the future.A21 Company claim Regulators had asked pay-later lenders to furnish both positive and negative data, and Affirm says it is the only major provider that reports all its loans.A24 Company claim
The change cuts both ways. It lets on-time payers build a record and reduces debt that is invisible to other lenders. It also makes every short plan a tradeline — at seven transactions a year, an active user accumulates many — and turns a missed Pay in 4 instalment into a mark on the file. With no late fee, the credit report is now the main consequence of falling behind. Analytical inference Credit-reporting disputes are the most common claim in the federal court dockets the research sampled; those are allegations, not findings.
Regulatory and legal record
The public record is clean of enforcement. It is not empty, and it is not proof of benign conduct.
No public CFPB, FTC, state attorney-general or state regulator enforcement action against Affirm was found in the research record.
That is a finding, and for a consumer lender of this size operating since 2012 it counts against a hostility reading. It does not prove the absence of problems: supervisory matters are not public, the FY2026 annual report's legal-proceedings item was not read verbatim, several court and regulator databases were unreachable, and the main federal regulator wound down its pay-later programme in 2025 before it produced enforcement against anyone in the sector. Fact, search-limitedCFPB market-monitoring orders to Affirm and four other providers. Its 2022 report named industry-wide concerns — inconsistent protections, data harvesting, debt accumulation — and did not single out Affirm; its late-fee concern does not apply to a lender that charges none.R1 Regulatory allegation, industry-level
Information demandCFPB interpretive rule treating pay-in-four accounts like credit cards for disputes, refunds and statements. It reached only Affirm's short interest-free products. Affirm publicly welcomed consistent standards, then objected in a comment letter to card rules. The rule was withdrawn in May 2025.R3R4
Rule, withdrawnFive US senators wrote to Affirm's chief executive, and seven state attorneys general led by Connecticut asked Affirm and five other providers about pricing, disclosures, ability to repay and disputes. Neither made an Affirm-specific allegation; Affirm's reply to the states and their status were not found.R5R6
Information requestsState pay-later laws. New York's Buy-Now-Pay-Later Act (signed May 2025) adds licensing, dispute and refund rules; law-firm summaries describe a 16% interest cap. It takes effect 180 days after final rules, which were not found. Illinois enacted a licensing framework (secondary source).R12
LegislationRate-exportation litigation. In a Colorado case, a federal appeals court is rehearing whether states can apply their own rate caps to loans made by out-of-state banks. Affirm is not a party, but its 36% rate is exported through banks in Utah and Missouri; Affirm says it voluntarily caps rates in some unnamed states.R11
Affirm not a partySecurities suits. Toole was dismissed in 2022. Kusnier, which alleged among other things that Affirm's statements were misleading because its service facilitated excessive consumer debt, was dismissed with prejudice in September 2025 for failure to plead falsity and intent; an appeal is pending. A dismissal on the pleadings is not a finding either way.R10
Dismissed · appeal pendingCard issuer. The Federal Reserve issued a cease-and-desist order against Evolve Bank & Trust, an Affirm Card issuer, in June 2024, and Affirm is a named defendant in the multidistrict litigation over Evolve's data breach; the outcome is not established.R9
Vendor bank · openAffirm Bank. Affirm applied to form a Nevada-chartered industrial bank. Consumer-advocacy and bank-industry groups filed opposing comments. No approval was found by 1 October 2026.A25
Application pendingConsumer litigation in federal courts is dominated by individual credit-reporting, debt-collection and telephone-marketing claims. These are allegations; no certified consumer class was found, and the volume of state-court and arbitration matters is unknown.
CHI institutional signals · SP001–SP004
The funding-cost episode is observable. The language did not come early enough to predict it.
CHI tests four frozen signals against a company's filings and shareholder letters, to see whether management language identifies a customer-economic change before it happens. The definitions were applied as written: an announcement on its own is not a predictive signal, and language that arrives with or after the action does not count as leading it. These are evidence tests, not a score, and they are not added together.
Charging for, restricting or metering access that used to be included.
Not supportedNo paid tier, subscription, usage cap or access fee was found. A $7.99-a-month “Affirm Plus” membership was reported in app code in 2023 and was never launched.
Drawing more revenue from existing customers through pricing, packaging, fees or upselling.
Weakly supportedThe 30% to 36% repricing is documented beyond doubt. But it applied to new loans only, no fee or premium tier was added, and the written language arrived after the action began.
Retiring, migrating or withdrawing a legacy product, feature or platform.
Weakly supportedAs to occurrence onlyAffirm Crypto (a beta), Returnly, the PayBright platform and the Debit+ name were retired or migrated. None was a core borrower product, and each was disclosed with or after the event.
An explicit link between an external cost pressure and a customer-economic action to pass it on.
SupportedPresent in one episode · not predictive“In order to offset our higher funding costs” is as explicit as this signal gets. It appeared at least four months after the higher ceiling was already in use, and only once.
On an earnings call, the finance chief lists pricing, shorter terms, approval changes and down payments as available tools. A capability, not a decision; spoken, not filed.
Language · spokenThe FY2022 annual report notes Celtic Bank permits rates up to 36%. A legal description; no intent stated.A6
Legal description10% of interest-bearing volume is already offered at up to 36%. Not disclosed until February 2023.
Action beginsThe shareholder letter calls funding costs a significant headwind. No price action is named; on the call, management says prices have not generally moved.
Cost language onlyThe explicit cost-to-price link, with the full list of measures. The rollout is 23% complete.A7
Announced mid-rolloutMore than 90% of interest-bearing volume at the new ceiling; rollout substantially complete.A10
Action completeFunding cost falls from 7.7% to about 5.8%. No letter links lower costs to lower prices.
Reverse leg absentThe funding-cost episode is observable, but the management language did not precede the action sufficiently to qualify as predictive under the frozen rules.
SP002 and SP004 describe one repricing from two angles and should not be counted twice. With one episode, reproducibility cannot be shown. The one place Affirm's language clearly led its actions — frequency and the card, named as strategy two to three years before they scaled — is outside the four frozen signals.The evidence, weighed
The strongest evidence on each side, mechanism by mechanism.
Strongest evidence in Affirm's favour
What the research could not get past
- A durable no-late-fee architectureContractual since at least 2016, repeated in every annual report, visible as the absence of any fee line, and uncontradicted in every period and market examined.
- No compounding, no deferred interest, fixed paymentsThe total cost and end date are shown before the borrower commits; there is no prepayment penalty; and no existing loan was repriced in the steepest rate rise in decades.
- Substantial 0% financing remains30% of FY2026 volume carried no finance charge, and 0% has grown faster than the platform since FY2024.
- Portfolio credit held up as frequency rose30+ day delinquency stayed in a 2.3–2.8% band while transactions per customer more than doubled.
- No major enforcement action foundNo public federal or state enforcement action, and every securities suit that reached a ruling was dismissed.
- Levers not pulledNo launched paid consumer tier, no subscription, no account fee; hardship terms that cap interest; payments paused in disputes; voluntary full-file credit reporting.
Strongest evidence against, or qualifying
What the defence could not explain away
- Higher borrower-interest exposureInterest income was 48% of FY2026 revenue and about 70% of volume was interest-bearing — while the consumer page speaks of interest on “some items”.
- A 30% to 36% pricing ceilingAverage APR at origination rose 210 basis points; more than 40% of one 2024 quarter's origination volume was at 30–36%.
- An explicit funding-cost pass-throughWith no documented reversal as costs fell — and an undisclosed average APR that means none can be tested.
- Higher frequency and persistent accessA declared strategy, now at seven transactions a year, with the card in 5.2 million wallets and no published data on concurrent loans or exposure per customer.
- Interest kept after returnsInterest already paid is not refunded, and payments can fall due before goods ship.
- Contractual post-purchase rights, decided by AffirmWhere a credit-card holder's equivalents are statutory; three different filing deadlines; no published outcomes.
- A broader ecosystemApp, card, deposit account, debit-card integrations with other banks, a bank-charter application and a planned product (Affirm Edge, pilots expected in FY2027) to embed Affirm lending in other banks' apps: Affirm increasingly sits in the wallet, not at one checkout.
Both lists are true at once. The first establishes a structurally safer contract than a revolving card, from a lender that does not earn money from customers' failure. The second establishes that the business is not what it was at its IPO and that its interest-bearing majority is not cheap. Neither establishes consumer harm.
What remains unknown
What this investigation could not establish.
These gaps are material. Most of them are figures Affirm holds and does not publish; several would settle the questions this page leaves open.
The average APR
Never disclosed in any document read; only the 2.1-point rise is. Rating-agency reports on Affirm's securitisations are the likely route, and none was read.
Whether pricing fell as funding costs eased
Cannot be tested without the average APR. The page says only that no statement of a reduction was found.
Per-customer exposure and concurrent loans
No distribution, no count of loans held at once, and no stated cap. Affirm cites an average outstanding balance of about $660 per user, unaudited.
The current pay-now share
Not disclosed since early 2024, and a December 2025 letter and an unverified summary of the August 2026 call conflict. How much card use is borrowing is unknown.
Servicing and collection detail
Help-centre text was unreadable and no loan, card or deposit agreement was obtained: accrual after a missed payment, collection agencies and debt sales, hardship approval rates, and the window for converting a card purchase into a loan.
Unresolved historical wording
The margin-target wording in the August 2026 shareholder letter is unconfirmed: extractions disagree on whether the long-standing target range was restated, and the letter has not been checked by eye. Several FY2026 annual-report rewordings, and whether the score-band table was dropped, are also unconfirmed.
Also open: Affirm's approval rate; whether merchant fees reach prices paid by all shoppers; and the content of the card agreement's reserved “optional fee-based programs”.
Conclusion
Affirm did not abandon its original customer proposition.
It still charges no late fees, no penalty rate and no deferred or compounding interest, and it shows the full cost before a customer commits. Those are real protections, and they matter most to the borrowers most likely to fall behind.
What changed is the scale around the promise. Borrower interest now pays for much of the business. The ceiling rose to 36% when funding got dearer, and the record does not show it coming back down. Through frequency targets and the Affirm Card, a lender built for one purchase at one checkout has become a credit option its average active customer uses seven times a year.
A credit system that remains unusually clean on penalty mechanics is becoming more important to its customers' everyday financial behaviour. That is the basis for a narrow investigation. It is not, on this evidence, a finding of hostility, and this page does not make one.
Sources and method
How this page was built, and how far it can be relied on.
What was done
Fourteen research streams ran in parallel, including a comparator stream and an independent stream tasked with building the strongest case against a negative reading. A reconciliation pass fixed one value for every figure used more than once and recomputed all arithmetic. Two independent verification passes made 58 primary-source spot-checks: 53 confirmed, 5 partly confirmed, none contradicted.
Limits
Quotations were retrieved by automated text extraction and have not been proof-read by eye against the rendered source, so this page paraphrases wherever it can. Risk-factor, MD&A and legal-proceedings sections of the annual reports were not read in full. Most pre-2023 call transcripts and the Wayback Machine were unavailable. Items marked secondary rest on press or third-party reporting.
Conventions
Affirm's fiscal year ends 30 June. Computed ratios were recalculated by the research team. Industry research — mostly on interest-free pay-in-four products — is not treated as an Affirm finding. The only comparator used is a general-purpose credit card; comparison with other pay-later providers is outside this page's scope. Evidence labels are explained above.
Affirm documents
Affirm Terms of Service, dated 24 February 2016 (superseded version still served by Affirm). affirm.com legal disclosuresPrimary
Affirm Terms of Service (US), dated 28 July 2026. Refunds, extended ship dates, billing rights, arbitration, amendment. affirm.com legal disclosuresPrimary
“Why Affirm” and How It Works pages, retrieved 1 October 2026. affirm.com/how-it-works/why-affirmPrimary
Affirm Holdings, Form S-1/A, 11 January 2021 (original S-1 filed 18 November 2020). sec.govPrimary
Affirm Holdings, Form 10-K for fiscal 2026 (year ended 30 June 2026), including the income statement and Note 1. sec.gov 10-K · income statementPrimary
Affirm Holdings, Form 10-K for fiscal 2022, filed 29 August 2022 — originating banks and permitted rates. sec.govPrimary
Shareholder letter, fiscal Q2 2023, 8 February 2023 — 30% to 36% ceiling, funding-cost pass-through, rollout to date. sec.govPrimary
Shareholder letter, fiscal Q3 2023, 9 May 2023. investors.affirm.comPrimary
Shareholder letter, fiscal Q4 2023, 24 August 2023. sec.govPrimary
Shareholder letter, fiscal Q1 2024, November 2023 — rollout “substantially completed”. sec.govPrimary
Shareholder letter, fiscal Q4 2024, 28 August 2024. investors.affirm.comPrimary
Shareholder letter, fiscal Q1 2025, 7 November 2024 — 30–36% share of originations; delinquencies and expanded approvals. sec.govPrimary
Shareholder letter, fiscal Q4 2025, August 2025 — merchant funding of 0%, “Pricing has also been stable”. sec.govPrimary
Shareholder letter, fiscal Q2 2026, February 2026 — 0% share of transactions, offer optimisation, card 0% share. sec.govPrimary
Earnings supplement, fiscal Q4 2026, 27 August 2026 — frequency, card, delinquency and mix series. investors.affirm.comPrimary
Affirm Investor Forum materials, 12 May 2026 — cardholder targets (wording to be re-verified). investors.affirm.comPrimary
Affirm Business Hub, “Fully refunding a charge”. businesshub.affirm.comPrimary, merchant-facing
Affirm Business Hub, Dispute Resolution Policy. businesshub.affirm.comPrimary, merchant-facing
Affirm developer documentation, refunds. docs.affirm.comPrimary
Affirm developer documentation, promotional messaging and prequalification. docs.affirm.comPrimary
Affirm, expanded credit reporting to Experian for all pay-over-time products, effective 1 April 2025. investors.affirm.comPrimary
Affirm, expanded credit reporting to TransUnion, effective 1 May 2025. investors.affirm.comPrimary
Affirm comment letter to the FDIC, 22 January 2019. fdic.govPrimary
Max Levchin, “Lawmakers looking at buy now, pay later should raise the bar”, 9 December 2025. investors.affirm.comPrimary
Affirm, applications to establish an industrial loan company, 23 January 2026. investors.affirm.comPrimary
Peloton merchant agreement, Form S-1 Exhibit 10.6 (effective 3 September 2020), section 3(c). sec.govPrimary
Affirm Card pages and cardholder agreement. affirm.com/card · cardholder agreementPrimary
Shareholder letter, fiscal Q3 2025, May 2025 — who receives 0%, self-funded 0% on Affirm's own surfaces. investors.affirm.comPrimary
Regulators, courts and official data
CFPB, “Buy Now, Pay Later: Market trends and consumer impacts”, September 2022. consumerfinance.govRegulator
CFPB, pay-later borrower report, January 2025 — soft checks, simultaneous loans, cross-lender “blind spot”. consumerfinance.govRegulator
CFPB interpretive rule on digital user accounts for pay-later loans, Federal Register, 31 May 2024. federalregister.govRegulator
Withdrawal of interpretive rules and guidance, Federal Register, 12 May 2025. federalregister.govRegulator
Letter from five US senators to Affirm, 18 November 2025. banking.senate.govLegislative
Connecticut Attorney General, multistate inquiry into pay-later lenders, 1 December 2025. portal.ct.govRegulator
Federal Reserve G.19 Consumer Credit, release of 8 September 2026 — card rate for accounts assessed interest. federalreserve.govOfficial data
Federal Reserve, charge-off and delinquency rates, release of 25 August 2026. federalreserve.govOfficial data
Federal Reserve enforcement action against Evolve Bancorp and Evolve Bank & Trust, 14 June 2024. federalreserve.gov · MDL 3127 docketRegulator · Court
Kusnier v. Affirm Holdings, N.D. Cal. No. 3:22-cv-07770, and Ninth Circuit No. 25-6869. district docket · appeal docketCourt
National Association of Industrial Bankers v. Weiser, Tenth Circuit No. 24-1293, order of 2 April 2026. law.justia.comCourt
“New York enacts first-of-its-kind law to license buy now, pay later lenders”, Mayer Brown, June 2025. mayerbrown.comSecondary
The complete research package — a dossier of roughly 55,000 words with a 388-row evidence ledger, a 308-event timeline, a legal ledger, a 131-row gaps register and two verification reports — is preserved by CHI and is not reproduced here. Affirm and related marks belong to Affirm, Inc.; no logo is used, and no endorsement or affiliation is implied. This page is research, not legal or financial advice.
Reading this beside the rest of the index
Every company here is assessed against the same question. Affirm is the one where the promise held and the relationship moved.
Read it alongside the methodology, which explains why no number appears on this page, and the other company investigations in the Money sector.