Among the fintechs that publish their numbers, "take rate" means about 0.16% of processed volume at Adyen and 8.8% of GMV at Affirm. Same two words, a fifty-five-fold gap. So when a prompt hands you "a payments company with a 2% take rate," you do not yet know what business you are looking at, and the interviewer knows you don't.
Fintech cases are decided in the first two minutes, before any framework, by three questions: who pays that rate, what is netted out of it before it becomes revenue, and who eats the loss when a transaction or a loan goes bad. I have scored candidates who built a tidy profitability tree on a volume figure they never questioned. The tree was fine. The trunk was rotten.
Below: the three questions, the four models they sort a fintech into, real filings to calibrate against, and a worked case in which a neobank's top line is set by a regulator rather than by its pricing team.
Take rate is a range, not a number
Four sets of audited accounts. Volume and revenue are as reported; each percentage is one division you can redo.
| Company and period | Volume | Revenue | Revenue ÷ volume | What the rate is |
|---|---|---|---|---|
| Adyen, FY2024 | Processed volume €1,285.9bn | Net revenue €1,996.1m | 0.16%, about 16 bps | Acquiring fee after interchange and scheme fees are passed through |
| PayPal, FY2024 | TPV $1.68tn | Net revenues $31.8bn | 1.9% | Blended checkout, processing, and P2P; transaction margin dollars of $14.7bn are 0.9% of TPV |
| Chime, 2024 | Purchase volume $115.2bn | Payments revenue $1,276.6m | 1.1% | Gross interchange, collected by two partner banks and passed on |
| Affirm, FY2025 | GMV $36.7bn | Revenue $3,224m | 8.8% | Merchant fees plus interest; 4.8 points go back out as transaction costs, leaving 4.0% |
Adyen's figure is net: interchange never touches its revenue line. Chime's is gross interchange before the network, its banks, and fraud take their share. PayPal's 1.9% is 0.9% after it pays processors and banks and absorbs transaction and credit losses. Affirm books 8.8% and sends more than half of it straight back out.
The companies say so themselves. Affirm's letter states that GMV "does not represent revenue earned by the Company." PayPal's 2024 10-K says its active-account count "may not have a direct relationship to our operating results." Filings tell you which numbers to distrust. Prompts do not; that is what they are for.
Question one: who pays the rate?
Follow a $40 debit purchase. The merchant pays a discount rate to its acquirer, which passes most of it to the cardholder's bank as interchange, a slice to the network as scheme fees, and keeps the rest. Every fintech take rate is one of those pieces, and the piece decides what the company controls.
An acquirer's take rate is the remainder, and the merchant is a customer you can price or lose to a rival at 10 bps less. A neobank's take rate is the interchange itself, set by the network's schedules and, for issuers with $10 billion or more in assets, by the Federal Reserve. Chime's S-1 says it earns most of its revenue through "interchange-based fees paid via the card networks, not paid to us by our members." Nobody at Chime sets that rate. So when a candidate proposes to "raise the take rate" at a neobank, I stop listening for a few seconds. It is like telling a utility to raise its regulated tariff.
| Model | Who pays | Netted before revenue is real | Who eats the loss |
|---|---|---|---|
| Acquirer or processor | The merchant | Interchange, scheme fees | The fintech: merchant fraud and chargebacks |
| Interchange-funded neobank | The merchant, via the network | Network fees, the partner bank's share, processing | Usually the fintech, for fraud and advance losses; the bank holds the deposits and faces the regulator |
| Lender or BNPL | The merchant and the borrower | Funding cost, credit losses, processing | Whoever holds the loan: the fintech or its forward-flow buyers |
| Software with payments attached | The merchant, twice: subscription and processing | Interchange and processing, on the payments line only | The fintech, on payments; the software line has no loss |
One question sorts a prompt into a row in ten seconds: "Who sends the company its money, and does the company set that price?"
Question two: what is netted out before it is revenue?
Every fintech reports a gross volume, a revenue, and what it keeps from a transaction after paying everyone who made it possible. The third number has a different name everywhere: net revenue at Adyen, transaction margin dollars at PayPal, revenue less transaction costs at Affirm, transaction profit at Chime, defined in its S-1 as gross profit less transaction and risk losses. The gap is not rounding. In the first quarter of 2025 Chime reported an 88% gross margin and a 67% transaction margin: twenty-one points of revenue left as transaction and risk losses before a dollar of marketing was paid.
When I hand over an exhibit with GMV on it, I am waiting for one sentence: "what sits between this and revenue?" Candidates who ask get the second exhibit. Candidates who don't compute the wrong margin with great confidence, and the sheet records that too.
The denominator gets netted as well. "Active" means a money-movement transaction in the last calendar month at Chime (8.6 million Active Members in March 2025) and a transaction in the past twelve months at PayPal (434 million accounts, and one person may hold several) and at Affirm (23.0 million consumers). Nu reports 114.2 million customers and an 83.1% monthly activity rate, then prices the difference: ARPAC of $10.7 a month against $0.8 to serve each active customer. Same word, a twelve-fold difference in window. When a prompt says "4 million active users," ask "active over what window, doing what?" The case writer had to choose, so the interviewer has an answer.
Question three: who eats the loss?
In fintech, loss is cost of goods sold, and a good prompt hides at least one kind. Fraud and chargebacks scale with transaction volume. Credit losses scale with the loan book and arrive twelve to eighteen months after the growth that caused them. Program losses on overdraft advances and small-dollar loans scale with how generous the product is.
Calibrate before you compute. US commercial banks charged off credit-card balances at an annualized 3.82% in the second quarter of 2026, down from a 4.69% peak in the third quarter of 2024. Affirm's allowance for credit losses was 5.6% of loans held for investment at June 30, 2025. A consumer lender with a 1% loss rate deserves a question about what is being underwritten, not a compliment. One at 9% deserves a question about the 2023 cohort.
Then ask about the partner. When a fintech holds no license, someone else's regulator can reach into its P&L: in June 2024 the Federal Reserve issued an enforcement action against Evolve Bank & Trust for failing to manage the risk of its fintech partnerships. And Chime's S-1 notes that transaction and risk losses rose after it launched MyPay, a pay-advance product, in July 2024, and that transaction margin will stay flat or fall as its liquidity products scale. The loss line is the product.
The regulator is a line in the P&L
Most industry cases treat regulation as a risk bullet at the end. In US fintech it is a row in the revenue build.
| Date | What happened | Why it matters in a case |
|---|---|---|
| October 2011 | Regulation II takes effect: debit interchange for issuers with $10 billion or more in assets is capped at $0.21 plus 0.05% of the transaction plus a $0.01 fraud-prevention adjustment | Small issuers are exempt, which is why US neobanks issue cards through banks under $10 billion |
| December 2015 | The EU's Interchange Fee Regulation caps consumer debit interchange at 0.2% and consumer credit at 0.3% | 0.2% of a €40 purchase is 8 cents. No app lives on that, so European neobank cases run on subscriptions, FX, and deposit interest |
| October 2023 | The Fed proposes cutting the cap to $0.144 plus 0.04% plus $0.013; comments closed May 2024 | Not adopted as of September 2026. A live "what if" in any US debit case |
| August 2025 | A North Dakota federal court vacates Regulation II in Corner Post v. Federal Reserve and stays its ruling pending appeal; a Kentucky court upholds the rule in October 2025 | The cap applies while the Eighth Circuit appeal runs. Saying "uncertain as of 2026" earns points; pretending otherwise loses them |
The Fed's own data show what the two tiers are worth. In 2024, exempt debit transactions earned issuers $0.51 each on average, or 1.21% of value; covered transactions earned $0.23, or 0.47%. Exempt banks collect about 2.6 times as much per dollar spent. That ratio is the business model of a US neobank, and it is set in Washington.
Worked case: Kestrel and the $10 billion cliff
Kestrel is invented; its figures are plausible and borrowed from no one.
Prompt. Kestrel is a US consumer fintech with 6 million customers, $18 billion of annual card spend, and a 1.4% take rate. Its CFO forecasts transaction profit of $200 million next year. Is that credible, and what should the CEO do?
The three questions, answered. Who pays: merchants, via interchange, on debit cards issued by one partner bank with $8.5 billion in assets, growing 20% a year, largely because Kestrel's deposits sit on its balance sheet. What is netted: network fees, the bank's share, and processing, 12% of revenue. Who eats the loss: Kestrel, for fraud and advance losses, 20% of revenue last year. The denominator: 2.5 million of the 6 million moved money last month. Average ticket $40.
| Line | Math | Result |
|---|---|---|
| Transactions | $18.0bn ÷ $40 | 450 million |
| Revenue (gross interchange) | 1.4% × $18.0bn | $252m |
| Cost of revenue | 12% × $252m | $30.2m |
| Transaction and risk losses | 20% × $252m | $50.4m |
| Transaction profit | $252m − $30.2m − $50.4m | $171.4m, a 68% margin |
| Per active member | $252m and $171.4m ÷ 2.5m | $101 revenue, $69 profit a year |
| Per "customer" in the prompt | $252m ÷ 6m | $42, a number that describes nobody |
Grow volume 17% with nothing else changing and $171.4 million becomes $200.5 million. The forecast is today's model plus growth. Now the cliff.
Next year. At 20% growth the bank crosses $10 billion within twelve months, and Kestrel's debit interchange drops to the cap: $0.21 plus 0.05% of $40 plus $0.01 is $0.24 a transaction, or 0.60% of a $40 ticket, against $0.56 today. On flat volume, revenue falls to 450 million × $0.24 = $108 million, down 57%. Cost of revenue tracks transactions and stays at $30.2 million; losses track volume and stay at $50.4 million. Transaction profit is $27.4 million, down 84%. A 57% cut in the top line became an 84% cut in profit because the two lines beneath it did not move. Add the 17% growth back and you reach about $32 million. Not $200 million.
Recommendation. The forecast assumes a rate Kestrel does not set will survive an event Kestrel is causing. The levers it does control: split issuance across a second exempt bank before the threshold, as Chime does, which buys time and doubles compliance cost; shift mix toward credit, which the cap does not cover, since a secured card at, call it, 1.9% earns $0.76 on the same $40 ticket, three times covered debit, and Chime's credit interchange rose from 17% of revenue in 2022 to 21% in 2024; and price what it can price, such as out-of-network ATM fees or a subscription tier, each with its own loss line and its own regulator. Do not say "raise the take rate," and say in one sentence why not.
Ned's rule. Never use a fintech number until you can say who pays it and who eats the loss behind it. If you cannot name both in one breath, it is a vanity metric, and I score it as one.
How the first five minutes are scored
No firm publishes its scoring sheet, so this is mine. Four lines carry most of a fintech case, and the opening that earns them takes under a minute: "Three clarifications before I structure. Who pays the 1.4%: if it is interchange, Kestrel does not set it, and I want to know which bank issues the cards. Is the $18 billion from all 6 million customers or from those active last month. And which losses sit with Kestrel rather than the bank. Then my structure is interchange yield, volume per active member, and the loss lines that scale with volume."
| What the sheet records | Weak opening | Strong opening |
|---|---|---|
| Clarified definitions before computing | Took "customers" and "take rate" at face value | Three targeted questions in under a minute |
| Structure fits the business model | Generic profit tree; regulated interchange treated as a price | Yield, volume per active, and loss lines that move with volume |
| Numeracy | Revenue computed on 6 million "customers" | Computed per active; checked the rate against a $40 ticket |
| Judgment in the recommendation | "Raise take rate, cut costs" | Levers Kestrel controls: a second bank, mix, priced products |
My estimate, not a study: in a first round of eight, two or three candidates ask a definitional question before structuring a fintech case. They are usually the two or three who advance.
Practice this today
One assignment, thirty minutes. Open the latest annual report or S-1 of any fintech; Adyen, PayPal, Chime, Affirm, and Nu publish theirs free. Find the volume, the revenue, and the "what we keep" line; divide; write the rate in basis points. Find how "active" is defined and note the window. Then finish the sentence "This company is paid by ___, nets out ___, and eats ___." Three companies, and the vocabulary stops sounding like revenue.
Then redo Kestrel at a $60 ticket. The cap becomes $0.25, or 0.42% of $60, against $0.84 at 1.4%: the cliff deepens to about 70%, because the cap is mostly fixed cents and the exempt rate is a percentage. Explain that in one sentence and you understand US debit better than most people who work in it.
Converting 0.16% to 16 bps to 6 cents on $40 should take three seconds, not thirty; the math drill times you on exactly that.
When the arithmetic is automatic, run a full case. The case library has fintech unit-economics cases among its 64 originals, the five-minute first rep scores three typed turns without an account, and the live voice case ends in a seven-score debrief that includes whether you clarified definitions before you built. Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
CoachNed is independent and not affiliated with McKinsey, BCG, Bain, the Big Four, or any company named here.
Frequently asked questions
What is the difference between GMV and revenue in a fintech case?
GMV, TPV, and processed volume measure the value of transactions that passed through the platform; revenue is the slice the company is paid for handling them. Affirm booked 8.8% of GMV as revenue in fiscal 2025 and Adyen about 0.16% of processed volume in 2024. Ask which one an exhibit shows before you compute anything.
How do you calculate take rate in a case interview?
Divide revenue by volume over the same period and say which revenue and which volume you used. Convert to basis points below 1%, then sanity-check on one transaction: 16 bps on a $40 purchase is about 6 cents, which tells you whether the business lives on volume or on something else.
Is a fintech case the same as a banking case?
No. A bank case runs on net interest margin, deposits, and capital; a fintech case runs on volume, take rate, activity, and loss lines. If the fintech holds loans on its own balance sheet you need both toolkits, so switch to the financial services approach mid-case, deliberately and out loud.
What is the Durbin Amendment and why does it come up in fintech cases?
It is the 2010 US law under which the Federal Reserve capped debit interchange for issuers with $10 billion or more in assets at $0.21 plus 0.05% plus $0.01. Smaller banks are exempt and collected about 2.6 times as much interchange per dollar in 2024. Most US neobanks issue through exempt banks, so the threshold is a live risk to their top line, and the cap itself is in litigation as of September 2026.
Sources
- Adyen publishes H2 2024 financial results — FY2024 processed volume of €1,285.9bn and net revenue of €1,996.1m. Checked 2026-09-24.
- PayPal Reports Fourth Quarter and Full Year 2024 Results — FY2024 TPV of $1.68tn, net revenues of $31.8bn, transaction margin dollars of $14.7bn, 434 million active accounts. Checked 2026-09-24.
- PayPal Holdings 2024 Form 10-K — active-account definition and the "may not have a direct relationship to our operating results" statement. Checked 2026-09-24.
- Affirm FQ4'25 shareholder letter, SEC exhibit — FY2025 GMV of $36.7bn, revenue of $3,224.4m, transaction costs of 4.8% and RLTC of 4.0% of GMV, 23.0 million active consumers, 5.6% allowance for credit losses, GMV definition. Checked 2026-09-24.
- Chime Financial Form S-1, amended, 2025 — 2024 payments revenue of $1,276.6m on $115.2bn purchase volume; 8.6 million Active Members and their definition; reliance on the small-issuer exemption; 88% gross and 67% transaction margin in Q1 2025; credit interchange share; MyPay loss commentary. Checked 2026-09-24.
- Regulation II: average debit card interchange fee by payment card network — the $0.21 plus 0.05% plus $0.01 cap, the $10 billion exemption, and 2024 averages of $0.51 (1.21%) exempt versus $0.23 (0.47%) covered. Checked 2026-09-24.
- Federal Reserve proposal R-1818 on the debit interchange fee cap — the 2023 proposal to lower the cap; comments closed May 12, 2024. Checked 2026-09-24.
- ABA files amicus brief urging Eighth Circuit to reverse vacatur of Reg. II — status of the Corner Post litigation and the stayed vacatur. Checked 2026-09-24.
- Federal Reserve enforcement action against Evolve Bancorp and Evolve Bank & Trust — June 14, 2024 action on fintech-partnership risk management. Checked 2026-09-24.
- Charge-off and delinquency rates on loans and leases at commercial banks — seasonally adjusted, annualized credit-card charge-off rates by quarter. Checked 2026-09-24.
- Fees for card-based payments: summary of Regulation (EU) 2015/751 — EU caps of 0.2% for consumer debit and 0.3% for consumer credit. Checked 2026-09-24.
- Nu Holdings reports fourth quarter and full year 2024 results — 114.2 million customers, 83.1% activity rate, ARPAC of $10.7, cost to serve of $0.8. Checked 2026-09-24.