Context

Erik Torenberg hosts Affirm co-founder and CEO Max Levchin and a16z general partner Alex Rampell for a 25-year look back at fintech, from PayPal's early days through Affirm's founding to agentic commerce. Both have spent their careers inside payments, and the conversation is unusually candid about which ideas worked, which were right but too early, and why. For a PM, the value is less about fintech trivia and more about the reusable product lessons buried in the war stories: how a single up-funnel wording change unlocked Affirm's real market, why owning the customer relationship changes the entire economics of a business, and how a product built around a durable existing interface (the credit card) beats one that asks users to change behavior.

The Big Idea

In payments, the winning move is rarely a shinier interface. It is finding the moment in a purchase where your product changes the customer's decision, and building a business that owns the customer relationship from that moment forward.

Affirm did not win by replacing the credit card. It won by surfacing installment financing while people were still shopping, which lifted merchant conversion, and by taking over the customer relationship in a way merchants actively wanted. The interface stayed the same; the timing and the ownership changed.

Key Insights

1. No payments niche is under $100B

Levchin's core observation: payments is the world's largest market, and every corner of it that looks like a small niche turns out to be worth at least $100 billion. The counterintuitive twist is that revenue does not scale with transaction size. The bigger the dollar amount, the smaller the rake, because nobody pays 2% on a $40 trillion wire, while high-frequency small purchases (quick-serve restaurants, coffee) carry real margin. For a PM, this reframes market sizing: the profitable opportunity often sits in high-frequency low-value flows, not the headline-grabbing large transactions.

2. Convenience wins small, cost wins big

There is a clean rule for which attribute a payment product should optimize. As the dollar amount goes down, convenience trumps everything: at the bagel shop, if your crypto passphrase is too long, you reach for a card. As the dollar amount goes up, cost dominates: on a huge transfer you will happily spend time finding the cheapest, most secure route. Knowing where your product sits on that curve tells you whether to compete on friction or on price.

3. The credit card is a near-unbeatable interface

Levchin calls the card "the singular best user interface ever created," and the durability is the lesson. People have tried wands, biometrics, palm scanning (Amazon discontinued its palm payment), and none displaced the card, because each was only slightly faster than a thing that already works. The takeaway for PMs: beating an incumbent interface requires being dramatically better, not marginally better, and "slightly more convenient" almost never clears the bar to change entrenched behavior.

4. Up-funnel timing changed everything

Affirm's real product-market fit came from a wording and placement change, not a new feature. Early on, "pay later" appeared at checkout, and the one merchant using it complained Affirm was just cannibalizing card volume. Then Beautylish told shoppers up-funnel, while they were still browsing shampoos and perfume, that they could pay in installments. Conversion jumped 30%. That reframed the product: it was not solving the "no card on me" problem, it was solving a budget problem, and telling people early expanded what they were willing to buy.

5. Negative CAC plus customer ownership

Affirm's structural advantage is that merchants pay Affirm to bring them a customer, so customer acquisition cost is negative, and merchants actively want Affirm to own the ongoing relationship. A mattress brand does not want to send its own customer a "you're late" notice; it is happy for a third party to handle the 12 or 39 billing touchpoints of a long loan. That ownership, plus negative CAC, is what lets Affirm launch new products to 50 million-plus customers. Levchin contrasts this with Levchin and Rampell's earlier TrialPay, where merchants did not want the intermediary owning the customer and consumers had no idea who the intermediary was.

6. Hard-to-underwrite loans are a moat

Affirm is one of the few players that will write longer loans (around three and a half years, versus the typical six-week buy-now-pay-later). Doing that safely requires real machine-learning underwriting, not a FICO shortcut or "count your Facebook friends," which makes it hard for competitors to copy. The second benefit: a long loan gives many billing interactions ("39 shots on goal") to upsell the customer on new services. The hard, unglamorous capability is both the defensibility and the growth engine.

7. Every startup crosses the desert

Levchin insists there is no such thing as a company that hits product-market fit and starts selling 24 hours later. Every worthwhile startup has a long "40 years in the desert" period where it still makes sense to show up but you do not know if it ends in a great result or a great nothing. Affirm limped through that with a single grumpy merchant before Beautylish and then mattresses cracked it open. The lesson is to expect the meander and read weak early signals carefully rather than quitting.

Mental Models & Frameworks

The payment size curve

• Small amounts: convenience and interface dominate; users abandon anything with friction for a card or cash.
• Large amounts: cost and security dominate; users tolerate friction to save money.
• Revenue reality: rake shrinks as size grows, so the high-margin opportunities cluster in small, high-frequency transactions.

Use it to decide what your payment product should compete on, and to sanity-check where the actual money is before chasing the biggest-dollar flows.

Critical mass or nothing

Levchin argues payment innovation has no "okay" outcomes. A new method either reaches the tipping point where everyone must have it, or it disappears into history. His example: a car-key payment wand that made obvious sense but was only slightly faster than a card, so it never reached critical mass and died. When evaluating a network or payment bet, ask whether it can plausibly reach ubiquity, because a middling adoption result is effectively failure.

Satisfy demand versus create demand

Affirm shifted from satisfying existing demand (someone wants the bag, help them afford it) to helping merchants create and guarantee demand (a platform to launch new products to 50 million known customers). This is the long-predicted convergence of payments and advertising. Use it as a maturity lens: a payments or commerce product can move up the value chain from processing what customers already want to actively generating demand for merchants.

Pay with identity, the general-store model

The original Affirm idea was a modern version of the 1800s general store: extend credit to someone you trust even without their wallet present. Its historical cousins are the Israeli grocery tab and the Japanese business-card tab. The reusable idea is that trust plus a good underwriting signal can substitute for upfront payment, and reaching people who are not actively shopping for credit (rather than those Googling "need money") selects for better risk.

Decision Principles

Principle: Turn an industry abuse into a promise

• When: an incumbent practice quietly harms customers and everyone treats it as normal.
• Why: Affirm's no-late-fees, real-0% stance came directly from Levchin's rage at "fake 0%" deferred-interest cards, where a single late payment triggers retroactive interest back to day one. Building the honest opposite ("there will never be an asterisk on an Affirm 0") became a durable brand differentiator, not just a policy.

Principle: Sell where the merchant has margin

• When: deciding which merchants or verticals to target for a financing or fee-based product.
• Why: merchants with high gross margins and strong reasons to compel a purchase (mattresses at ~80% margin, seven-year replacement cycles) will happily fund a high merchant discount rate so the customer gets a true 0% offer. Margin structure predicts willingness to pay far better than category size alone.

Trade-offs & Nuance

Agentic payments, not agentic shopping

Levchin is bullish on agents handling the payment step but skeptical they will choose what you buy. People want to participate in purchases they care about; he will happily spend too long comparing two bike parts because he wants to. Rampell's nuance: once you already know the exact SKU, agentic commerce becomes "buy this at the lowest price," which suits people with more time than money (his example is camelcamelcamel, a top-100 US site most high earners have never heard of). The unresolved friction is trust and reliability: will the item actually arrive, from a reputable seller, on time.

We already outsource some purchases

The counterpoint they land on: grocery shopping via Instacart is effectively 100% agentic already. You tell a shopper "bring me milk" and accept their substitution without a second thought. So the behavior shift is not hypothetical; it is a question of which purchase types people will hand off and how fast, with high-consideration goods (a Friday-night outfit) being the last to go.

Common Mistakes

Mistake: Owning the customer against the merchant's will

Rampell's TrialPay connected Zynga and advertisers but neither the merchant nor the consumer wanted TrialPay in the middle; consumers had no idea who they were, and emails got blocked. The mistake is inserting yourself into a relationship the parties do not want you to own. Affirm worked because merchants genuinely wanted Affirm to take over the awkward parts (collections, late notices) of the customer relationship. Before building a B2B2C product, check whether both sides actually want you owning the end customer.

Mistake: Chasing high fees into bad risk

Affirm briefly served for-profit online education, where merchant discount rates can hit 50% because so many students default. It backed out fast, because the reason people did not pay was that the product itself (the education) was worthless, so losses were structurally high. The lesson: an unusually high fee is often pricing in an unusually bad outcome, and chasing the fee without understanding why it is high can be a trap.

Practical Application

Move your key offer up-funnel

Find the one message that changes a customer's buying decision (financing terms, a guarantee, a benefit) and surface it while they are still choosing, not at checkout. Affirm's 30% conversion lift came purely from telling shoppers earlier. Test moving your most decision-relevant information one step earlier in the funnel and measure conversion.

Map margin before you map market size

When picking which segments to sell a fee-based product into, rank prospects by gross margin and how badly they need to compel the purchase, not just by category size. High-margin merchants with strong conversion incentives (like the mattress brands) will fund offers that let you win the customer.

Build the unglamorous hard thing

Look for the capability that is genuinely hard to execute (Affirm's long-loan ML underwriting) and gives you repeated future contact with the customer. That combination, a real moat plus many upsell touchpoints, is worth more than an easily copied feature.

Audit your product for "fake 0%" moments

Find where your industry, or your own product, uses a technically-true-but-misleading practice customers resent. Consider making the honest opposite an explicit, marketed promise, the way Affirm turned "no asterisks, no late fees" into a brand pillar.

Questions to Consider

• Where in our purchase or signup flow does the customer actually decide, and are we surfacing our most persuasive information before that moment or after it?
• Which of our target segments have the gross margin and the conversion incentive to happily fund a fee, versus which are big but unwilling to pay?
• Is there a hard-to-copy capability we could build that also creates repeated contact with the customer, rather than a feature competitors can clone in a quarter?
• In any partnership where we sit between two parties, do both the business and the end customer actually want us owning that relationship?
• What "technically true but misleading" practice is normal in our category that we could turn into an honest, marketed promise?

Bottom Line

Affirm's story is a repeated lesson that product wins come from timing and ownership, not novelty: surface the offer that changes the decision while the customer is still deciding, target merchants whose margins let you give customers a genuinely good deal, and build the hard capability that lets you own the customer relationship at negative acquisition cost. The flashy new interface usually loses to the durable old one; the quiet reframe of when and how you show up usually wins.

Case Studies Mentioned

Beautylish: the up-funnel unlock

Beautylish, an online cosmetics retailer, was the first merchant to tell shoppers they could pay in installments up-funnel, while browsing, rather than at checkout. That single change produced a roughly 30% conversion increase and gave Affirm the realization that its product solved a budget problem, not a "no card present" problem. It became the template Affirm rolled out to direct-to-consumer brands.

Mattress brands: merchant-funded 0%

Casper, Purple, and similar memory-foam brands had very high gross margins and a painful seven-year replacement cycle, so they were willing to fund high merchant discount rates. That let Affirm offer customers a true 0% loan (subdividing a $1,200 mattress into small payments), which massively lifted conversion. The mattress wave was Affirm's springboard into the broader direct-to-consumer market.

The PayPal mafia

Levchin's account of why so many PayPal alumni went on to found major companies (YouTube, Yelp, LinkedIn, Palantir, SpaceX, and more): the team deliberately selected for aspiring entrepreneurs, and having known these people in their unpolished, stressed, doubtful states made their later ambitions feel achievable rather than god-like. The product-team lesson is that who you hire and how well you know each other shapes what people dare to attempt next.

Notable Quotes

"There are no niches in payments that are smaller than $100 billion." (Max Levchin)

"The card payment interface is the singular best user interface ever created." (Max Levchin)

"Convenience just trumps everything else as the total amount you're trying to send goes down." (Max Levchin)

"There will never be an asterisk on an Affirm 0." (Max Levchin)

People to Follow

Max Levchin

Co-founder and CEO of Affirm and a co-founder of PayPal, with roughly 30 years in payments and fraud-fighting machine learning. Worth following for deeply first-principles thinking on payments, underwriting, and where consumer finance meets product.

Alex Rampell

General partner at a16z and founder of TrialPay, a longtime fintech operator and investor. Known for early theses on the convergence of payments and advertising and for sharp frameworks on consumer business models like negative CAC.