Summary of Affirm's Max Levchin Breaks Down How Buy Now, Pay Later Really Works

Overview of Affirm's Max Levchin Breaks Down How Buy Now, Pay Later Really Works

This Odd Lots episode (Bloomberg) features Max Levchin, founder and CEO of Affirm, explaining the nuts-and-bolts of Buy Now, Pay Later (BNPL): why Affirm was created, how its underwriting and pricing differ from traditional credit cards and other BNPL players, how the merchant economics work, the company’s stance on reporting to credit bureaus, and practical uses of AI inside the business. The conversation mixes Levchin’s personal origin story (PayPal → credit-card pain → Affirm) with concrete product, regulatory, and competitive detail.

Key takeaways

• Affirm was founded to fix the opaque, compounding, fee-heavy structure of traditional credit cards—principally by offering fixed, transparent repayment plans and banning late fees and deferred-interest tricks.
• Core product/mission principles: no hidden fees, no interest compounding into principal, clear payment schedules and per-transaction underwriting.
• Underwriting is real-time and individualized: uses credit bureau data when available, plus cash-flow insights from bank account data and merchant/context data (item type, useful life). Levchin stresses many small, legal variables rather than a single “magic” signal.
• Affirm furnishes both positive and negative repayment data to credit bureaus (unique among major BNPLs). Levchin argues all BNPL lenders should do this to avoid stacking and to reward on-time borrowers.
• Merchant economics: merchants pay fees to Affirm (range: from below card rates up to low single-digit percentage points when merchant absorbs interest). Merchants care about fee, approval rates (incremental sales), and protecting brand reputation (no predatory collection practices).
• Consumer metrics: average Affirm transaction ≈ $300; strong retention (high percent of repeat customers); Levchin claims delinquency materially lower than credit cards and a large share of customers pay on time.
• AI is in production: used heavily in customer service automation, contract and ad copy review, internal workflows (finance/legal); AI augments staff and enables specialization rather than wholesale layoffs.
• Levchin is skeptical that stablecoins/crypto/brand-stablecoins currently unlock meaningful benefits for Affirm’s core business.

How Affirm’s BNPL works (product & underwriting)

• Two product flavors:
  • Short-term, interest-free installment plans (merchant typically covers the interest).
  • Longer-duration loans (12–36 months) that can be zero interest (merchant subsidized) or consumer-paid interest.
• Design rules: fixed payment schedules disclosed up-front, no late fees, no hidden or deferred interest.
• Underwriting process:
  • If sufficient bureau data exists: use custom credit model to render an instant decision.
  • If bureau data is sparse: request access to bank account transaction flows to assess cash inflows/outflows and affordability.
  • Merchant-provided info (what’s being purchased) influences underwriting (useful life vs. repayment term matters).
  • Compliance: avoid prohibited-basis attributes (race, gender, etc.); avoid variables that act as proxies for prohibited attributes.
• Decisioning is per-transaction (real-time) — Affirm approves or declines at the point of purchase.

Business model & economics

• Revenue sources:
  • Merchant fees (often the primary revenue when merchant subsidizes interest).
  • Consumer interest when consumer elects a financed plan that includes interest.
• Cost/capital:
  • Affirm funds loans via capital partners; cost of capital adjustments are typically smoothed in contracts (changes pass through gradually).
  • If funding costs rise, levers include passing costs to merchants, adjusting consumer pricing for interest-bearing loans, and contracting terms with capital providers.
• Merchant value proposition:
  • Fee vs. incremental sales math: merchants weigh paying a fee to capture a marginal buyer they otherwise would lose.
  • Approval rates and brand safety are often more important than headline fee rates.
• User behavior:
  • High repeat usage — Levchin describes strong loyalty and frequency (multiple transactions per year for active users).
  • Average ticket ~ $300; use cases include higher-ticket consumer goods and occasionally groceries/party purchases.

Risks, stacking, and regulation

• Stacking (customers borrowing across multiple BNPL providers) is a concern industry-wide.
• Affirm’s approach: reporting both on-time and late payments to credit bureaus to create a permanent record, which should reduce stacking risk and reward responsible borrowers.
• Levchin’s critique: some BNPL providers may avoid furnishing data because their revenue depends on late fees; reporting reduces that source of profit.
• Levchin calls for broader industry adoption of credit reporting and stronger regulatory clarity around underwriting variables and disclosure.

Technology & AI applications

• AI is used in production at Affirm for:
  • Customer support automation (handling high-volume, straightforward inquiries).
  • Contract and ad copy review to ensure merchant advertising of Affirm terms is accurate.
  • Helping internal teams (finance/legal) process and monitor enormous volumes of custom contracts and merchant statements.
• AI is positioned as augmentative: it handles routine tasks and enables human specialists to focus on more complex issues; Affirm has not used AI to indiscriminately cut staff.
• Engineering, finance, and legal are major AI users internally.

Product innovations: the Affirm Card

• Affirm issues a physical/digital card that can switch modes:
  • Debit mode: pay now (debit from bank).
  • Credit mode: set up a loan in the app and the card then charges the loan when tapped.
• Adoption: Levchin reports quick organic uptake among users (no heavy mass-market ad spend).

Views on crypto / stablecoins

• Levchin is skeptical that dollar-pegged stablecoins currently solve material problems for Affirm:
  • E‑commerce is mostly domestic; cross-border gains are limited today.
  • Rewards schemas and user behavior don’t naturally lead to consumers holding multiple branded dollar-pegged tokens.
  • If cross-border commerce grows materially, use cases could change, but not yet compelling.

Notable quotes

• “We put two principles down: we will never change the plan, and we will not charge late fees.”
• “Make a plan you commit as a lender that you'll never change and don't charge late fees. And that's it.”
• “If you are a BNPL provider and you're not furnishing data, you kind of have no excuse. Furnish your damn data.”
• “We don't make nearly as many mistakes, or perhaps we are not willing to let people go late because we don't benefit from it.” (on lower delinquency)

Actions & recommendations (for different audiences)

• Consumers: prefer BNPL providers that disclose full schedules up-front, avoid late fees, and report positive history to credit bureaus (those records can help build credit).
• Merchants: evaluate BNPL providers on approval rates and brand protection, not just headline fees—better approvals can drive incremental sales that justify higher fees.
• BNPL providers: report on-time and late payments to credit bureaus to reduce stacking and improve long-term underwriting accuracy.
• Regulators/policymakers: consider rules for disclosure/reporting and standardized underwriting transparency to protect consumers and reduce perverse incentives (e.g., late-fee-driven revenue).

Bottom line / Verdict

Affirm’s pitch is that BNPL can be a consumer-friendlier alternative to credit cards if built with transparent pricing, per-transaction underwriting, and an alignment of incentives (no late fees, no compounding). Levchin emphasizes industry responsibility—particularly around furnishing bureau data—and practical, measured uses of AI. The product is positioned less as a payments novelty and more as a principled redesign of consumer credit rails.