a16z Podcast: Fintech for Startups and Incumbents

Episode Summary

Executive Summary: Alex Rampell argues startups can attack incumbent fintechs in three ways: by cherry-picking the best customers using psychological and pricing wedges, by using new data sources/ML to underwrite more accurately, and by changing customer behavior so previously unprofitable users become profitable. Incumbents should respond with sub-brands, dynamic underwriting, and “turn down traffic” partnerships rather than chasing one-size-fits-all scale.

Main Topics: Wedge 1: Positive selection via customer psychology (Priority: 5/5): Startups can attract the healthiest, safest, or most creditworthy customers by making pricing feel fair and differentiated, leaving incumbents with a worse risk pool. Wedge 2: New data sources and smarter underwriting (Priority: 5/5): Startups can use alternative signals—especially mobile data and behavioral traces—to estimate willingness/ability to repay better than legacy models and overcome coarse pricing constraints. Wedge 3: Changing behavior over time (Priority: 4/5): Companies like EarnIn, microfinance models, and usage-based insurance can nudge or monitor behavior so initially risky users become better risks. Incumbent response: sub-brands and segmentation (Priority: 5/5): Large firms should not use a single brand for all customers; they should create targeted sub-brands aimed at distinct low-risk segments and price accordingly. Incumbent response: turn-down traffic partnerships (Priority: 4/5): Rejected customers can be routed to startups, creating a monetization channel for the incumbent and a customer-acquisition channel for the startup. M&A and acqui-hire strategy (Priority: 4/5): Incumbents should buy existential threats early or acquire failed startups for their talent/process, especially when the startup solved a problem the incumbent still lacks.

Key Arguments: In financial services, more customers is not always better; in insurance and lending, each incremental customer can be a coin flip between profit and loss. Startups can exploit the psychological unfairness of equal pricing by offering lower rates to demonstrably better risks, which improves marketing and risk selection simultaneously. HealthIQ and SoFi are examples of startups that won by identifying and appealing to premium customer segments that incumbents priced too broadly. Alternative data can replace blunt underwriting and help lenders avoid charging illegal or excessive rates when they cannot distinguish good borrowers from bad. Fair-lending rules and adverse-impact concerns constrain the use of some variables in the U.S., pushing innovation to other markets or more carefully structured models. Branch-style lending shows how phone/app/usage signals can reveal hidden creditworthiness; more information can reduce rates and expand lending safely. EarnIn illustrates behavior-shaping fintech: it lends against earned wages, uses real-time data, and incentivizes better outcomes without traditional fees or interest. Big companies should create multiple sub-brands because a single umbrella brand cannot credibly target both elite low-risk customers and risky edge cases. Turn-down traffic is a practical partnership model: the incumbent sends rejected customers to startups, recapturing value from otherwise lost traffic. For M&A, incumbents should value process and problem-solving ability over raw outcomes when a startup tried to build something strategically important but failed. A startup that failed because of lack of distribution may still be highly valuable to an incumbent that already has distribution and capital. Incumbents should not wait too long to invest if top startups are competitive, but overpaying can still be rational when an existential threat is possible.

Data Points: U.S. life expectancy example: 79.6 years - Used as a rough average in discussing risk pooling and mortality distributions. Car insurance discount example: 50% discount - Illustrative price cut a startup might offer to best-in-class drivers to attract low-risk customers. Geico advertising spend: $1.2 billion per year - Cited as the scale disadvantage incumbents can turn into a distribution asset via turn-down traffic. Usury cap example: 36% APR - Utah cap mentioned as a constraint on lending rates and a driver of alternative underwriting. Hypothetical short loan APR: About 9,000% APR - A $9 loan repaid with $10 after a few days was used to show why APR can mislead for tiny short-duration loans. Illustrative loan ladder: $1 → $2 → $4 → $10 - Branch-style credit ladder example showing how repeated repayment can expand trust and loan size. Branch model: Phone data and app signals - Branch was described as using mobile-device behavior and unusual features to predict repayment. Behavioral health target: Lose 6%–7% of body mass - Example of a health intervention where peer pressure/group accountability improves outcomes.

Pivotal Quotes: "you actually want one-tenth as many customers as Geico" — Alex Rampell: Explaining that a startup may be more profitable by serving only the best risks rather than maximizing customer count. "you have to over-allocate on process and you want to weight outcome to almost zero because you're buying the outcomes that were in fact zero" — Alex Rampell: Describing how incumbents should evaluate failed startups they may want to acquire for talent and process. "turn down traffic strategy" — Alex Rampell: Naming the approach where incumbents redirect rejected customers to startups instead of losing them entirely.

Implications: Fintech incumbents must rethink scale: profitability comes from selecting, pricing, and shaping risk—not just adding customers. Winning strategies include segmentation, alternative data, dynamic incentives, and strategic partnerships or acquisitions.