Mar 8, 2019, 9:35 PM ETFireside Chat, Interview

3 Ways Startups Are Coming for Established Fintech Companies -- And What To Do About It

Unlike most industries where "more customers equals more revenue," financial services operate on risk pools where adding unvetted customers can decrease profitability due to the "coin flip" nature of underwriting.
Insurance and lending rely on a cross-subsidization model where profitable "good" customers (e.g., healthy gym-goers, safe drivers) subsidize unprofitable "bad" customers (e.g., those prone to illness or accidents).
Incumbents often suffer from negative economics if they cannot distinguish risk, forcing them to charge average rates that drive away the best customers and leave them with a portfolio of losing propositions.
Startups can exploit the psychological unfairness of the current system where low-risk individuals pay higher average rates to subsidize high-risk individuals.

Startups attack incumbents by using "positive selection" to target the "good" end of the risk distribution curve, leaving incumbents with the "bad" end.
SoFi pioneered this with "HENRYs" (High Earning, Not Rich Yet), offering lower rates to borrowers with high employment prospects who were previously priced identically to high-default-risk students.
Health IQ utilized behavioral data (e.g., running a 9-minute mile, bench pressing body weight) to identify long-living individuals, offering them significant discounts (up to 50%) that incumbents could not match.
Debt Settlement Firms (e.g., Freedom Financial) use "willingness to repay" as a selection criteria, negotiating settlements with creditors where borrowers prove they want to exit debt, creating a symbiotic relationship with creditors who prefer partial recovery over default.
This strategy allows startups to operate profitably with fewer customers by eliminating the "bad" risk entirely rather than cross-subsidizing it.

Traditional lending relies on FICO scores and credit bureaus, creating a "bin and ball" problem where lenders cannot distinguish risk without charging usurious rates or rejecting borrowers entirely.
New lenders bypass regulatory interest rate caps (e.g., Utah's 36% cap in the US) by using alternative data to price short-term, small-dollar loans where the absolute interest cost is low, even if the APR appears high.
Branch utilizes phone data (e.g., number of apps installed, battery consumption patterns, text message volume) to identify counterintuitive correlations between behavior and creditworthiness.
Machine learning models detect "strange correlations" in high-dimensional data (9,000+ dimensions) that human underwriters cannot perceive, allowing for precise pricing of unbanked or thin-file customers.
Lenders use "induction" on loan amounts (e.g., $1 → $2 → $4) to test a borrower's "willingness to repay" over time, building proprietary credit histories for customers ignored by traditional banks.

Startups move beyond static risk assessment to actively change customer behavior to improve profitability over time.
Earnin provides access to earned wages before payday, using real-time streaming data of work hours and location to verify employment, removing the need for traditional credit checks.
Dynamic Underwriting allows insurers to re-price policies based on real-time telemetry (e.g., monitoring a car's speedometer to penalize reckless driving immediately after signing up).
Social pressure mechanisms, such as Vouch, leverage social networks where friends co-commit to loans, making default socially costly rather than just financially costly.
Community-based accountability (e.g., weight-loss challenges using public scales) is adapted to lending to encourage repayment through peer monitoring rather than collection calls.

Sub-Branding: Incumbents should create distinct sub-brands for specific demographics (e.g., "Mormon Car Insurance" or specialized brands for safe drivers) to avoid diluting the core brand and to capture positive selection in niche markets.
Turn-Down Traffic Monetization: Instead of simply rejecting applicants, incumbents should partner with startups to offer rejected customers to niche lenders who have better underwriting models for those specific profiles, monetizing the customer acquisition cost already spent.
Acquisition Strategy (Facebook Model):
- Buy existential threats early, accepting high valuations for probabilistic upside (e.g., buying Instagram/WhatsApp).
- Acquire failed startups ("failed triers") specifically for their talent, not their product, as they possess the process and tenacity required to innovate.
Talent Integration: Place leaders from failed startups in charge of successful teams or new divisions to leverage their "process-oriented" mindset over the incumbent's "outcome-oriented" culture.
Investment Timing: Late-stage investment is often preferable to early-stage for incumbents; early deals may suffer from adverse selection where the best assets are overpriced, while later deals have more validation.
Distribution Leverage: Incumbents must utilize their massive distribution and cost-of-capital advantages to support startups, acting as a partner rather than a competitor by sharing traffic and infrastructure.