Information-Based Strategy

Credit is a curve, not a cutoff

The framework treats a screening threshold as a curve that can be found empirically by testing, rather than a fixed line drawn in advance. It is recounted on the Cheeky Pint, the podcast hosted by Eric Glyman and John Collison, where investor Alex Rampell describes Capital One's founding playbook under Rich Fairbank and Nigel Morris.¹ In the early card world, access was a linear cutoff: a very high credit score earned a card with benefits, and everyone else got a debit card. Rampell recounts the founders' insight that the business was lucrative but not serving most of the country, and that "there must be a curve, and it's testable." Cards might go to scores above 800, but the open question was 790, then 780, and the cost of lending lower could be carried by charging a higher interest rate until a borrower proved efficacy, which Rampell calls "the BNPL of its day."

Micro-testing pockets of data

The method, in Rampell's account, was to use pockets of data to micro-test before the era of big data: a person with a given balance at a given bank receives an offer at a specific rate or with a defined no-interest window, and the mathematical return is then measured. It is a continuous experiment that finds the price of risk one cohort at a time. He describes the path as a slog, with the founders pitching many banks through the 1980s and being rejected until Signet Bank let them run it as an internal division, a decade-long build that was eventually spun out as Capital One in 1994.

Hire for slope, not intercept

Rampell draws a hiring corollary from the same story: "you want to hire for slope, not intercept." He recounts that the company did not hire banking people but hired smart people, and became the place smart people went, comparing the resulting talent lineage to a musical one in which many later fintech risk leaders trace back to it. The distinction he draws is that intercept is current credentials while slope is trajectory and learning rate, so a domain veteran may know the job well yet lack slope. This connects to hire for spikes as the general form of optimizing for raw capability and trajectory over domain pedigree, and the testing-your-way-to-truth posture connects to growth as experimentation.

Where the account acknowledges its limits

The framing carries its own tensions, which the source keeps in view. Experimental pricing of marginal borrowers is also how a lending book becomes fragile, and Rampell notes the company later had to buy a bank for stable deposits precisely because rate and credit shocks can break a lending-funded model, so the experimentation later ran within regulatory constraints. He also flags that "hire for slope" is easy to assert and hard to measure, and that the talent-lineage claim is survivorship-flavored, since the risk leaders remembered are the ones who came from the winner. The through-line the account offers is that the durable edge was not the product, which competitors could copy, but the accumulated information advantage and the talent compounding around it.