Capital One Labs

Overview

Capital One Labs was a fifty-person hardware, software, and design innovation lab built around a single question: how can we use technology to help our customers spend wisely? When I joined, my boss asked me one thing — can and should we build a credit card that can display customer spending?

The First Prototype

I built it single-handedly: a 3D-printed hollow credit card with a laser-cut lid, holding a small low-resolution screen and a push button, both wired to a battery-driven Arduino. The Arduino talked to a small Node server that streamed real-time spending data to the card.

I presented that scrappy prototype to C-level leadership and secured a multi-million dollar budget to take it into production. A PM and I then hired a nine-person cross-functional team — hardware and software — and took the product from that prototype through to initial manufacturing at Jabil in San Jose.

Two Prongs

My work ran on two tracks at once:

The card as a platform — if other Capital One teams could build apps for it, what would that product become?
The first app — personal credit spending. Do customers actually want their spending on their card? What information, and when?

Every Hat

Throughout the product development process, my role spanned:

User researcher — focus groups, interviews, user studies, experiments
Product designer — UX, UI, product vision
Packaging designer — the unboxing experience
Industrial designer — 3D printing, laser cutting, 3D modeling, material sourcing
Software engineer — a web server streaming real-time spending data
Hardware prototyper — Arduino, sensor explorations, embedded C

The Sticker Study — Prototypes as Filters

A key challenge: how do you reliably test hypotheses when your prototypes are so far from the real experience?

We knew the card-as-a-platform would eventually need a color display. But for the first iteration in personal credit, did we actually need color? Color could show spending status — red, yellow, green — but it cost significantly more and changed the screen thickness.

Instead of building the hardware to find out, I designed a simple two-week study. I set up a data pipeline that pulled each participant's real credit card spending every morning. I gave ten participants a set of colored circular stickers — red, yellow, green. Each morning, based on their actual spending against their budget, they placed a sticker on their physical credit card, right where the screen would go.

Two weeks of data told us:

The psychological effects of color in the context of spending.
Whether people actually noticed the sticker pulling the card from their wallet or using it at a terminal.
Real user privacy concerns about others seeing their status.

All of those insights, for the cost of a few stickers — no hardware development needed. The method is called prototypes as filters: take one core part of the experience and test just that, as cheaply as possible.

The Mint Scraper — Research Infrastructure

The sticker study needed real spend data, and so did everyone else. The entire lab was running research studies without access to actual user spending — Capital One's own systems were COBOL, and this was pre-Plaid.

So I built one. A simple Node.js server connected to Mint.com: test users linked a Mint account to their credit card, and the server pulled their read-only spend data every fifteen minutes, then stored it or sent it as a text or email to someone at the lab.

This little tool got adopted across the entire lab and was used in multiple projects for about a year — until Capital One finally landed a Plaid partnership. It also unlocked something internal data never could: non-Capital One test users.

"His ability to single handedly conceive, prototype, test and revise technical products means he can complete the work of multiple designers and developers."

— Adam Koepple, Product Manager, Capital One Labs