Credit in Minutes, Not Weeks — Alt-Data Underwriting at Kabbage

When I joined Kabbage in 2012, the company was still proving a radical thesis: that a small business's live financial and social footprint could tell you more about creditworthiness than a FICO score ever would. Traditional banks took weeks to decide. We aimed for minutes — and built the platform, the models, and eventually the teams that made it real. By the time I left, engineering had doubled, three teams were shipping on a weekly cadence, and the underwriting engine I architected was the core of the lending platform.

The Problem Traditional Banks Couldn't Solve

Small business owners needed working capital to buy inventory, make payroll, or seize a growth opportunity. Banks wanted personal credit scores, collateral, and paperwork — a process measured in weeks, with a hard cutoff that excluded millions of viable entrepreneurs. Kabbage saw an opening: connect to a business's accounting, banking, and e-commerce data, supplement it with public social engagement signals, and underwrite in the time it takes to finish a cup of coffee.

From Engineer to Team Lead

I joined as an application developer. As the platform grew, the bottleneck shifted from writing code to coordinating the people writing it — so I stepped into leadership as an engineering manager: still hands-on, but now responsible for the people, practices, and release cadence that kept our platform trustworthy at scale.

We built on a service-oriented architecture: ASP.NET MVC for customer-facing and internal applications, REST services composing the platform core, and a data integration layer connecting to 15+ external platforms — Amazon, eBay, QuickBooks, PayPal, UPS, social APIs — that turned live business activity into a proprietary underwriting asset. From the start we invested in CI/CD pipelines and automated test harnesses, because a lending platform that goes down or ships a bad deploy has real consequences for real businesses.

Alt-Data Risk Models

The underwriting engine combined structured financial data with engagement metrics from a business's Facebook or Twitter presence — signals of legitimacy: Is this a real, active business with customers who interact with it? Combined with cash-flow patterns from connected accounts, our models could approve or decline in minutes while maintaining profitability across a portfolio that grew to serve thousands of SMBs.

That dataset — built transaction by transaction — became one of Kabbage's most durable assets, powering follow-on products long after the initial line-of-credit offering proved the model.

Scaling the Organization

As the company grew toward unicorn status (and eventual acquisition by American Express), I helped scale engineering from a handful of developers to a multi-team organization:

• Maintained a weekly release cadence across three teams running staggered three-week sprints
• Doubled engineering capacity while reducing production incidents by roughly 30%
• Established engineering practices — code review, automated testing, deployment safety — that outlasted any single project

Impact That Outlasted the Exit

Kabbage didn't just move fast; it moved capital to business owners that banks wouldn't touch — first-time entrepreneurs, seasonal retailers, online sellers without decades of credit history. The work directly supported the growth of thousands of small businesses and helped prove that alternative data could power underwriting at scale.

Scaling a platform, a dataset, and an engineering organization at the same time — without losing release discipline — became a playbook I've re-run ever since: first at SunTrust, teaching a century-old bank to move at startup speed, and today for companies making the jump from a handful of engineers to a real engineering organization.