Financial Recovery with Automation

Context

Payment disputes are one of the most expensive corners of a payments platform — and the recovery side of disputes is the corner of that corner where money quietly leaks. When a merchant wins a dispute, the funds owed back to them have to be pulled from the cardholder side, reconciled, and credited cleanly. Without an automated path for that recovery, every successful win turns into an operational ticket that may or may not ever close.

At Braintree, the dispute pipeline had two real problems at the top end: enterprise clients with very high dispute volumes — Amazon, Meta, Airbnb, DoorDash, Uber, Microsoft — and no merchant-facing system for recovering disputed funds. Wins existed on paper; the money didn't always make it back.

Role

Senior Product Manager. I owned the recovery workstream end to end — problem framing, the Payment API surface, the state machine inside the dispute lifecycle, the SFTP reporting contracts for enterprise clients, and cross-functional alignment across engineering, risk, ops, and the named enterprise accounts.

What we shipped

Four initiatives layered into the same recovery surface:

1. GraphQL mutations for merchant-driven recovery. Built into the Payment API so merchants could manage dispute recoveries directly — submit, track, query — with full transparency into where each case sat in the recovery flow.
2. Automated fund recovery. Designed and shipped the orchestration that pulled won funds back through Braintree's funding infrastructure without manual intervention. The merchant didn't have to chase ops to get their money.
3. Custom enterprise reporting. Tailored daily reports for Amazon, Meta, Airbnb, DoorDash, Uber, and Microsoft, delivered over hardened SFTP pipelines with the schema each finance team actually wanted — not a generic dispute dump.
4. State-machine-backed dispute lifecycle. Rebuilt the internal state model so the dispute, its evidence, and its recovery sat in one auditable flow. Every transition was logged, every recovery had lineage.

Product decisions worth writing down

Recovery is its own product. Inside the larger dispute org, recovery was treated as a downstream cleanup task. Reframing it as a first-class merchant-facing product — its own API surface, its own reporting, its own SLA — was the move that unlocked the dollar number.

The SFTP report is the relationship. Enterprise finance teams don't poll APIs. They consume files. The reporting schema we negotiated with each client became, in practice, the contract that defined the relationship — more than any product spec.

State transitions, not flags. Booleans on a dispute row are how recovery systems quietly accumulate bugs over the years. Modeling the lifecycle as a real state machine made it possible to ship new rules and exceptions later without breaking the audit trail.

Outcome

Over $10M recovered annually for merchants, sustained. The GraphQL mutations moved recovery from an opaque ops process to a merchant self-service flow. The enterprise reporting pipelines became the thing the Amazons and Metas of the world depended on at month close. Next on the roadmap: AI-driven insights to predict which disputes are recoverable before the case even closes.