DoorDash faced inventory management challenges as it moved into the Convenience and Grocery business, requiring the management of online inventory per merchant per store, with refreshes multiple times a day. To handle the scale of this inventory platform, the DoorDash team built a write-heavy inventory platform that could manage millions of items. The platform needed to be highly scalable, reliable, low latency, and offer high observability with many guardrails.
The architecture included an API controller, raw feed persistence, hydration, price calculation, and an out-of-stock predictive classification. The pipeline was asynchronous using Cadence, providing reliability and durability characteristics. The solution was incrementally improved over time, including a change from an item-level API to a batch API, database table optimization, and reducing column count. The platform was built using CockroachDB and Kafka for data storage and processing.















