Building an Inventory Data Pipeline to Address Omnichannel Stock Consistency Limitations

Oliveyoung operates a real-time inventory management system to support its omnichannel strategy connecting online and offline channels. It collects order data from in-store POS and changes from internal management systems via Kafka and CDC (GoldenGate), then processes and serves this data as data models such as store inventory, time-series inventory, and out-of-stock status.

However, as the business scale grew, the system faced three structural limitations. First, an increase in base inventory generation lead time due to growing data volume; second, a synchronous loading bottleneck caused by high coupling between data models within a single pipeline; and third, excessive Read load resulting from a Pull-based approach in which all services directly call the API.

To address these issues, Oliveyoung is pursuing a transition to an event-driven architecture (EDA) that optimizes the role of the main inventory service and distributes data according to each service's purpose, aiming to secure data consistency and system scalability.