Applying DDD to Refactor Keep's E‑commerce Supply‑Chain System

Any business architecture accumulates historical problems; the key is to redesign it with appropriate architectural ideas. The article presents Keep's practical DDD‑driven reconstruction of its e‑commerce supply‑chain system, originally plagued by unclear boundaries, tangled layers, inaccurate inventory, and poor extensibility.

The supply‑chain (or inventory) system manages the purchase‑→‑stock‑→‑sale chain, acting as a bridge between order fulfillment and warehouse management. Legacy systems (Skynet and ERP) became heavy, with unclear domain boundaries, mixed use‑cases, and tangled dependencies, leading to inventory inaccuracy, over‑selling, and difficulty tracing changes.

To address these issues, Keep performed three major analyses:

Inventory business scenarios

Bounded contexts

Inventory domain model (occupied, available, physical, in‑transit, frozen stock)

Based on the analysis, the team selected a hexagonal architecture to separate domain models from infrastructure, applied CQRS to isolate commands from queries, and introduced an anti‑corruption layer (ACL) to prevent context leakage.

EventStore was chosen as the event‑sourcing backbone. Domain events are published, persisted, and processed with explicit error handling. Subscription groups register and declare the events they handle, ensuring decoupled processing.

Unit testing was emphasized: domain tests for core models, command‑execution tests for business interfaces, Mockito for mocking, H2 in‑memory database for isolation, and CI‑driven automatic test execution after each push.

The refactor yielded five major benefits: accurate inventory, easier functional extensions, rapid issue localization, a reusable EventStore component, and a mature DDD best‑practice package for future system redesigns.

Overall, the case demonstrates how DDD complements micro‑service architecture by clearly defining business boundaries, enabling clean modularization, and providing a solid foundation for scalable, maintainable backend systems.