How Oliveyoung Cut Delivery Time by 14 Hours with a Multi-Center System

Oliveyoung shifted its order processing structure from a single-center model to a multi-center system, and in the process built a new delivery optimization system that determines the fulfillment center in real time. Previously, even as centers increased, orders kept concentrating on the Yangji center, and every time orders were transferred to the Gyeongsan center, manual Excel work was required, leading to repeated delivery delays and operational inefficiency.

The new system acts as a Control Tower between OMS and WMS, determining the optimal fulfillment center from order creation up to just before the shipping instruction is issued. The core of the design is as follows.

• Unified judgment layer: Inventory, per-center CAPA, and delivery region information are comprehensively judged in one place
• Control before confirmation: Validation is completed at the stage right before handing off to WMS, blocking error propagation
• Policy flexibility: A Rule-based engine allows operational policy to be reflected by simply changing DB settings
• Delivery lead time optimization: The shipping route is dynamically selected based on customer experience

Order Distribution modularizes logic using the strategy pattern, and by combining Resilience4j circuit breaker, Caffeine local cache, and Redis remote cache, it was built to make judgments within 100ms even for large order volumes. Order Transfer applies Distributed Lock, section-by-section failure isolation, and Compensating Transaction, enabling stable transfer of tens of thousands of orders, while operators can check progress and success/failure status in real time via a monitoring UI.

The results were clear. Based on comparisons during the Oliveyoung sale, the system recorded a 43.5% increase in maximum daily shipment volume, a 34.4% increase in total shipment volume, a 49.2% reduction in manual intervention, and a 100% distribution automation rate achieved. Most importantly, delivery lead time—the metric customers directly feel—was cut by an average of 14 hours, and overall lead time improved by 16.5%.

The roadmap ahead goes further.

• Advancing automatic switching between fulfillment centers and delivery methods
• A split delivery architecture that handles partial stockouts
• Accurate ETA calculation reflecting inventory, load, and region
• Multi-channel scalability encompassing MFCs, stores, and external logistics partners

Ultimately, the goal is to provide customers with a wait-free delivery experience, and to provide operators with a logistics operation environment where the system judges for itself.