Intelligent Replenishment and Inventory Theory for Alibaba Retail Platform

Alibaba Retail, a key component of Alibaba's new retail ecosystem, aims to build an intelligent distribution platform and enable millions of small stores to embrace the digital transformation era. It connects brand merchants with small store owners via an app and provides wholesale channels, while also empowering stores through Tmall mini‑stores and POS systems to quickly respond to consumer demand.

Ensuring continuous and stable product supply is a primary goal of the supply chain. Stockouts lead to lost GMV and reduced store revenue, whereas over‑stocking increases inventory turnover time and holding costs. Therefore, intelligent replenishment is essential to balance supply and demand.

Inventory theory addresses three core questions: (1) what will demand be in a future period, (2) how much inventory should be prepared, and (3) what should the inventory level be each day. These questions are captured by the "procure‑sell‑stock" (进‑销‑存) model, which requires continuous decision making as each replenishment decision influences the next period's inventory.

The total cost at any time includes ordering, transportation, and holding costs. The objective of the supply‑chain replenishment model is to minimize this total cost while satisfying demand, which can be expressed as a function of order quantity, inventory level, and demand.

Assuming demand follows a normal distribution, the safety‑stock model introduces the concept of service level (SL). For a desired SL (e.g., 95%), the required safety stock can be derived using the error function and standard normal tables. The safety stock depends only on the service level and demand standard deviation, not on the mean demand.

Real‑world constraints complicate the simple model. Vendor lead time (VLT) must be considered, as orders are placed before goods arrive. Minimum order quantity (MOQ) and warehouse receiving capacity further restrict feasible replenishment plans. Additionally, replenishment quantities must be integer and non‑negative.

These considerations lead to an integer linear programming (ILP) formulation. Decision variables represent replenishment quantities for each SKU. Constraints enforce service‑level requirements, MOQ, warehouse capacity, and integrality. The objective is to minimize overall inventory turnover days, which serves as a proxy for total cost under the assumption of stationary demand.

Although the ILP abstracts many complexities, it captures the essence of supply‑chain replenishment. Future work will focus on demand forecasting using big‑data and machine‑learning techniques to improve the accuracy of the demand distribution inputs.