The Optimization Method For Inventory Placement Of Related Product Category With Multi-Warehouse System In Large-Scale Online Supermarkets

Large-scale online supermarkets always apply for the multi-warehouse network systems,which results in the so-called order splitting problem that is a multi-item customer order is often split into several sub-orders and fulfilled by different warehouses.Order splitting has been a common phenomenon in practical operations.Serious order splitting problem with multiple deliveries of multiple packages for one order intrigues some challenges such as high fulfillment cost for online retailers,high delivery disturbance for customers,high environmental pollution for the whole society.The decision to optimize the vast products inventory placement between multiple warehouses is the basis and the first procedure of order splitting fulfillment.The quality of this decision will directly affect the efficiency of subsequent order fulfillment(including order picking,packing,distribution,etc.).In practice,the complexity of this problem is that the super large-scale of the problem,which is due to the huge customer orders(average more than one million orders per day)and the vast variety of products(total more than one million products),and the explosive increased number of feasible product categories placement solutions with the increase in the number of orders and the number of items in each order.Thus,the problem of online supermarket product categories related-placement in multiple warehouses has several characteristics: multiple influence factors,complex decision space structure,and huge solution space,which is a complex large-scale combination optimization problem and is difficult to be solved by existing theoretical methods.We propose the method of optimizing product categories placement in multiple warehouses and regard the product categories as the link between orders and warehouses.For the sake of decreasing the solution-space to better matching orders and warehouses and using the clustering idea,we try to place the most related products into the same warehouse.It can effectively solve the problem of huge solution space and complex relationships among decision elements brought by massive orders and products,which can further reduce the order fulfillment cost,and provide a novel idea for large-scale online supermarkets’ order splitting problem from the source fulfillment procedure.This research synthetically applies the theories of combinatorial optimization,multi-commodity network flow,and graph theory.Following the procedure as “How to place product categories between multiple warehouses → Which product category to place in each warehouse” studying the problem,we formulate a multi-commodity network flow model for the related product category placement problem of the multi-warehouse network system,transform the order splitting minimization of the related product category placement problem to a special clustering problem,and develop a high efficient K-links Clustering Algorithm(K-LCA)for generating near-optimal solutions of the related product category placement problem of the multiwarehouse network system with the smallest splitting orders to fulfill the customer orders.This work can help achieve effective plans for supporting the decision-making process of online order fulfillment.The research includes following the aspects:(1)Problem statement and decision-making analysis of related product category placement for large-scale online supermarkets with the multi-warehouse network system.First,this section describes the order fulfillment process and demonstrates the reason for order splitting in largescale online supermarkets,clarifying the necessity of product category placement optimization in a multi-warehouse network system for solving order splitting.Then,it analyzes the key factors(orders,products,warehouses,etc.)that affect the decision-making of product category placement and summaries the relevant characteristics,verifying the reasonability of studying the product category instead of the product stock keeping unit.Finally,it analyzes the process of placing product category decisions in multiple warehouses,clearly describes and defines the problem,and makes a complexity analysis of its decision optimization problem.(2)An integer programming optimization model for related product category placement in large-scale online supermarkets with the multi-warehouse network system.First,this section analyzes the idea of the model construction for product category placement problem using the theory of multi-commodity network flow.Then,through analyzing the order fulfillment cost,it formulates the mathematical programming models with the objective of minimizing the number of order splitting.Finally,it analyzes the complexity of the model which is an NP-hard problem and proposes the heuristic solution idea.(3)A designed heuristic clustering algorithm for related product category placement solutions in large-scale online supermarkets with the multi-warehouse network system.First,this section defines a new product categories correlation index “link” based on the analysis of the fixed-charge multi-commodity network flow model,and transforms the objective of minimizing splitting orders in the model to the criteria of minimizing the total out-links between multiple warehouses in the algorithm.This paper proposes a K-links Clustering Algorithm(KLCA)to solve the product category placement problem.It can quickly and effectively generate near-optimal product category placement solutions of a multi-warehouse system in online supermarkets.(4)Numerical experiment and case study.Based on the characteristics of a large online supermarket in China and analysis of the actual order and product data,this section presents numerical experiments to verify the effectiveness and optimality of the K-LCA product category placement approach proposed in this paper,as well as the applicability and efficiency of the algorithm in dealing with the real large-scale product category placement problem.According to the computational results,management insights are provided for the warehouse operation practice of the e-commerce retail industry.This research conducts a useful exploration for solving the product category placement problem for reducing splitting faced by large-scale online supermarkets with a multi-warehouse network system.Moreover,the solution procedure provides a new theoretical method for solving such combinatorial optimization problems with large scale,complex decision structure,and huge solution space,which is conducive to improving the scientificity and practicability of complex management decision-making problems.From the perspective of application value,the research results provide a new solution idea and decision support for the order splitting fulfillment optimization problem in an online supermarket with a multi-warehouse network system.It can effectively reduce the number of splitting orders,reduce the cost of order fulfillment,greatly improve the online supermarket order processing efficiency and service level,and help to improve customer satisfaction.Furthermore,it can reduce carbon emissions and decrease the possibility of environmental pollution due to the decrease of the order splitting resulting in fewer abandoned express cartons,which is useful for China’s e-commerce retail industry towards precise,efficient,and low-carbon sustainable and healthy development.