TL;DR
This paper tackles the challenge of optimizing offline physical store locations for online retailers adopting a "Buy Online and Pickup in Store" (BOPS) omnichannel strategy. By modeling customer channel choice behavior using the Multinomial Logit (MNL) model and developing an improved genetic algorithm, the research provides a framework to minimize total enterprise costs while meeting diverse customer demands.
What Problem Does This Paper Address?
The retail landscape is rapidly evolving towards omnichannel models, where customers can interact with businesses across various channels. For online retailers, a key strategy is implementing BOPS, allowing customers to purchase online and pick up items at physical stores. However, determining the optimal locations for these physical stores is complex. Traditional location optimization models often overlook crucial customer behavior, such as their preference for different channels (online, BOPS, or traditional offline purchase). This paper addresses this gap by explicitly incorporating customer channel choice behavior, modeled using the MNL framework, into the physical store location optimization problem. The goal is to find a store location strategy that minimizes overall enterprise costs, including store construction, transportation, and return penalties, while satisfying customer demand across all channels.
Methods
The core of this research lies in two main components:
- Customer Choice Modeling: The Multinomial Logit (MNL) model is employed to capture how customers choose between three distinct purchase channels: the purely online channel, the BOPS channel, and the traditional offline physical channel. The utility functions for each channel are defined based on factors like freight costs, return rates, pickup distance, service levels, and shopping time.
- Location Optimization Model: A mathematical optimization model is constructed with the objective of minimizing the total enterprise cost. This cost comprises:
- Construction Cost of Offline Physical Stores: The expense associated with building new physical stores.
- Vehicle Transportation Cost: Costs associated with delivering goods from distribution centers to physical stores (large vehicles) and from physical stores to demand points (small vehicles).
- Penalty Cost of Return: Costs incurred due to unsatisfied customer demand and product returns from the online channel.
- Improved Genetic Algorithm: To solve the complex optimization model, an improved genetic algorithm is designed. This algorithm features enhancements in chromosome selection, crossover, and mutation strategies to improve convergence speed, avoid local optima, and achieve better overall solutions compared to traditional genetic algorithms.
Datasets
The study uses a simulated dataset for verification and analysis. This dataset includes:
- 30 Demand Points: Characterized by their coordinates and the quantity of goods demanded.
- 10 Alternative Physical Store Locations: Defined by their coordinates.
- 1 Distribution Center: Specified by its coordinates.
The paper does not introduce a new benchmark dataset but rather uses synthesized data to demonstrate the model's effectiveness.
Key Insights and Ideas
- Customer Choice is Paramount: Ignoring customer channel preference significantly impacts the effectiveness of physical store location decisions. The MNL model provides a robust way to integrate this behavior.
- BOPS and Offline Channels Dominate: In an omnichannel setting, the BOPS and traditional offline channels collectively account for a substantial portion of customer demand (around 72% in the example), highlighting their importance for retailers.
- Strategic Freight and Service Levels: Increasing freight costs for the online channel and improving the service level of physical stores can effectively shift customer demand towards BOPS and offline channels, thereby reducing overall enterprise costs.
- Return Rate as a Driver for Physical Stores: A higher return rate in the online channel makes expanding offline physical stores more necessary and ultimately leads to lower total enterprise costs. This suggests that physical stores can act as a buffer and a solution for online channel inefficiencies.
Key Results
- The optimal location scheme identified four physical stores (nos. 2, 7, 8, and 9) that effectively serve customer demand across all channels.
- The study demonstrates that with the proposed location scheme, BOPS and offline physical channels account for approximately 72.03% of the total demand.
- Sensitivity analysis reveals that increasing freight levels and physical store service levels can reduce total enterprise costs.
- A higher online return rate necessitates more physical stores and leads to lower overall costs.
- The improved genetic algorithm showed faster convergence and higher optimization efficiency compared to the traditional genetic algorithm, achieving a lower optimal total cost and average running time.
Limitations and Future Work
The authors acknowledge several limitations and suggest avenues for future research:
- No Product Shortage Assumption: The current model assumes no product shortages in any channel. Future work could incorporate scenarios where stockouts in one channel lead customers to switch to others.
- Absence of Competitors: The model does not consider the presence of competing enterprises. Future research could integrate competitive analysis into the consumer choice and location models.
- Alternative Choice Models: While the MNL model is effective, exploring other consumer choice behavior models could provide further insights.
- Dynamic Factors: The study could be extended to consider more dynamic aspects, such as fluctuating demand or evolving customer preferences over time.
Summary updated