Assortment Pack Allocation Using Cluster-Based Retail Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining an optimal packaging and distribution of item assortments through a supply chain to balance shipping cost and timeliness while meeting predetermined ideal allocations at each store is challenging, especially in large retail enterprises with numerous locations.
Innovation Solution
A scalable ensemble optimization approach utilizing k-means clustering, simulated annealing, and linear programming to efficiently determine optimized assortment pack configurations and allocations, minimizing computation time and error while ensuring accurate demand fulfillment across multiple retail stores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If custom assortment packs are created for each individual retail store, then product availability and demand fulfillment are optimized, but device complexity and computation time increase significantly
Solution Approach 1:
The patent segments the retail store population into clusters based on demand similarity, creating groups of stores with comparable product requirements. This allows the system to develop a manageable number of standardized pack configurations that can serve multiple stores, rather than creating unique configurations for each store. The segmentation reduces complexity while maintaining the ability to fulfill specific demand patterns for different store types.
Solution Approach 2:
The system changes the parameter of pack configuration from store-specific to cluster-based, transforming the optimization problem from thousands of individual configurations to a smaller set of standardized configurations. This parameter change enables the system to balance product availability with computational feasibility by adjusting the granularity of customization.
2Measurement precision
If optimized pack configurations are developed for each store, then allocation accuracy improves, but computation time increases by over 80%
Solution Approach 1:
The system performs preliminary clustering of stores based on demand patterns before developing pack configurations. This preliminary action groups stores with similar requirements together, allowing the optimization algorithm to work with representative cluster configurations rather than every individual store. This reduces computation time while maintaining allocation accuracy through the use of cluster centroids that capture the essential demand characteristics.
Solution Approach 2:
The patent uses cluster centroid configurations as templates or copies that can be applied to multiple stores within each cluster. Instead of performing independent optimization for each store, the system creates standardized pack configurations based on cluster characteristics and allocates them to member stores. This copying approach dramatically reduces computation time while preserving allocation accuracy through the representative nature of cluster centroids.
3Device complexity
If a universal pack configuration is used across all stores, then device complexity is reduced, but allocation accuracy and demand fulfillment deteriorate
Solution Approach 1:
The system applies local quality by tailoring pack configurations to specific store clusters rather than using a single universal configuration for all stores. Each cluster receives customized pack configurations based on its demand characteristics, ensuring that local product availability needs are met. This approach maintains relatively low complexity by limiting customization to the cluster level rather than the individual store level.
4Measurement precision
If comprehensive optimization is performed for all stores simultaneously, then allocation accuracy improves, but scalability is compromised
Solution Approach 1:
The patent divides the comprehensive optimization problem into smaller, manageable segments by clustering stores into groups with similar demand patterns. The optimization is then performed separately for each cluster rather than for all stores simultaneously. This segmentation enables the system to scale to large numbers of stores while maintaining allocation accuracy within each cluster through the use of cluster-specific optimization.
Solution Approach 2:
The system creates universal pack configurations for each cluster that can be applied to all stores within that cluster. These cluster-level universal configurations serve multiple stores with similar characteristics, enabling the system to scale efficiently. The multi-functionality of cluster-based configurations allows the same optimization logic to serve numerous stores without proportionally increasing computational burden.
Data Source
AI summary
Methods and systems for optimization and deployment of assortment pack configuration and allocation are disclosed. Different items, e.g., variants on a single core item type, may be packed together to reduce shipping costs and limit the impact of stochastic elements in the supply chain. In accordance with example aspects of the disclosure, a scalable ensemble optimization approach is used which initializes, iterates, and finalizes a packing solution in an accurate, computationally-efficient manner. In some examples, the approach described herein utilizes an iteration phase which explores a bilinear structure of the underlying problem and exploits randomness to explore the possible solution space.