Demand Planning for Substitutable Items via Category Elasticity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demand planning models are inefficient and impractical for retailers due to the need to calculate cross elasticities for every pair of similar products, leading to scalability issues and unreliable data, which results in incomplete item interactions and inaccurate category constraints.
Innovation Solution
A demand planning method that eliminates the need to estimate cross elasticities for each pair of items by using a model that calculates item and category elasticities, allowing for linear scalability with the number of items and maintaining asymmetry, enabling faster processing and more accurate price elasticity calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cross elasticity is calculated for every pair of similar products, then demand forecasting accuracy is improved, but computational complexity and processing time increase exponentially
Solution Approach 1:
The patent segments the monolithic cross-elasticity calculation into two distinct components: category-level elasticity estimation and item-level elasticity derivation. By first estimating elasticity at the category level using aggregate sales data, then deriving individual item elasticities from the category elasticity and item-specific price effects, the system divides a computationally intractable problem into manageable segments that can be solved efficiently
Solution Approach 2:
The patent merges the estimation of multiple item-specific elasticities into a single category-level elasticity estimation. Instead of calculating n*(n-1) individual cross-elasticity values for n items, the system estimates one category elasticity parameter that captures the aggregate substitution effects, then derives individual item elasticities algebraically from this single estimate combined with item price data
2Loss of information
If cross elasticity is calculated for every pair of similar products, then complete item interactions are captured, but data reliability decreases due to insufficient observations
Solution Approach 1:
The patent combines sparse item-level observation data with aggregate category-level data to produce reliable elasticity estimates. By pooling sales and price data at the category level, the system accumulates sufficient observations to estimate category elasticity reliably, then uses this robust category-level estimate as the foundation for deriving individual item elasticities, ensuring that item interaction effects are captured without relying on insufficient item-pair observations
3Adaptability or versatility
If traditional cross elasticity methods are used, then detailed product relationships are modeled, but scalability to large retailers with hundreds of items is lost
Solution Approach 1:
The patent segments the elasticity estimation problem into hierarchical levels (category and item) that scale differently with product assortment size. The category-level estimation step scales with the number of categories rather than the square of the number of items, while the item-level derivation step scales linearly with the number of items. This segmentation enables the system to handle large retail assortments with hundreds or thousands of items across multiple categories without exponential computational growth
Solution Approach 2:
The patent creates a universal category elasticity estimation framework that can be applied across different categories and retail contexts. The category-level elasticity model serves as a universal foundation that captures general substitution patterns within categories, which can then be adapted to specific items through the derivation step. This universal approach allows the system to scale to retailers with diverse and expanding product assortments without requiring reconfiguration of the core modeling framework
Data Source
AI summary
A system and method are disclosed for planning a product assortment based on a sales forecast without using a cross elasticity by receiving a percentage pricing change for at least two substitutable products of an inventory in a supply chain network having one or more supply chain entities, and at least two substitutable products are grouped in the same product category and at least one of at least two substitutable products is grouped in a product assortment, calculating an average percent pricing change for the product category including at least two substitutable products and a direct effect factor and cross-effect factor for each of at least two substitutable products, and identifying an item of at least two substitutable items to be removed from the product assortment based, at least in part, on a substitutable demand calculated by modeling a price increase of a substitutable item to infinity.