Price Elasticity Modeling for Sparse Retail Demand Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for computing price elasticity (PE) values are inadequate for products with limited historical data, seasonal variations, and volatile sales, leading to suboptimal pricing recommendations.
Innovation Solution
A method and system using unsupervised reinforcement learning models, specifically Thompson Sampling (SPECTS), to compute PE values by preprocessing data, selecting models, and updating price elasticity distributions based on prior and likelihood distributions, ensuring alignment with market dynamics and user objectives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If regression-based models using historical data are used to compute price elasticity values, then accurate PE values can be obtained for products with sufficient price changes in history, but the method fails for products with very few price changes, items with very few sales points, new items and items with no history
Solution Approach 1:
The system performs preliminary actions by pre-computing price elasticity values for products during periods when sufficient data is available, and stores these values for later use during periods when data is scarce. This allows the system to provide accurate PE values even when current historical data is insufficient.
Solution Approach 2:
The system introduces an intermediary component that bridges the gap between regression-based models and products with limited data by using alternative data sources, proxy products, or synthesized information to compute PE values when traditional methods fail.
2Device complexity
If traditional regression-based methods are used to compute price elasticity values, then the computation process is simple and straightforward, but the method cannot capture variations in PE values across different phases of sales and seasonal variations
Solution Approach 1:
The system transitions from static regression-based PE computation to dynamic PE computation that adapts to different phases of sales and seasonal variations. The system continuously updates PE values based on current market conditions, product lifecycle stage, and temporal patterns, allowing it to capture variations across time while maintaining computational feasibility.
Solution Approach 2:
The system performs preliminary analysis to identify seasonal patterns, product lifecycle phases, and temporal variations in PE values. These pre-computed temporal characteristics are then used to adjust PE computations in real-time, capturing seasonal and lifecycle variations without requiring complex real-time computations.
3Productivity
If approximate PE values are used in an ad-hoc manner, then the computation process is quick and easy, but suboptimal price recommendations are generated
Solution Approach 1:
The system performs preliminary computations to establish baseline PE values and identifies patterns in the data that can be reused. By pre-processing data and computing PE values in advance when conditions are favorable, the system maintains high accuracy while enabling fast real-time pricing recommendations.
Solution Approach 2:
The system maintains continuous computation and updating of PE values rather than relying on static ad-hoc approximations. By continuously refining PE estimates as new data becomes available and maintaining ongoing computations, the system ensures both speed and accuracy in price recommendations.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
In retail industry, retailers need accurate price elasticity (PE) values to offer optimal prices for products to business constraints. Existing approaches provide approximate PE values leading to suboptimal price recommendations. This disclosure relates to method of computing PE values for optimal pricing based on sequential price elasticity computation is provided. Transaction data and attribute data of products are processed to obtain preprocessed data. One or more selected models are determined based on the preprocessed data. Priors at one or more levels are computed and likelihoods are computed based on one or more monetary objectives from historical transaction data of the products. One or more parameters of one or more price elasticity distributions are iteratively determined through one or more component approaches based on the priors and the likelihoods. The PE values are derived based on ensemble techniques applied to one or more parameters of the one or more PE distributions.