Dual-Layer Demand Transference Model for Assortment Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demand transference prediction models, such as the Scan*Pro model, often produce unrealistic and unintuitive results, leading to reduced retailer confidence and increased implementation complexity due to the need for ad hoc adjustments to avoid these issues.
Innovation Solution
A dual-layer machine learning model is implemented, combining a modified Scan*Pro model with a Multinomial Logit (MNL) model to predict demand transference, ensuring that total sales do not increase when items are removed from an assortment and providing a mathematical guarantee against unrealistic outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a standard demand transference prediction model (e.g., Scan*Pro) is used, then the model can be implemented with relatively simple structure, but it produces unrealistic and unintuitive results leading to reduced retailer confidence
Solution Approach 1:
The model is segmented into two distinct layers: a top layer MNL model that ensures realistic demand transference predictions by preventing total sales increase, and a bottom layer log-linear retail sales model that captures historical sales patterns. This segmentation allows each layer to address specific requirements while combining their strengths.
Solution Approach 2:
The patent creates a composite modeling approach by integrating two different modeling paradigms (MNL and log-linear) into a unified dual-layer structure. The top layer uses MNL for theoretical correctness and the bottom layer uses log-linear for empirical accuracy, creating a hybrid model that leverages the advantages of both approaches.
2Reliability
If ad hoc adjustments are made to the model to avoid unrealistic outcomes, then the reliability of predictions improves, but the implementation complexity increases
Solution Approach 1:
The MNL top layer is designed to preemptively prevent unrealistic outcomes (such as total sales increase when items are removed) before they can occur. By embedding the theoretical constraints directly into the model structure, the patent eliminates the need for post-hoc adjustments and ad hoc fixes.
Solution Approach 2:
The dual-layer model is self-regulating through its mathematical structure. The MNL top layer automatically ensures that demand transference predictions are theoretically sound without requiring external intervention or manual adjustments, making the model self-correcting and easier to implement.
3Measurement precision
If a more complex dual-layer model is implemented to ensure realistic predictions, then prediction accuracy improves, but computational efficiency may decrease
Solution Approach 1:
By segmenting the model into two layers with distinct responsibilities, the patent enables efficient computation where the top layer handles theoretical constraints and the bottom layer handles empirical pattern recognition. This division allows for optimized computational approaches in each layer.
Solution Approach 2:
The model transforms the complex problem of ensuring realistic predictions into parameter estimation problems that can be solved efficiently using standard statistical techniques. By changing the approach from constraint satisfaction to parameter optimization, computational efficiency is improved while maintaining accuracy.
Data Source
AI summary
Embodiments determine demand transference for an item assortment of a retailer. Embodiments receive historical sales data for a category of items corresponding to the retailer and receive hierarchy data for the category of items corresponding to the retailer. Based on the historical sales data and the hierarchy data, embodiments estimate first variables of a multinomial logit (“MNL”) model. Based on the historical sales data and the hierarchy data, embodiments estimate second variables of a log linear retail sales model.