Capturing Cannibalization and Complementarity Effects in Retail. A Multimodal Deep Learning Approach
Posted: 6 Feb 2024
Date Written: February 2, 2024
Abstract
This paper presents a novel two-stage multimodal approach for capturing cannibalization and complementarity effects within the retail sector. Traditional demand methods often assume independence between items and hence struggle to capture the intricate interplay between products. In response, we propose a framework that combines product information (textual, visual, and numerical features), and ticket data to provide a holistic understanding of product relationships. The model is validated using real-world retail data, demonstrating its ability to predict and quantify the impact of cross-item effects accurately. The findings derived from this approach enhance our understanding of cannibalization and complementarity dynamics and offer valuable insights into retail operations, such as assortment optimization, pricing, and inventory management.
Keywords: cross-item effects, multimodal deep learning
Suggested Citation: Suggested Citation