Diverse Counterfactual Explanations (DiCE) Role in Improving Sales and e-Commerce Strategies

• Published in: Journal of theoretical and applied electronic commerce research Vol. 20; no. 2; p. 96
• Main Authors: Oprea, Simona-Vasilica, Bâra, Adela
• Format: Journal Article
• Language: English
• Published: Curicó MDPI AG 01.06.2025
• Subjects: Automation; Bridge maintenance; Business to business commerce; Cameras; Consumers; Consumption; Corporate finance; counterfactual analysis; Credit scoring; Decision making; Decisions; Design optimization; Discounts; Elasticity of demand; Electronic commerce; Explainable artificial intelligence; Machine learning; machine learning post-action; Market positioning; Market segmentation; Market segments; Market strategy; Marketing; Price elasticity; Price increases; price strategy; Pricing; Product design; Product differentiation; Product specifications; Profitability; Regression models; Statistical analysis; Statistical methods; Trends; what-if explanations
• ISSN: 0718-1876; 0718-1876
• DOI: 10.3390/jtaer20020096

Pricing strategy is a critical challenge in e-commerce, where businesses must balance competitive pricing with profitability. Traditional pricing models rely on historical data and statistical methods but often lack interpretability and adaptability. In this study, we propose a novel approach that leverages Diverse Counterfactual Explanations (DiCE) to enhance pricing strategies for mobile phones. Unlike previous research that applied counterfactual analysis in customer segmentation, energy forecasting, and retail pricing, our method directly integrates explainability into product-level pricing decisions. Our approach identifies actionable product features, such as improved hardware specifications, that can be modified to increase the predicted price. By generating counterfactual explanations, we provide insights into how businesses can optimize product attributes to maximize revenue while maintaining transparency in pricing decisions. This framework bridges explainable AI with pricing strategies, allowing companies to justify price points and improve market positioning dynamically. Furthermore, we identify other features that could lead to the same price goal. The linear regression model achieved an R2 score of 96.15% on the test set, along with a mean absolute error (MAE) of 108.31 and mean absolute percentage error (MAPE) of 5.43%, indicating strong predictive performance. Through DiCE, the model identified actionable modifications (e.g., increasing front camera resolution and battery capacity) that effectively raise predicted prices by 15–20%. This insight is particularly valuable for product design and pricing optimization. The model provided a ranking of features based on their impact on price increases, revealing that front camera and battery capacity are more influential than RAM in driving pricing changes.