When Are Customers Willing To Pay For Better Reliability And How Much: Empirical Study On E-Commerce Observational Data Through Causal Inference And Natural Language Processing
Résumé
Product reliability is widely acknowledged as a crucial aspect for both enterprises and customers. However, solely investing in improving product reliability does not guarantee enhanced sales performance. Various factors, including reliability, price, brand, and customer sentiment, contribute to a product's success, with intricate interactions among them. This paper introduces a statistical model leveraging observational data from e-commerce websites to explore these interactions and their impact on sales. Our study employs explanatory variables such as reliability, price, brand, and customer sentiment, with product sales as the dependent variable. Popularity and sales figures are approximated using the number of reviews per unit of time in the product life cycle. Reliability and customer sentiment are extracted from customer reviews through two natural language processing (NLP) models. Recognizing the limitations of observational data, where statistical association may differ from causality, we apply the propensity score matching approach to estimate the average treatment effect of each explanatory variable. Causality is crucial for decision-makers seeking to understand why a product succeeds or fails. Our approach addresses this by offering insights into the factors influencing product popularity. The analysis reveals that 'customer sentiment' is the most significant factor, followed by 'reliability ' and 'price.' In contrast, a comparable study, neglecting causal effects, prioritizes 'reliability,' followed by 'customer sentiment' and 'price.' The results underscore the importance of considering causal inference, as our model corrects biases introduced by confounders in observational data, aiding decisionmakers in prioritizing features to enhance a product's popularity.
Domaines
| Origine | Accord explicite pour ce dépôt |
|---|---|
| Licence |