Bayesian Network Evidence on Price, Electronic Word of Mouth, and Seasonality in Amazon Product Sales Performance

This study examines the associations of price, eWOM, and seasonal review activity with product sales performance in cross-border e-commerce. The analysis uses 3,886 products from Amazon's All Beauty category and treats sales rank as a proxy for relative sales performance. A compact Bayesian network links price level, eWOM strength, peak review season, and sales performance. Conditional probabilities are estimated with Laplace smoothing, while chi-square tests and Cramer's V assess the stability and strength of the observed associations. Products with high eWOM strength have a 48.9% posterior probability of high sales performance, compared with 19.4% for the medium group and 14.3% for the low group. Price differences are small, and seasonal differences are also limited. The strongest multi-factor combination consists of medium price, high eWOM strength, and a spring review peak, with a posterior probability of 58.7%. The results indicate that credible review activity is more closely associated with strong sales performance than price level or seasonality. The findings support a management approach that gives priority to genuine review quality, uses controlled price tests, and treats seasonal timing as supporting evidence rather than a primary decision rule.