Analysis of Consumer Sentiment and Problems in Online Shopping
Authors
Keywords
consumer sentiment, consumer review, online shopping, Amazon, Trustpilot, VADER, LDA
Summary
This study analyses consumer sentiment and the main sources of dissatisfaction in online shopping by examining 1,254 independent customer reviews of Amazon published on Trustpilot between October 2025 and January 2026. The paper applies lexicon-based sentiment analysis (VADER), correlation and regression analysis, and topic modelling (LDA).
Opinions are polarised: 54.63% of reviews carry negative sentiment, 41.95% positive and 3.43% neutral. About half express a categorical attitude, while the rest contain mixed or restrained assessments. The relationship between the composite sentiment score and the rating is moderate and positive (r = 0.413), so the text and the stars convey similar but not identical information. One third of one-star reviews receive a positive composite score, which marks the limits of the lexicon-based approach. Controlling for the reviewer's country leaves this relationship almost unchanged, while the ratings themselves differ systematically across countries.
The topic analysis identifies five areas. Reviews about delivery to the address and about contacting customer service carry the most negative sentiment, while those about returns and refunds have the lowest average rating. Together these three areas cover 57.18% of the reviews and concern stages that follow the order. The findings direct attention to the last mile of delivery and to problem resolution after purchase.
Pages: 17
DOI: https://doi.org/10.58861/tae.di.2026.3.03


