<![CDATA[Online fresh-produce retailing presents distinctive quality and delivery uncertainties that influence consumer evaluations and repurchase intentions. This study contributes a leakage-controlled, polarity-aware text-mining framework for examining consumer sentiment and repurchase signals in unsolicited online reviews. Using 113 public Tokopedia reviews covering 31 fresh-produce products, this pilot study employed lexicon-based sentiment classification and logistic regression with product-clustered standard errors. The findings reveal predominantly positive consumer sentiment, although explicit repurchase intentions appear less frequently. Consumer sentiment demonstrates a significant positive association with repurchase signals, while product-quality mentions show no significant relationship. Unexpectedly, positive delivery mentions exhibit a negative association with repurchase signals, potentially reflecting review-length constraints rather than unfavorable delivery experiences. These findings extend the methodological application of consumer-generated content analysis by demonstrating the feasibility of extracting distinct behavioral signals while addressing classification bias. Online retailers can leverage this framework to monitor consumer experiences and identify opportunities for service improvement.]]>
Copyrights © 2026
