SENTIMENT ANALYSIS OF CONSUMER REVIEWS ON DARAZ: PREDICTING PRODUCT RETURN RATES

Ahmad Sohail; Dawood UL Hassan

Authors

Ahmad Sohail Author
Dawood UL Hassan Author

Keywords:

Sentiment analysis, consumer reviews, Daraz, product returns, e-commerce, Pakistan

Abstract

Purpose

This research investigates the application of sentiment analysis on consumer reviews from Daraz to forecast product return rates. The rapid expansion of Pakistan's e-commerce sector has led to product returns becoming a significant challenge that impacts profitability, logistics, and consumer trust. This research seeks to connect consumer sentiments found in reviews with return behavior, aiming to offer practical insights for minimizing reverse logistics costs and improving customer satisfaction.

Design/Methodology

The research employs a secondary quantitative approach, utilizing consumer reviews that are publicly accessible from Daraz. We utilized natural language processing (NLP) techniques to extract sentiment polarity, intensity, and review length, and performed statistical analysis with JASP software. We utilized descriptive statistics, correlation analysis, paired t-tests, and contingency tables to evaluate the connections among review sentiments, ratings, and return frequency.

Findings

The findings showed a significant link between negative sentiment and increased product return rates, especially in the fashion sector, where returns surpassed 28%. It was frequent to observe discrepancies between star ratings and written reviews, suggesting that ratings by themselves do not adequately predict consumer dissatisfaction. Reasons for returns primarily revolved around size discrepancies (35%), subpar quality (29%), and inconsistencies in descriptions (26%). The analysis revealed a significant positive connection (r = 0.68) between negative sentiment and how often returns occur.

Practical Implications

The results indicate that Daraz and similar e-commerce platforms have the opportunity to utilize automated sentiment tracking to detect unhappy customers promptly, enhance logistics, bolster after-sales support, and minimize expensive returns.

Originality

This research distinctly combines sentiment analysis with predictive return modeling within the realm of e-commerce in Pakistan, providing valuable methodological and practical insights for studies in emerging markets.