Optimizing return management analysis in e-commerce: topic modeling of customer reviews using BERTopic and ChatGPT

Abstract

The online product sales business, known as e-commerce, has evolved a lot over the last few years, focusing more and more on improving the customer experience. One of the big issues in this paradigm is the lack of perception of the product when it is bought online, leading to a higher ratio of product returns than in traditional stores. This product return involves a great cost, both in logistics and product management. That is why an effective return management system is needed. In this thesis, attention is drawn to the importance of analyzing the reviews of returned products and the fact that they contain very valuable information. The study is based on a new topic modeling technique known as BERTopic and the application of NLP in order to extract the main causes of product returns from a review database of one of the biggest retail companies around the world, Amazon. In addition, different data treatment techniques are implemented to optimize and automate this process by using AI, more specifcally ChatGPT. Our findings revealed significant insights for e-commerce platforms and retailers, high- lighting the importance of modeling the reviews divided by product category due to the distinct behaviors and return reasons associated with different product types. The study's results can be extended to other product categories or sectors, offering a versatile solution for minimizing product returns and improving customer satisfaction. In conclusion, this paper introduces an innovative tool to extract and interpret e-commerce reviews to provide actionable insights for stakeholders, with the aim of reducing the product return ratio.

Outgoing