Improving e-Retail Rating Approach for Product Market Research Open Access
Downloadable ContentDownload PDF
In e-retail industry, online Word-of-Mouth (e-WOM) includes vast amounts of text-reviews and star-ratings of products. Such valuable information does not only help consumers, by enabling informed purchasing decisions, but also benefits businesses and engineering professionals, who are involved in product design and development. Current star-rating approaches allows individuals to rate a product. On the other hand, the reasoning behind such star-ratings continues to be hidden in large volumes of text-reviews. Mitigating this gap empowers all stakeholders to understand product market better. Consequently, shortcomings of products can be addressed, new products can be innovated based on market demands, and favored product features can be recognized and maintained. In this praxis, (1) Natural Language Processing (NLP) applied to hundreds of thousands of Amazon reviews to identify relevant product attributes, and (2) Ordinal Logistic Regression (OLR) is used to assign star-ratings to those attributes. As a result, a more granular and predictive star-rating approach emerged that can derive key information from online reviews and deliver such insight quickly via assigning numerical scores, star-ratings, to product features. Improvement to the star-rating process provides critical information to key stakeholders, such as manufacturers, data scientists, product owners, and engineering managers, to understand how well a given product meets the associated market demand. In addition, mining online text-reviews to derive product qualities can determine superior product qualities to maintain, as well as shortcomings of the products to enhance. Such insight can also boost innovation by providing a snapshot of product market and relevant demand. Engineering managers can initiate projects to address deficiencies of existing products, as well as innovate new ones. Furthermore, they can apply such NLP techniques to any given large dataset to derive valuable information automatically.