Investigating the Heterogeneity of Product Feature Preferences Mined Using Online Product Data Streams

Abstract

This work investigates the “must have” and “deal breaker” product feature preferences expressed by users of online platforms (e.g., customer review websites or social media networks) in order to inform designers of product features that should be investigated during the next iteration of a product’s launch. Existing design literature highlights the risks of aggregating group preferences, and suggest that design teams should instead, focus on maximizing enterprise value by optimizing the attributes of a product. However, design knowledge about products and product attributes are influenced by market information, which is dynamic and difficult to acquire. The use of online product review platforms has emerged in the design community as a viable source of product data acquisition and demand model prediction. However, as the heterogeneity of product preferences increases, so does the complexity of understanding which product attributes should be optimized by the design team to maximize enterprise value. These challenges are exacerbated in product preference acquisition techniques that rely on mining online data, as the customer is typically unknown to the designer, which limits the amount of follow up data available to be mined. By quantifying the degree of “must have” and “deal breaker” product preferences expressed online, designers will be able to understand what product-features should be omitted from next generation product design optimization models (i.e., “deal breaker” features) and what product features should be considered (i.e., “must have” features). A case study involving customer electronics mined from online customer review websites is used to demonstrate the validity of the proposed methodology.

Cyber-Empathic Design: A Data-Driven Framework for Product Design

“…Recently, attempts have been made to extract user experience information from nontraditional sources of information including online product reviews. This information is then translated using data analytics in order to extract product features and quantitatively investigate product feature preferences [39][40][41][42][43]. In preference modeling, the mathematical techniques traditionally rely on linear mapping.…”

Using Machine Learning to Improve Lead Times in the Identification of Emerging Customer Needs

“…Stop words, URLs and curse words (as in [89]) are removed due to having a low likelihood of being related to a customer need.…”

Extracting Customer Perceptions of Product Sustainability From Online Reviews

“…Singh and Tucker used sentiment analysis to determine "must have" and "deal breaker" features for products [15]. "Must have" features are those that are popular while "deal breaker" features are those that are unpopular.…”