Consumers’ Perceived Benefits and Costs for Amazon Go Based on Social Media Data Using Text Mining

This study aims to identify the perceived benefits and costs of Amazon Go, the first unmanned store with advanced technology, based on the value-based adoption model using a text-mining technique. We collected 15,435 documents posted on Twitter from January 21, 2018, to September 1, 2021, employing “Amazon Go” and “#Amazon Go” as keywords. Frequency analysis, clustering analysis, CONCOR, and semantic network analysis were conducted using Python and R programming. The major results are as follows. First, we extracted 20 attributes of Amazon Go, which are classified into four dimensions (functionality, no humanity, privacy risk, and self-service). Second, overall, consumers perceived greater benefits than costs of Amazon Go. Consumers perceive “automation” as the most beneficial attribute of Amazon Go, and attributes referring to “frictionless payment,” “tracking of past purchase history,” “no humanity,” and “no cash acceptance” were also observed positively. However, negative perceptions regarding each of these attributes, such as “hard,” “bad,” and “worried,” were also detected. Third, we found concerns about privacy infringement, indicating that consumers may resist new technologies due to privacy concerns. In addition, consumers’ worries about “shoplifting” were extracted as costs. Finally, terms related to price perceptions were not extracted, reflecting that the concept of Amazon Go is still new to consumers and its users are still at the trial stage. This study offers implications that may help unmanned retailers with the advanced technology to deliver services that correspond to consumer values and expectations, thus increasing consumer utility and satisfaction.