Developing and validating e-retailing satisfaction scales with text-mining

Purpose The purpose of this study includes two parts. First, it introduces a machine-based method for model and instrument development and updating that integrates large sample qualitative data. Second, a new model and instrument for e-commerce customer satisfaction are developed. Design/methodology/approach The research occurs in two phases. In Phase 1, data collection occurs with a literature-based quantitative model and instrument that includes at least one qualitative scale item per construct. Data analysis of the resulting data includes factor analysis (FA) and latent semantic analysis text mining to generate an updated model and instrument. In Phase 2, data collection uses the new model and instrument. Data analysis in Phase 2 includes exploratory data analysis with FA, exploratory structural equation modeling and partial least square modeling. Findings As a result of the information gained by the integration of qualitative scales in the literature-based survey, the final model departs substantially from the initial research-based research model. It integrates many of the constructs known to impact a website and software usability from information systems research into a new e-retail satisfaction model. Originality/value The research method, as presented here, offers a strategy for integrating large scale qualitative data for refinement of models and the development of instruments. It is essentially a method of gaining the wisdom of crowds economically while simultaneously reducing the biases and laborious effort commonly associated with qualitative research.