Identifying Constellations of Website Features: Documentation of a Proposed Methodology

A “feature preference constellation” is a set of commercial website features similar in appeal to shoppers. That is, the features in a constellation: covary in attractiveness (i.e., correlate), are similar in elevation (mean attractiveness), are treated as intact entities rather than being redefined as the partial reflections of underlying factors or dimensions, and can represent more than a single function or meaning. Identification of constellations is useful to practitioners as a guide for developing new websites and to scientists as a predictor/explanation for adoption and diffusion of sites throughout a market. A constellation is a different construct from a “dimension” and cannot be gauged via the factor analytic procedures popularly used to assess dimensions. A 10-step procedure was developed to identify constellations. Dependent upon sample representativeness and research objectives, feature preferences may first be adjusted for sample (here, demographic) characteristics by the use of “rescaled residual scores.” Feature similarity indices, Lin coefficients sensitive to both elevation and covariance, are subjected to multidimensional scaling. Features, as points in multidimensional space, are grouped by a two phase hierarchical/partition clustering procedure. Comparison with the most popular factor analytic approach showed that the proposed procedure yielded meaningful feature groupings more homogeneous in elevation while comparable in within group covariation and in separation among groups. It was concluded that the procedure can operationalize the constellation construct, and do so more effectively than can the popular factor analytic approach. It was suggested that the constellation concept, because of its unique suitability for particular applications, be considered by both scientists and practitioners. A condensed version of this study appears in the RRCB paper, “Website feature preference constellations: Conceptualization and measurement” by the present authors. Identifying Constellations of Website Features: Documentation of a Proposed