Tonks, D G (2004) Identifying market segments in consumer markets: variable selection and data interpretation. Working Paper. The Department of Marketing, Lancaster University.
Abstract
Market segmentation is often articulated as being a process which displays the recognised features of classical rationalism but in part; convention, convenience, prior experience and the overarching impact of rhetoric will influence if not determine the outcomes of a segmentation exercise. Particular examples of this process are addressed critically in this paper which concentrates on the issues of variable choice for multivariate approaches to market segmentation and also the methods used for identification of the resulting market segments. 'Market segmentation' is a key phrase if not a central notion for marketing academics and marketing practitioners alike. For example, Hooley and Saunders (1998) propose that '… segmentation is a logical extension of the marketing concept itself.' In use however, the notion of market segmentation can be problematic. Although science-like in its procedures and applications, market segmentation presents significant operational problems and these include the selection of segmentation variables and the interpretations of resulting data - such that the managerial and also the broader cultural associations of market segmentation might both be likened to a latter day 'bureaucratic form of Sophism' (Laufer and Paradeise 1990) where the impact of rhetoric in its classical sense can be very powerful. The desires for order, for control and also for belonging can all be connected with the need to make sense of diversity and with the resulting quest to classify and seek associations. These are all underlying features of market segmentation as either academic reflection or as managerial deliberation and subsequent action. (Similar comments also apply to the agency of rhetoric). With the practice of market segmentation, a first step is the choice of a variable or variables for identification, description and association. Multivariate approaches may take the form of numerous variables each taken individually but there are many clustering type methodologies which seek to look for comm