Essays in international market segmentation

The primary objective of this thesis is to develop and validate new methodologies to improve the effectiveness of international segmentation strategies. The current status of international market segmentation research is reviewed in an introductory chapter, which provided a number of methodological and substantive issues that need further attention. These issues are critically assessed and methodologies are developed as potential solutions.

In chapter 1, previous research in international segmentation is classified according to three dimensions depicted in Figure 1. In the figure, the first dimension relates to the segmentation basis, the second to segmentation objects, and the third to segmentation methodology. All three dimensions affect the effectiveness of international segmentation strategies. Two key research directions for improving the effectiveness of international segmentation were formulated along these dimensions.

The first direction concerns the integration of targeted product and communication strategies by linking product-specific bases with general consumer-level bases. A new methodology is developed to identify cross-national market segments using means-end chain theory. Based on theory founded in consumer behavior, the means-end chain links values (a general consumer-level basis) with benefits and attributes (product-specific bases).

Figure 1

Three dimensions of international segmentation

Such an approach has the potential to combine product development and communication strategies at the international segment level and may serve as a guiding principle for international marketers to tailor products and advertising messages to the desires of global consumer segments. Chapter 4 provides a model-based methodology for identifying such segments. An international segmentation model was developed that estimates relations between product attributes, benefits of product use, and consumer values at the international segment level, and at the same time identifies those segments. The model builds upon methodological issues that were addressed in chapters 2 and 3 and rests on mixture methodology that, due to its capability of deriving segments based on models of consumer behavior, is particularly effective. In particular, it accounts for the international sampling design and the heterogeneity of response tendencies across countries and consumers.

The segmentation model was applied to identify segments in the European yogurt market, using a large sample of European consumers. Four segments were identified, of which one was truly pan-European and the other segments were cross-national. The segments were found to represent distinctive means-end structures and the pattern of links between attributes, benefits, and values gave rise to strategic implications with respect to product development and communication. The segments were found to be related to socio-demographics, consumption patterns, media consumption, and personality data, which contributes to the identifiability and accessibility of the segments. The results suggest that the proposed model-based international segmentation methodology, combining product- and consumer-level bases, has the potential to identify segments of consumers in different countries that are actionable towards product development and advertising strategy.

In chapter 5, a different direction is proposed that seeks to improve the effectiveness of international target market selection of expanding companies, by improving the geographic configuration of segments. Whereas consumer segments are more responsive, their typical geographic configuration does not make them accessible with cost efficient logistic operations. Especially if physical distribution represents a major component of total production and marketing costs, it is important that a geographic segment defines one particular area as opposed to dispersed segments that may arise in previous segmentation approaches. A flexible model-based segmentation approach is developed that identifies contiguous geographic segments based on consumer-level data. The model is based on multi-attribute theory of preference formation and accommodates a broad set of strategic restrictions on the segments. Moreover, the model accounts for heterogeneity that is likely to exist within geographic segments.

The methodology is illustrated in the international retailing domain, where geographic expansion is an important strategy to attain growth. Based on the importance that consumers attach to different attributes of store image, five geographic segments were identified across regions in seven countries of the European Union. The segments were distinctive in terms of their patterns of image attribute importances, which provides opportunities for expanding retailers to delineate geographic areas to enter and to develop an appropriate image in such areas. The results also demonstrated the accessibility of the segments through advertising media and logistics. In addition, no significant differences were found between the original model and a nested model that does not take the contiguity into account. This means that the actionability of restricting segments to be contiguous does not substantially harm the responsiveness of these segments.

Given the often limited rigor of statistical and measurement techniques applied in the area of international segmentation, special attention has been given to methodological issues. Several issues were addressed that may negatively affect international segmentation research findings and methods were developed to deal with these issues.

The first issue concerns the segmentation method . International segmentation research demonstrates an excessive reliance on heuristic segmentation techniques, such as cluster analysis. These techniques provide limited flexibility for international segmentation and may not be very effective in recovering response-based segments. The international segmentation methodologies developed in this thesis are model based and rely on insights from state of the art statistical techniques such as mixture and hierarchical Bayes models. Three international segmentation models are described in chapters 2, 4, and 5, and are successfully applied to empirical data. Chapter 5 provided a Bayesian formulation of a new international segmentation model that accommodates within-segment heterogeneity and complex restrictions on the configuration of segments. In chapter 4 it is empirically shown that a new mixture model approach outperforms standard clustering approaches that are traditionally employed in international segmentation.

A second methodological issue is related to the estimation of international segmentation models. The importance of international sampling designs had not been acknowledged in the literature on international segmentation and mixture modeling. Previous international segmentation studies did not account for the implicit stratified sampling designs encountered in cross-national data collection. In this thesis the effects of international sampling designs on maximum likelihood estimation of segmentation models are investigated and a framework for accommodating those effects is proposed. A pseudo maximum likelihood procedure is introduced that accommodates complex sample designs for maximum likelihood estimation of finite mixture models. In addition, modified or pseudo-information criteria are suggested for correct estimation of the number of international segments.

The effects of not accounting for the sampling design were empirically assessed in an international value segmentation study. The pseudo-maximum likelihood approach was compared to standard maximum likelihood estimation that does not account for the sampling design. The results show that the estimates of segment sizes and segment-level parameters may be severely biased when not accounting for the design in standard maximum likelihood estimation. In addition, the empirical application demonstrated that the use of standard information criteria leads to incorrect inferences about the number of segments. This means that standard estimation methods in international segmentation research may lead to incorrect conclusions and erroneous managerial action.

The international segmentation methodology in chapter 4 was based on MEC theory. The traditional measurement technique for means-end chains (laddering) is not suitable for international segmentation. A necessary condition for the validity of international segments is that the basis for segmentation is measured in a valid and reliable way. Measurement instruments should allow collecting large and representative samples and standardization across countries. In this thesis a MEC measurement technique is developed that meets those criteria. The technique is denoted as the association pattern technique (APT), and its validity is further assessed. Two key issues were investigated that may hamper the validity of APT. First, APT implicitly assumes that attribute-benefit and benefit-value links are independent because it measures these links in two separate tasks. The second issue is the convergent validity of APT as compared to the more traditional laddering interview. Consistent support for independence of attribute-benefit and benefit-value links was found across four product categories. Statistical tests of convergent validity of APT and laddering demonstrated that the basic structure revealed by both methods is similar. This suggests that APT is valid for measuring means-end chains and can be used for identifying international consumer segments. The APT method is successfully applied in an international segmentation study in 11 countries.

The final methodological issue addressed in this thesis is related to response tendencies , which may hamper the identification of cross-national segments. The APT method may be prone to a respondent's propensity to choose any link. Therefore, the international segmentation model in chapter 4 accounted for differences in those tendencies that may exist between respondents. Based on item response theory, a response threshold approach was developed that allows testing those differences between countries, but also within countries. The results demonstrated that the differences in response tendencies were significant between countries, but also within countries. This means that it is important to account for response tendencies in international segmentation but in domestic segmentation as well.