Data for a study of U.S., Japanese, and British competitors were obtained from detailed personal interviews with chief executives and top marketing decision makers for defined product groups in90 companies. Tocontrol for market differences, the methodology was based upon matching 30 British companies with their major American and Japanese competitors in the U.K. market. The study involved 30 triads of companies, each composed of a British, American, and Japanese business that competed directly with one another.
Most of the data on the characteristics of the companies’ performance, strategy, and organization were collected 0.1 5-point semantic differential scales. The first stage of the analysis involved factor analysis of variables de .bing the firms’ strategies and marketing activities. The factor scores were used to identify groups u similar companies using Ward’s hierarchical clustering routine. A six-cluster solution was developed.
Membership in the ixclusters wasthen interpretedagainsttheoriginalperformance, strategy, and organizational variables.AlI the clusters contained some successful companies, although some contained significantly morethanothers. The clusters lent support to the hypothesis that successful companies were similar irrespective of nationality, because American, British, and Japanese companies were found in all the clusters.There was, however, a preponderance of Japanese companies in the more successful clusters and a predominance of British companies in the two least successful clusters. Apparently, Japanese companies do not deploy strategies that are unique to them; rather, more of them pursue strategies that work effectively in the British market The findings indicate that there are generic strategies that describe successful companies irrespective of their industry, Three successful strategies can be identiflCd. The first is the Quality Marketing strategy. These companies have strengths in marketing and research and development. They concentrate their technical developments on achieving high quality rather than pure innovation. These companies are characterized by entrepreneurial organizations, long-range planning. and a well-communicated sense of mission. The second generic strategy is that of the Innovators. who are weaker on advanced R&D but are entrepreneurial and driven by a quest for innovation. The last successful group is the Mature Marketeers, who are highly profit oriented and have in-depth marketing skills. All three appear to consist of highly marketing-oriented businesses. Foreign investment in the United Kingdom continued to be robust. In 2009. the United Kingdom was leading other European countries in terms of foreign direct investments. The United States and Japan continued to be major investors.
Applications of Nonhierarchical Clustering
We illustrate the nonhierarchical procedure using the data in Table 20.1 and an optimizing partitioning method. Based on the results of hierarchical clustering. a three-cluster solution was prespecified. The results are presented in Table 20.4. The “Initial Cluster Centers” are the values of three randomly selected cases. In some programs. the first three cases are selected. The classification cluster centers are interim centers used for the assignment of cases. Each case is assigned to the nearest classification cluster center. The classification centers are updated until the stopping criteria are reached. The “Final Cluster Centers” represent the variable means for the cases in the final clusters. In SPSS Windows. these are rounded to the nearest integer.
Table 20.4 also displays “Cluster Membership” and the distance between each case and its classification center. Note that the cluster memberships given in Table 20.2 (hierarchical clustering) and Table 20.4 (nonhierarchical clustering) are identical. (Cluster 1of Table 202 is labeled cluster 3 in Table 20.4, and cluster 3 of Table 20.2 is labeled cluster 1 in Table 20.4.) The “Distances Between the Final Cluster Centers” indicate that the pairs of clusters are well separated. The univariate F test for each clustering variable is presented. These F tests are only descriptive. Because the cases or objects are systematically assigned to clusters to maximize differences on the clustering variables, the resulting probabilities should not be interpreted as testing the null hypothesis of no differences among clusters.
The following example of hospital choice further illustrates nonhierarchical clustering.