The car market is a high-involvement, high-information market, in which consumers are expected to go through extensive searches. Cars are highly symbolic artefacts. The marque and model say a lot about the owner, and evidently, a car is far beyond a purely rational, functionally based purchase. However, car manufacturers face a serious problem as worldwide marque loyalty levels, from purchase to purchase, average below 50%, and tend to decline over time. Evidently, the analysis of factors affecting car marque loyalty is a research topic of significant managerial importance. This study attempts to empirically address the structure of marque loyalty in the car market and has a dual objective: First, to relate marque loyalty to a set of consumer characteristics under a theoretical framework, and second, to examine the impact of current car’s attribute-level performance on loyalty. In this direction, this study illustrates the value of Adaptive Network-based Fuzzy Inference System (ANFIS), as a bridge between qualitative and quantitative approaches, in an attempt to identify alternative complex antecedent conditions that give rise to marque loyalty in the car market. The proposed approach offers to conventional correlational quantitative approaches three benefits: (1) asymmetry (i.e., relationships between independent and dependent variables are treated as non-linear/asymmetric), (2) equifinality (i.e., multiple pathways may lead to the same outcome), and (3) causal complexity (i.e., combinations of antecedent conditions lead to the outcome, and hence, the focus is not on net-effects, but on combinatorial-synergistic effects). To demonstrate these merits, ANFIS is compared to a conventional econometric forecasting technique, namely logistic regression.

One of the basic strategic decisions a retailer must make involves the determination of the assortment to offer. Product assortment planning (PAP) involves important decisions related to the determination of variety (i.e., number of categories), depth (i.e., number of stock-keeping units within a category) and service level (i.e., amount of merchandise inventory within a category) in a retailer’s product portfolio (Mantrala et al., 2009; Hübner & Kuhn, 2012). By making optimal PAP decisions, retailers hope to satisfy customers’ needs by providing the right service in the right store at the right time (Nogales & Suarez, 2005). Despite the importance of PAP, several limitations and gaps can be found in existing literature. First, existing research tends to examine analytical solutions that deal almost exclusively with questions of depth, whilst it completely fails to address issues related to variety and service levels (see for example, Mantrala et al., 2009). Second, current literature focuses on a single category of products or services and fails to examine the interplay among various categories that are offered by a retailer. Third, although in reality a retailer might have a different assortment at each store format, the academic literature has focused on determining a single assortment for a retailer, which could be viewed as either a common assortment to be carried at all stores or the solution to the PAP problem for a single store (Kök et al., 2006). Finally, Private Labels (PLs) have been widely neglected in existing PAP literature, despite the fact that retailers consider PLs as a powerful competitive tool (Nogales & Suarez, 2005). This paper corrects for omissions of existing PAP research by introducing a new innovative method, namely Differential Evolution (DE). More specifically, the proposed mechanism facilitates simultaneously, strategic PAP decisions, related to the determination of a) optimal variety of PL categories in a retail grocery store, b) optimal service level of PL merchandise within each category, and hence, c) optimal balance between PLs and National Brands (NBs) in a retailer’s product portfolio. The interrelated issue of assortment adaptation across different store formats is also considered. Differential Evolution (DE) is an evolutionary, population-based algorithm, for global optimization over continuous spaces. It was first introduced by Storn and Price (1997), and has been extensively applied to a wide domain of optimization problems due to its ability to efficiently handle non-differentiable, nonlinear and multimodal cost functions. DE is based on the Darwinian theory of Evolution (Engelbrecht, 2007). In a world with limited resources and stable populations, each individual competes with others for survival. The individuals with the best characteristics will more probably survive and reproduce. Those desirable characteristics (a) are passed on to their offspring, (b) are inherited by the subsequent generations, and (c) over time will become dominant among the population. During the production process of a child organism, random events may cause random changes to its characteristics. If these altered characteristics benefit the organism, then the likelihood of survival for the organism is increased. In accordance to this, DE works with a group (population) of candidate solutions to the problem (individuals). The algorithm searches for the global optimum through an iterative process. In each algorithm’s iteration the individuals produce offspring through crossover, and some individual’s characteristics are randomly altered through mutation. The strongest (fittest) individuals of the new population survive to the next generation. Our proposed mechanism is implemented to empirical data that have been collected for the purposes of a large-scale telephone survey research examining consumer buying behaviour in the grocery market of a European metropolitan area. A highly structured questionnaire was developed and data were collected from a random sample of 1,928 supermarket customers. The telephone survey was conducted by the Computer Assisted Telephone Interviewing (CATI) facilities of a local university. In total, we examined consumer preferences for a set of twelve product categories that are usually available in a typical supermarket. We implement our DE algorithm to find optimal solutions (i.e., PL service level per category) in the entire dataset and for three store-formats separately (i.e., large supermarkets, discount supermarkets, small local supermarkets). The derived optimal solution for the entire dataset suggests that retailers should mainly focus their efforts on providing extensive PL service levels in product categories such as disposable paper products and packaged foods, and also maintain a decent PL presence in categories such as bakery, laundry, household cleaning products, tea-coffee, and non-alcoholic beverages. On the other hand, the introduction of PLs in categories such as frozen foods, personal hygiene products and clothing products would not be advisable. Regarding the adaptation of PL service levels across store formats, interesting conclusions can be drawn. For example, managers of large mainstream supermarket chains must offer extensive PL service levels in categories such as disposable paper products and packaged foods, whilst they should maintain a decent PL presence in categories such as laundry and dairy products. In line with our expectations, discount retailers are expected to provide broader varieties of PLs, because in addition to the PL categories offered by mainstream supermarkets, discounters must also provide extensive PL service levels in household cleaning products. It is suggested that discounters not only must offer broader varieties of PLs, but also more extensive service levels within those varieties. Finally, the derived optimal PL service levels in most product categories of local supermarket chains are extremely low. This finding indicates that local grocery stores should concentrate their efforts in providing a narrow variety of PLs, by focusing on few categories, such as packaged food and laundry products. In the light of the entire discussion, we suggest that evolutionary analysis can reveal exciting opportunities not merely for new research, but for novel, revolutionary views of market behavior.