رویکرد تقسیم بندی پویا برای هدف قرار دادن و سفارشی کردن کمپین های بازاریابی مستقیم

An important aspect of customer relationship management is the targeting of customer segments with tailored promotional activities. While most contributions focus on the selection of promising customers for targeting, only few authors address the question of which specific differential offers to direct to the selected target groups. We focus on both issues and propose a flexible, two-stage approach for dynamically deriving behaviorally persistent segments and subsequent target marketing selection using retail-purchase histories from loyalty-program members. The underlying concept of behavioral persistence entails an in-depth analysis of complementary cross-category purchase interdependencies at a segment level. The effectiveness and efficiency of the proposed procedure are demonstrated in a controlled field experiment involving the targeting of several thousands of customers enrolled in the loyalty program of a “do-it-yourself” retailer. Our empirical findings provide evidence of significant positive impacts on both profitability and sales for segment-specific tailored direct marketing campaigns.

The retail industry is characterized by fierce price competition among companies that offer rather similar product assortments and pursue aggressive promotional policies within given retail formats (Corstens& Corstens, 1995; Kahn & McAlister, 1997). Retailers are collecting huge amounts of personally identifiable point-of-sale (POS) transaction data dissembling rich information about customers’ purchasing habits (e.g.,sizes, spending values, or compositions of shopping baskets). The individual purchase histories collected from the customers enrolled in the program can be linked back to store and marketing data, sociodemographic background characteristics, and additional survey or feedback information (if available). Within advanced concepts of customer relationship management(CRM), this database of consolidated data sources plays a central role in analyzing and planning targeted direct marketing actions (Winer, 2001).As our brief review in the next section will show, considerable advances have been achieved in the field of target segment selection for direct marketing purposes;however, the authors could not detect any contributions to the academic marketing literature that convincingly address the question of which specific differential offers (in terms of merchandise types or product categories to be featured or subjected to rewards) to direct to the customer segments that turn out to be worth targeting.This article attempts to utilize the multicategory nature of choice decisions made by individual shoppers throughout their shopping trip histories to assist direct marketers in selecting who to target with what specific offer(s). In doing so, both the process of segment formation and the customization of targeted cross- and upselling campaigns are based on a measure that quantifies a customer’s “interest” in particular(combinations of) product categories. As in the case of the Tesco Clubcard program (Humby & Hunt,2003) or the indications provided by Pearson and Gessner (1999), resolution of this issue is sometimes claimed by practitioners, but there is a lack of a more thorough treatment in the academic literature.Furthermore, specification of such “interest measures”and especially consideration of cross-category effects are mostly accomplished on a rather ad hoc basis or guided by pure managerial intuition.In contrast, we advocate a more general and datadriven approach for quantifying a specific customer’s tendency to symptomatically (re)purchase distinctive combinations of product categories included in a retail assortment. The latter is denoted as “behavioral persistence,” which will be evaluated based on an indepth exploratory analysis of shopping basket histories.The remainder of this article proceeds as follows:Following a brief discussion of current practices in target market selection within loyalty programs and their specific deficiencies, an overview of previous research on analyzing cross-category purchase interdependencies based on shopping basket data is provided.Next, we propose a flexible and dynamic approach to derive segments of customers who are behaviorally persistent in the aforementioned sense.The proposed two-stage modeling framework includes a data-compression step of the observed shopping basket data and resolves the subsequent target group selection. In an empirical application study, targeting effectiveness and efficiency of the proposed procedure are evaluated in a controlled field experiment involving segment-specific adapted direct mailings to a large customer sample of a “do-it-yourself” (DIY)retailer. Finally, we discuss conclusions and outline some suggestions for future research.

We introduced a novel approach for assisting retail marketing managers in planning segment-specific,customized direct marketing campaigns. In contrast to most existing approaches, both target-segment selection and customizing the content of a direct mailing are addressed. The proposed two-stage procedure consists of a data compression step that serves for deriving prototypes of distinguished cross-category interdependencies among the categories included in a retail assortment. Due to their adaptive nature, these prototypes can be updated continuously for evolving time periods. Equally, the second stage entailing the construction of behaviorally persistent segments is flexible enough to be dynamically adjusted. Furthermore,the degree of behavioral persistence responsible for segment formation can be controlled easily by management via suitable choice of a threshold parameter.The empirical performance of our approach is demonstrated for two different target segments selected from a DIY retailer’s customer database. Both segments exhibiting a certain minimum degree of behavioral persistence were selected for targeting in a controlled direct marketing experiment. The empirical findings support the usefulness of the proposed procedure in terms of impacts on both sales and profits. Compared to a randomized customer sample, our recommended segments were targeted more effectively and efficiently.A research agenda for further validation of the empirical performance of the presented methodology should include the following tasks: First, an extension of segment-specific targeting campaigns to segments other than the two illustrated in the present application study is advisable. Second, the approach should be evaluated for different target segments resulting from varying degrees of the behavioral persistence measure proposed in this article. This also could include a combination with more traditional methods for target-segment selection such as RFM or CLV metrics. Third, the linkage with predictive approaches to market basket analysis of the type proposed by Manchanda et al. (1999) or Russell and Petersen(2000) possibly could further enhance the managerial implications of our approach to segment-specific category selection for target marketing. Some nonexperimental findings in this direction are available in a recent work by Botzug and Reutterer (2006).Finally, a comparison to the impacts of one-to-one targeting strategies and applications to other retail or nonretail industries would be helpful.From a more conceptual perspective, our proposed measure of behavioral persistence is restricted to the diagonal elements of a shopping basket classes’ switching matrix for the purchase histories of the available customer database. Hence, further research endeavors also could be devoted to studying customers’ latent basket class switching behavior. In addition to behaviorally persistent segments, identification of significant switching paths across the derived partition of prototypical shopping basket classes could serve as a valuable basis for targeting switching segments in accordance with their basket class transition probabilities.