A Methodology for Customer Segmentation Using Existing Product Category Schemes and The SAS ® System

Nearly every retailer (or e-tailer) has a methodology for categorizing the products that they sell. However, most of them do not have a clearly defined system (or any system for that matter) for categorizing their customers by the types of products they buy. This is unfortunate since, in the rapidly evolving world of customer relationship management, this level of customer understanding is critical. This paper will develop a methodology for turning product categorization on its ear (so to speak) and into a customer segmentation scheme that is not only straightforward in its production, but also simple in its interpretation. In addition to the development of customer segments, this paper will also present methods for evaluating the performance of segments from a financial perspective. Enhanced targeting is a natural consequence of this, as well as similar evaluation techniques that will suggest themselves. The paper is designed to be ‘soup-to-nuts’ covering everything from the data manipulation necessary to begin the analysis, all the way through to the financial calculations used to evaluate the segmentation performance. As such, the PROC SUMMARY and PROC SQL from BASE SAS will be prominently utilized, as will PROC FASTCLUS from SAS STAT. The Issue: In most retail marketing shops the focus is product centric. That is, customers get offers based on what they bought most recently, or sometimes what they bought only once. The problem with this type of marketing is that it is highly susceptible to unusual or seasonal purchases (such as gifts). In the brick and mortar world this can be annoying to customers who receive mail offers for products they have no interest in buying. In the ecommerce world, this can be devastating. Because of the relative ease with which customers can opt-out of all future email solicitations, an e-commerce company that does not target effectively will find it’s lifeblood list of customers to whom it may solicit rapidly dwindling. One solution to this is to develop a profile of the types of products that customers tend to buy. Pragmatic Segment Development: One way to develop profiles of customers is to examine, individually, all the products that each customer has purchased over his/her lifetime of interaction with the company. While this approach is certainly thorough, it has the drawback of not being very easy to use since the entire order history of each customer under consideration must be scrutinized each time a new campaign is run. Furthermore, relatively new customers are not treated well under this scenario since they have little history on which to base offer generation. Another approach, which this paper advocates, assumes that while all customers are certainly different, there exist only a few broad categories in terms of the type of items customers tend to purchase. By leveraging all available customer order history, general segments can be developed which can then be used to predict the profile of new customers. This scheme is also easy to use since only a few customer segments and their corresponding profiles need be considered. Hypothetical Example: The easiest way to understand the process proposed herein is through an example. Let us suppose that the online camera store ‘Shutterbugs’ wishes to develop a profile of the types of customers that shop at their site. They currently have a system for classifying their products through a 3-byte field stored in their database; The first byte indicates the high-level type of product, the second it’s brand, and the third indicates the specific kind of product. That is, every product can be completely defined by these 3 hierarchical qualifiers. The possible values for these bytes are shown in Table 1. Note that this example is intentionally over-simplified for ease of presentation. Actual schemes may be significantly more complicated. Statistics and Data Analysis