In almost all grocery and retail stores, sales are recorded by scanner checkouts. The recording of receipt data provides retailers with large aggregated1 data volumes (Boztuğ and Silberhorn 2006, p. 109 f.). With each purchase, the customer receives a receipt for his purchase, which is primarily used to control the purchase. In the context of receipt data analysis, the receipt can be interpreted as the customer’s vote on the retailer’s range of services (Hertel 1999, p. 38).
In bond data analysis, the content of physical shopping carts is examined. At this point, it should be mentioned that the terms bond data analysis and shopping cart analysis are largely considered synonyms (Buhr 2006, p. 55). The reason for this is that market basket analysis and bond data analysis can hardly be distinguished from each other. At the functional level, market basket analysis includes consumer-oriented analyses. The bond data analysis however covers functional and technical aspects (Städler and Fischer 1999, p. 339). In this paper, the terms bond data and shopping cart data as well as bond data analysis and shopping cart analysis are used synonymously.
Bond data analysis enables retailers to draw conclusions about the buying behavior of customers across all goods (Städler and Fischer 1999, p. 345). In addition, bond data analysis gives retailers the opportunity to show which products are purchased in combination. On the basis of this data, it is accordingly possible to enable better spatial placement of the products in the retail company (Boztuğ and Silberhorn 2006, p. 107).
As a result of constant further development, technology has reached a level at which the processing and analysis of data is no longer a problem in most cases. Nevertheless, the data is not used enough in the planning and control of marketing measures. The existing possibilities are not exhausted because of problems with “questions”, “data sources” and “methods”, among others. Users must be clear about which formulated questions are to be answered with which data and with which method (Schröder and Rödl 2004, p. 519).
In their definition, Boztuğ and Silberhorn assume that bond data ana-lysis is used to determine models that can predict the selection probability of a bundle of many items during shopping. They assume that the construction of a shopping basket implies dependent selection decisions. The analysis of the shopping basket data should reveal and quantify the interconnectedness relationships within a shopping basket and also estimate the selection probability of a particular shopping basket (Boztuğ and Silberhorn 2006, p. 108). An important component of shopping basket analysis is the shopping basket, which is defined as follows: “In its original literal sense, the shopping basket refers to the physical collection of goods procured by a consumer in the course of a shopping transaction at one or more shopping locations” (Knuff 2008, p. 98).
A shopping cart thus represents a customer’s examination of a retailer’s overall offering. This refers to the totality of the items that the customer places in his shopping basket during his shopping trip and purchases at the end. As soon as a customer takes his shopping basket to the checkout and wants to pay for the goods, the goods are recorded and stored with the help of scanning checkouts. This data is referred to as shopping cart data or receipt data (Knuff 2008, p. 99).
At the end of the purchase process, the customer receives a receipt for his purchase in the form of a voucher on which an image of his physical shopping cart is listed with the items he has purchased (Buhr 2006, p. 53). The data generated during the checkout process can be used to determine what the customer purchased at what time and in what quantity. The identification of customers plays an important role here. As soon as they can be identified, the behavior of customers over time can be analyzed (Olbrich 2012, p. 87).
Customers can be identified in several ways. On the one hand, purchases can be assigned to customers or households on the basis of pseudonymized sales data. A pseudonym can be the card number when using a customer card or EC card or credit card. The pseudonym can be supplemented with data about the customer if the customer allows this. If this is not the case, he remains anonymous. In the case of identification by EC cards, it can happen that the card loses its validity and this leads to the customer having a new card number. It is then difficult to perform an analysis over a longer period of time. In the case of loyalty card systems in stationary retailing, in most cases the address data and names can be assigned to the pseudonymous data, so that personalized transaction data are available (Schröder and Rödl 2004, pp. 521-525). At this point, it should be noted that more opportunities arise for receipt data analysis when customers step out of anonymity and their identities are available to retailers (Schröder and Rödl 2004, p. 519).