A Systemic Approach to the Database Marketing Process

The role of database marketing (DBM) has become increasingly important for organisations that have large databases of information on customers with whom they deal directly. At the same time, DBM models used in practice have increased in sophistication. This paper examines a systemic view of DBM and the role of analytical techniques within DBM. It extends existing process models to develop a systemic model that encompasses the increased complexity of DBM in practice. The systemic model provides a framework to integrate data mining, experimental design and prioritisation decisions. This paper goes on to identify opportunities for research in DBM, including DBM process models used in practice, the use of evolutionary operations techniques in DBM, prioritisation decisions, and the factors that surround the uptake of DBM. Introduction Database marketing (DBM) is often viewed as a tactical marketing tool rather than a strategic one, and as a capability that is easily developed (Cooke, 1994). In fact, DBM is a sophisticated process that requires a systemic approach that integrates a number of organisational departments. Without a systemic approach to DBM, the marketing capability of an organisation may be underdeveloped. The term ‘database marketing’ will be used in this paper to include business-to-consumer applications of direct marketing, where direct contact is made with a customer. DBM is an interactive approach to marketing using individually addressable media and channels (Shaw, 1993). The terms ‘data mining’ and ‘predictive modelling’ are used interchangeably to mean the process of analysing customer data to produce models that predict the behaviour of the population in question. Much of the advanced practice in DBM is performed within private organisations, which may explain partly the lack of articles published in the academic literature that study DBM issues (Blattberg, 1987; Bult and Wansbeek, 1995; DeTienne and Thompson, 1996; Dwyer, 1989). The successful use of DBM as a strategic tool can give competitive advantage to an organisation. The benefits of DBM include improvements in the ability to control costs, plan budgets and measure marketing plan performance, the identification of strategic advantage through better use of customer and market information, the facilitation of long-term customer relationships that increase lifetime value and loyalty, improved accuracy in customer segmentation, and the ability to relate customer characteristics to behaviour (Desai et al., 2001; Desai et al., 1998; Lewington et al., 1996). However, a major inhibitor to DBM is that it requires strategic and organisational changes beyond technological superiority to gain many of the benefits of its capabilities. ANZMAC 2002 Conference Proceedings 2941 The Database Marketing Process DBM has its origins with catalogue retailers based in the USA selling directly to customers. The principal channel used was direct mail, and mailing of new catalogues usually took place to the whole database of customers (DeTienne and Thompson, 1996). Analysis of the results of the mailings led to the adoption of techniques to improve targeting, such as CHAID (ChiSquared Automated Interaction Detection) and logistic regression (DeTienne and Thompson, 1996; Magidson, 1988; Schoenbachler et al., 1997). The addition of centralised call centres and the Internet to the DBM mix introduced the elements of interactivity and personalisation (Kelly, 2000). During the 1990s, the data-mining boom popularised such techniques as artificial neural networks, market basket analysis, Bayesian networks and decision trees (Berson et al., 2000; Coates et al., 1994). The increasing sophistication of DBM requires extension of the existing process models. Current DBM capabilities are still evolving at different speeds within different industries and geographical locations. Barriers to the adoption of DBM include the cost of development, poor data and system quality, a lack of a DBM strategy and a lack of a company-wide marketing orientation (Desai et al., 1998). A research opportunity exists to examine the uptake of DBM and factors that encourage uptake. Database Marketing Models Early DBM models concentrated on a one-off marketing campaign, where there were a number of potential marketing treatments to be offered to different customer segments. This model was extended for a campaign containing new, untested treatments, such as the launch of a new product (Morwitz and Schmittlein, 1997). Different treatments were offered to a sample of the target population in a testing phase before offering the most effective treatments to the remainder of the eligible customer base (Levin and Zahavi, 1996). In Blattberg’s model (1987), marketing is continued to responsive customers. A refinement to this model occurs with the addition of a predictive modelling step to refine the targeting of customers rather than using simple metrics of responsiveness and average order value (Berson et al., 2000). The three-element model proposed in Lewington et al. (1996) identifies the database, a market modelling process, and a mechanism for performance measurement feedback as being the important elements for effective DBM. Although it covers the major elements of DBM, this model does not illustrate the complexity of the process required for successful DBM. Taking a more detailed view of DBM is the ‘closed loop’ model, in which DBM is seen as part of a process of organisational learning (Hirschowitz, 2001). Hirschowitz’s model identifies the key steps as being idea generation, analysis, customer targeting and treatment allocation, campaign creation and response measurement, campaign execution, campaign monitoring and campaign refinement. The monitoring of campaigns feeds information back to the campaign refinement and idea generation stages. Metrics to evaluate the success of the campaign are defined as part of the campaign development process. However, the closedloop model fails to consider the integration of experimental design and predictive modelling within an environment in which each campaign is a part of an overall marketing strategy. A recent addition to the existing paradigm, assisted by technological innovation and the advent of the Internet, is for the marketing effort to take place opportunistically: that is, when ANZMAC 2002 Conference Proceedings 2942 the customer contacts the organisation, a list of campaigns for which the customer is eligible is delivered immediately (Herschel, 2001). A Systemic Model In the transformation of DBM from being a mass marketing channel, marketers are making frequent, highly targeted offers over a variety of marketing channels, including direct mail, call centres, email and the world-wide web (Doyle, 2000). Consequently, the amount of DBM activity is increasing within an organisation. We contend that as this marketing activity increases, a number of issues become paramount in importance: once a marketing campaign has been designed and developed, execution of the campaign process must require little or no human intervention (Levin and Zahavi, 1996); technology used must support development and management of a large number of predictive models, but in a way that supports analysts rather than replacing them (Hand, 1998); the process of monitoring of the campaign and predictive model performance must be automated; and there must be a capability to manage many campaigns concurrently in line with business policy, and to plan customer-focused communications. The systemic process proposed in this paper has been formulated to cope with some of these issues, specifically the management and monitoring of predictive models and campaign performance, extending the models found in the literature. Figure 1 outlines the steps in the systemic process model. Each step takes place at one of the three levels: the program level, the campaign level and the treatment level. The program level is the overarching strategic level at which campaigns and customer communications are coordinated. At the program level there can be a number of campaigns, and a campaign may consist of a number of treatments. At the campaign level, treatments are coordinated to achieve a particular strategic or tactical objective. Predictive model development takes place at the treatment level, identifying those customers most likely to respond to the treatment favourably. Models are refined as responses to treatments are captured and fed back to the marketing database. Experimental design principles are used at the program and campaign design stages to test customers’ responses to varying marketing stimuli, and to collect data for the measurement of campaign performance and the refinement of predictive models. The prioritisation stages manage multiple predictive models to ensure that optimised decisions are made in terms of the allocations of customers to communications channels, campaigns and treatments. Lewington et al. (1996) contend that a comprehensive model of DBM is of interest both to the academic community and practitioners. However, the systemic model – introduced in this present paper – more accurately reflects sophisticated DBM practice than the models found in the earlier literature, including Lewington et al. (1996). It also provides practitioners with a template from which organisations can build their internal DBM capabilities. Our model’s specific areas of innovation are to take an overarching view of DBM that explicitly integrates data mining with experimental design and the need for program-, campaignand customerlevel prioritisation. A more detailed framework for integrating experimental design, predictive modelling and prioritisation decisions for event-triggered and opportunistic campaigns is given in another paper by the current authors. Research remains to be done on the extent to which companies involved in DBM adhere to the process model in Figure 1. ANZMAC 2002 Conference Proceedings 2943 Program Campaign Treatment Strategic alignment of program with strategy Program development Idea generation Predictive model development Campaign design Treatment development Program design Campaign / channel / treatment prioritisation Campaign execution Customer–treatment allocation