تجربه گرایی هدفمند: نحوه مدل سازی تصادفی شکل گرفته در پژوهش بازاریابی صنعتی

It is increasingly recognized that progress can be made in the development of integrated theory for understanding, explaining and better predicting key aspects of buyer–seller relationships and industrial networks by drawing upon non-traditional research perspectives and domains. One such non-traditional research perspective is stochastic modeling which has shown that large scale regularities emerge from the individual interactions between idiosyncratic actors. When these macroscopic patterns repeat across a wide range of firms, industries and business types this commonality suggests directions for further research which we pursue through a differentiated replication of the Dirichlet stochastic model. We demonstrate predictable behavioral patterns of purchase and loyalty in two distinct industrial markets for components used in critical surgical procedures. This differentiated replication supports the argument for the use of stochastic modeling techniques in industrial marketing management, not only as a management tool but also as a lens to inform and focus research towards integrated theories of the evolution of market structure and network relationships.

When asked why he robbed banks, Willie Sutton, a notorious bank robber, is reputed to have replied by saying “that's where the money is”. The saying still resonates as an injunction to heed the most likely explanation. Indeed, physicians are taught “Sutton's Law” as a warning to seek the most likely diagnosis first1 (Chang, 2009). What lessons does “Sutton's Law” have for developing and integrating theory in industrial marketing management research? The search for a valuable contribution should start in the areas where that contribution is most likely to be found — to look for the banks. General theories seek to integrate middle level theories in order to explain a wide range of behavioral phenomena, independent of context (Hunt, 1983). This paper argues that well-established empirical regularities provide a starting point for integrating theory and form a solid foundation for higher level explanation — that is, they show us where the banks are. The observation of empirical patterns is an opportunity for guiding further research in order to uncover causal mechanisms and to “delve into the underlying processes so as to understand the systematic reasons for a particular occurrence or non-occurrence” (Sutton & Staw, 1995, p. 378). When those same patterns are repeated in different contexts, industries, firms and relationships, then we have the basis for integrating the causal mechanisms over these different situations. While it is entirely possible that similar empirical patterns may arise from completely different causal mechanisms in different contexts, we argue that the most likely solution starts from the assumption that similar phenomena have similar generative mechanisms and that integrated general theory is most likely to emerge from a research program guided by a common understanding of the ‘explananda’ and the nature of the theories that provide the explanations. The early Industrial Marketing and Purchasing (IMP) research is an excellent example of “Sutton's Law” in action. The original IMP researchers first located their bank — the emerging body of empirical evidence indicating the existence of stable long term relationships between individually significant buyers and sellers (Håkansson & Wootz, 1979). The empirical evidence arising from a study of almost 900 buyer–seller relationships across five European countries (Cunningham, 1980) provided the starting point for the research which resulted in the interaction approach (Håkansson, 1982) which in turn became the “most likely solution” at the heart of the IMP research program. As more researchers adopted the IMP approach, the theory became broader and deeper but at its heart it retains the principles of the interaction approach and the associated assumptions about the nature of buyers and suppliers and their network relationships (Ford & Håkansson, 2006). In a similar manner, stochastic theories of consumer choice have emerged to describe, model and explain regular patterns of buyer behavior. Such patterns have been observed across a wide range of consumer markets, from packaged goods to durables (Uncles, Ehrenberg, & Hammond, 1995). Applied to organizational markets, established consumer modeling techniques can provide insights into the dynamic nature of the portfolio of relationships between buyers and suppliers. Analysis of the exchange behavior in multiple buyer–seller relationship dyads detects patterns of structural change and provides a market “norm” against which to benchmark individual relationships (Gadde & Mattsson, 1987). However, previous studies have focused on multiple category suppliers to a single focal firm (Dubois et al., 2003 and Kamp, 2005). In contrast, the analysis in this paper presents a study of multiple buyers and suppliers operating in a single category, demonstrating the power of analysis of the macroscopic patterns of behavior to identify and interpret structural changes and the impact of these changes on individual buyer–supplier relationships. This paper argues that in addition to describing emergent aggregate behaviors in an organizational purchasing context, the use of such models can direct further research and development of theory to explain behavioral phenomena that repeat across firms, industries and business types. We present two empirical examples that use a stochastic model to analyze behavior in public healthcare procurement. Our approach is to compare the patterns predicted by the chosen stochastic model with actual purchasing behavior. Different forms of market structure are characterized by different observed purchasing patterns, informing and guiding further research to help to uncover the structures and generative mechanisms that help explain the observed phenomena. The deviations between the model predictions and the observed behaviors can also be interpreted in terms of the assumptions underlying the stochastic model; this in turn provides insights into the nature of interactions in industrial markets. Our approach helps to address the problem of limited progress in attaining theoretical unity in the understanding of buyer–seller relationships and industrial networks through the use of a stochastic model as an integrating mechanism for theory development. The paper is structured as follows. Following this Introduction, we provide an explanation of how purposeful empiricism helps integrate theory and hence “contributes to general theory development in industrial marketing research” (Peters, Pressey, Vanharanta, & Johnston, 2013). In Section 2.1 we present a brief overview of the NDB-Dirichlet stochastic model (subsequently referred to as the Dirichlet) before examining in Section 2.2 how its core assumptions can be interpreted within a context of extended networks of long term interorganizational relationships in business-to-business markets (subsequently referred to as the markets-as-networks approach). Within this section we indicate how the Dirichlet provides a theoretical lens through which to view any market, focusing attention on large scale regularities that repeat across different contexts and so contributing to the goal of “attaining theoretical unity” in our understanding of buyer–seller relationships (Peters et al., 2013). Two empirical examples are presented in Section 3 illustrating how the large scale patterns predicted by the Dirichlet provide a mechanism to describe market structures that can be discussed in terms of relationship interdependence and connectedness. The empirical data are taken from a three year longitudinal study of purchasing surgical consumables in a public sector collaborative procurement organization. The study identifies two management interventions designed to influence purchasing behaviors, one initiated by a supplier and the other by the purchasing organization. The analysis of the purchasing patterns before, during and after these interventions provides valuable insights into market making and the extent to which purchasing patterns can be changed within the constraints of an established network of relationships. Section 4 discusses how the observation of regular patterns of purchasing behavior, and just as importantly, deviations from these regular patterns can direct further exploratory and explanatory research to uncover the underlying portfolio of relationships, structures and generative mechanisms that give rise to the regular patterns, representing the purposeful empiricism in the paper's title. In contrast to blind empiricism and the development of theory in isolation, this purposeful empiricism directs the development of theory towards explaining empirical regularities that are replicated across different firm, business and industry contexts, with an increased likelihood of developing more unified theoretical understanding. The paper proposes using the large scale regularities described by the Dirichlet as a guiding structure to direct and integrate further research. If phenomena repeat across different business contexts, the most likely explanation is that the phenomena have similar underlying mechanisms, thereby providing a basis for more general theory. The paper makes three contributions to the industrial marketing research literature. First, we present a highly differentiated replication of the Dirichlet in an organizational market where the patterns of buyer–supplier interaction are dynamic. Our second contribution is to show how the theoretical benchmarks predicted by the Dirichlet can deliver insights into changing market structures and thereby identify changes in the network of relationships. Our third contribution uses the empirical examples to demonstrate how the Dirichlet model provides a theoretical lens to focus analysis on specific situations. In particular, deviations from the Dirichlet benchmarks point to violations of the Dirichlet assumptions which in turn can direct analysis towards the underlying reasons for why the assumptions may not hold in specific circumstances.

The paper contributes to industrial marketing management research firstly via a differentiated replication of the Dirichlet in a collaborative purchasing environment, characterized by substantial changes in market share and purchasing frequency over an extended period (Uncles & Kwok, in press). The resulting empirical analyses show how the phenomena observed through stochastic modeling techniques represent fruitful areas for the development of explanatory theory. The second contribution of this paper is to provide empirical evidence to demonstrate how regular patterns of behavior typical of Dirichlet markets are generated in industrial networks and how such patterns inform our understanding of interactions in network relationships. In particular, the analysis of the large scale regularities that emerge from interactions within a “network of overlapping networks” of multiple buyers and sellers in two product categories provides an overview of market structure that is not possible from a focal firm perspective. The application of the stochastic modeling approach over an extended period also permits changes in the relationship portfolio to be monitored and analyzed. Understanding how changes are manifest in observable purchasing behavior allows the underlying mechanisms that result in the formation of new relationships, the disruption of existing relationships and the stability of other relationships to be studied. Stochastic modeling and industrial marketing share important assumptions about the individuality of network participants and the constraints on independent actions imposed by network interdependence or the steady purchasing behaviors of experienced buyers (McCabe & Stern, 2009) and this paper extends the argument that both research perspectives can and should be used together to identify and explain organizational marketing and purchasing behavior. In our third contribution we demonstrate how the deviations between the observed purchase behaviors and the theoretical predictions for the network provide another source of phenomena for further investigation and explanation, improving our understanding of network relationships. In order to develop more general theories of industrial marketing, the onus is on researchers to focus empirical analysis and theory development in areas that build and extend the generalizability of existing theories or explain generalizable phenomena. This paper provides an empirical basis for the sequential application of stochastic modeling and in-depth analysis to make progress in the development of more integrated theory. The Dirichlet acts as an integrating framework to identify large scale regularities of interest that repeat across different industries, firms and relationships. Uncovering and explaining the associated structures and mechanisms that bring about the observed phenomena support the development of theories that apply across a wide range of contexts. This sequential application is entirely consistent with the realist ontology of much industrial marketing research (Mingers, 2003). In addition to these direct implications for researchers in industrial markets and stochastic modelers, the adoption of modeling techniques has important implications for simulation of business networks. Agent-based computer simulations may use the regular patterns of the Dirichlet and the theoretical ideas about the generative mechanisms that bring about these patterns to simulate network behaviors (Welch & Wilkinson, 2002). Such virtual networks have the potential to be valuable tools for experimental research and teaching. There are also important implications for management. The empirical analysis has shown how managerial intervention can disrupt established relationships and change long run purchasing behaviors. The understanding of how such interventions can work and of the nature of successful and unsuccessful interventions can provide normative guidance for managers seeking to bring about change in networks. Interventions must be sympathetic to the relationship conditions (Ritter et al., 2004) and must go with the grain of existing purchase behavior. A key lesson of “Dirichlet” markets with steady long run purchase rates is that gaining market share in a specific product category will generally mean following Willie Sutton's dictum and “finding new banks” i.e. acquiring more customers rather than persuading existing customers to buy more. The combination of stochastic modeling and industrial marketing research presents an iterative process of theory development and empirical analysis that will inform our understanding of relationships in industrial networks. In turn, stochastic modeling gains from a deep and rich analysis to develop theoretical insights into the generative mechanisms underlying observed patterns and regularities. As Epstein (2008, p. 12) observes, “models can surprise us, make us curious, and lead to new questions”. Purposeful empiricism takes these new questions and uses them to move the research agenda beyond structure and isolated studies towards a more integrated understanding and explanation of behavior in organizational markets.