Fuzzy set theory models have considerable potential to address complex marketing and B2B problems, but for this methodology to be accepted, models require validation. However, there is relatively little detail in the literature dealing with validation of fuzzy simulation in marketing. This limitation is compounded by the difficulty of using case-based and qualitative evidence (data to which fuzzy models are well suited) when applying more general validation. The chapter illustrates a fuzzy model validation process using small-N cased based data and concludes with recommendations to assist researchers in validating their fuzzy models.
TopIntroduction
Researching B2B marketing poses difficult issues, particularly when investigating business networks and their interactions. We are often faced with a small number of organizations (small-N) and a large number of variables, many of which are ‘fuzzy’. For example, actors themselves often take multiple roles such as manufacturer, new product developer, and information broker. Therefore, even classifying actors is problematic when using multivariate research techniques. Case study techniques overcome these issues as they allow for rich variable descriptions (Piekkari, Plakoyiannaki, & Welch, 2010). Yet, cases have disadvantages, notably the difficulty in conducting case comparisons or investigating multiple possible scenarios. Fuzzy set theory overcomes many of these issues (Donzé & Meier, 2011). Yet, even though fuzzy set theory opens up new ways of investigating business networks, it can only be achieved if the fuzzy models that are used are validated (and verified) within real-world contexts.
Fuzzy set theory has not been used extensively within the marketing and management disciplines, with even less papers reporting model validation (and verification), despite its importance (Richiardi, Roberto, Saam, & Sonnessa, 2006). The diversity of views about validation processes has undoubtedly contributed to this omission. As Petty (2009, p. 40) observes, “the practice of verification and validation is as varied as the subjects of the model involved”. In the social sciences, various authors have documented the difficulties associated with this task observing that the context to be modeled is usually large and complex with “only a small number of observable instances and data sufficiently detailed for validation” (Petty, 2009, p. 140).
Previous literature highlights the need to develop a suite of “best practices” to validate and verify computational simulation models (Louie & Carley, 2008; Richiardi, et al., 2006; Wilensky & Rand, 2007; Windrum, Fagiolo, & Moneta, 2007), yet given the unique characteristics of fuzzy set theory some of these “best practices” cannot be applied in their current form. Reporting validation and verification practices is critical if fuzzy set theory models are to be accepted by the wider academic community and their findings used by practitioners (Louie & Carley, 2008; Maguire, McKelvey, Mirabeau, & Öztas, 2006). This chapter outlines a process for business marketers to validate their fuzzy set theory models (at least in part).
The validation process we describe has a number of advantages for fuzzy set theory modelers, in that it: (1) allows model builders to develop the fuzzy rule based system without working from a full factorial set of rules, thus improving validation efficiencies; (2) checks for consistency and coverage of the fuzzy rule based system improving model sufficiency; and (3) validates the simulation model within real world data sets. We note that working from the full factorial set of rules is extremely resource demanding and that model builders must still complete a validation process to compare to real world data.
The chapter begins by reviewing general processes of verifying and validating models and goes on to identify techniques and methods that are particularly relevant for fuzzy set theory models. We then consider how researchers can establish the extent to which their model fits the “real-world” and contribute to the discourse on model validation by illustrating a process that uses case-based evidence. Here we focus on the use of linguistic information, consisting of nuances and variation, in fuzzy simulation methods. Fuzzy set theory is well suited to dealing with the ambiguity and natural language used in this type of data (Donzé & Meier, 2011). The case describes the innovation and commercialization processes for various photovoltaic technologies and focuses on a lead group of researchers who have impacted the development of solar technologies on the world stage. We conclude the chapter with insights and recommendations for researchers when embarking on the validation process.