1 Introduction

Knowledge created by mankind seems to grow exponentially. The amount of knowledge accumulated in the past one hundred years is possibly higher than the one previously created in the whole period of human civilization. Available information reaches astronomical numbers nowadays, as a mere internet search can verify. How can we classify this knowledge according to its origin or the way it was generated? How much of this accumulated information is, in fact, reliable, valid, trustworthy? Such questions refer to how we acquire knowledge and they point to possible types of knowledge: popular, theological, philosophical, and scientific knowledge.

Popular knowledge simply stems from human observation and interaction with everyday phenomena. It is based on personal experiences gathered non-systematically through simple inference, without critical consideration of the observed phenomenon. As an example, a person looks at the sky and infers that it will possibly rain. Or a farmer learns how to grow lettuce based on the results of his or her previous experience.

Theological knowledge, on the other hand, is based on the core aspect of faith. One believes that the resulting propositions correspond to an absolute truth. This type of knowledge is assumed to be infallible because it is considered a supernatural truth. One example is eternal salvation.

The third type is philosophical knowledge. Its core derives from reflection, which leads to concepts and ideas focused on immaterial, subjective issues that should be relevant to human beings. This contemplation seeks some logical coherence with reality, asking questions such as: what is happiness, social welfare, democracy?

Finally, scientific knowledge results from an understanding derived from systematic use of complementary methods that allow to empirically test, probe, and verify phenomena from different knowledge fields. One must use the appropriate methods, following logical, critical, analytical thinking, to analyze facts, situations, and observations, so that their veracity or falsifiability can be proven or refuted. Fallibility is, therefore, the crux of scientific knowledge creation.

Scientific knowledge is the focus of this article, which we hope will be useful for students, researchers, and scholars of applied social sciences (business administration in particular) in their research efforts. We expect it to be especially helpful in studies in the fields of marketing and consumer behavior.

CAPES (Coordination for Improvement of Higher Education Personnel) is the governmental agency responsible for organizing, evaluating, expanding, and consolidating Master’s and PhD programs throughout Brazil. They have a database of all the master theses and doctoral dissertations defended in the last decades. In Business Administration, data are available from 1987 on. Between 1987 and 2017, 63,870 works (11,374 dissertations and 52,496 theses, including MBA dissertations) were stored there (Capes, 2018). The production from the last 5 years corresponds to about 30% of the total, in synch with the expansion of the number of graduate courses in Business Administration in Brazil. The volume of dissertations and theses produced in this area in Brazil is steadily growing. The main question is about the quality of these works: which theoretical and /or methodological contribution do these dissertations or theses bring, which are the implications for business management, and the impact of their results for society in general? This is certainly a vast and relevant field of study, especially considering the amount of public and private resources invested in the education of these researchers.

It is not the purpose of this article to discuss this broad and complex question. The idea is to present a methodological tool hoping to contribute to improving the quality of these works. The reflections exposed here are the result of almost forty years of doing research in marketing, acting as doctoral advisor, examining theses and dissertations, and evaluating articles submitted to congresses and scientific journals. This experience has helped devise a way to better articulate and connect theories, methods, results and implications of empirical research.

After finishing a research job, when your head ‘hits the pillow’, it is usual to think about the potential weaknesses of your research, what you could have done differently or better. But, at that point, the work is done. You cannot turn back in time, only use this experience to improve planning and execution of future empirical research. Therefore, it may be a good idea to suggest a methodological tool to help researchers critically consider the ‘interconnection’ of theory, method, results, and research implications. This is the core issue of this article. Even if this connection is almost obvious, in practice we find inefficiencies that could have been avoided with better planning of the research steps.

This article has three sections. The first one addresses the process of marketing knowledge creation. The second section discusses a proposition of a methodological association matrix for use in empirical research. Then final considerations for marketing researchers are discussed.

2 Marketing knowledge creation

Knowledge is created when we try to clarify a fact, situation, or phenomenon that we observe or interact with. In marketing studies, the locus of research lies in transactions, relationships, and consumption of products, services, ideas, or locations. Understanding why, how, where, in what situations, and what underlying factors influence these processes is the focus of marketing research. The core of scientific research is a search to find valid, reliable answers to questions about the elements, factors, interactions, and results of these processes. And these questions must be clear and objectively formulated and delimited.

As the process of knowledge generation, dissemination, and disposal is dynamic in time, we must always ask whether there are adequate secondary data to answer the doubts or problems we face. If the answer is yes, this means the end of the research process, as we already have the answers to the problem. However, secondary data do not always exist, and if they do, they may not fully answer the questions or have the desired quality. Therefore, secondary data should be carefully evaluated in terms of their quality and adequacy to address a research problem. In case the risk threshold is inadequate, answers can only be obtained through primary data. Figure 1 illustrates this situation.

Let us consider a situation, fact, or phenomenon requiring collection of primary data for its analysis. Based on studies by Apostel (1960), Barberi (1967), Mitroff et al. (1974), and Lilien (1975), and following a framework by Bunge (1974), a Master´s thesis (Mazzon, 1978) introduced a conceptual framework to evaluate and compare marketing models. This framework comprised 52 variables associated with four dimensions: conceptualization of the phenomenon under study, modeling, solution, and implementation of the results.

From a hypothetical-deductive perspective, the first step in studying a marketing phenomenon is to properly circumscribe it. What do we want to study? What is the study expected to consider and what should be dismissed? In which conditions should the phenomenon be investigated? Numerous phenomena can be studied in the field of marketing, particularly in consumer behavior. Some examples: what is the value of a brand from the perspective of several stakeholders? How will a target audience respond to an ad on social media and how much will each factor influence this response? Every issue of the main academic marketing journals presents countless phenomena, situations, and facts of the most varied nature duly analyzed using the scientific method.

After properly defining the phenomenon of interest for the study in terms of research problems (the answers are expected to fill a relevant gap in the state-of-the-art knowledge on the theme), the next step is to design a framework or theoretical model based on related theory. The role of theory is to describe the structure and organization of knowledge in a specific field. The model aims to represent a particular construction using the theory as a reference framework. Theories deal with observed, observable, and even unobservable elements, whereas models deal with references to an observable universe and evidences from the observed world.

This activity of conceptualization is the field of marketing philosopher researchers, who essentially reflect critically and conceptually on the phenomenon under scrutiny. Ideas and concepts are defined and explored. The systematized body of conceptual knowledge serves as a foundation for building theoretical models whose main purpose is to propose theory-based research hypotheses about problems from a particular field.

Following the hypothetical-deductive perspective, from a theoretical model one can design an operational model that allows testing the proposed research hypotheses using real-life evidence. Modeling to create operational models is the field of work of the marketing scientist, whose job is to test and verify whether the formulated hypotheses are consistent with market evidence. This is one of the most developed areas in marketing. Multiple models have been designed over the past 60 years, such as consumer judgment and brand choice models, consumer product or brand attitudes, advertising sales response models, new product sales forecast, consumer responses to service failures, and brand loyalty, among countless others. Such models are operationalized mostly through questionnaires using adequate scales, scenarios and manipulations in experimental designs, and observation processes and instruments, among others.

The third activity is to find a solution for the operational model. This means yielding results, usually through mathematical or computing techniques such as factoring, classification, discrimination, prediction, conjoint analysis, canonical correlation, multidimensional scaling, structural equation modeling, simulation, association and variance statistical tests, as well as the more recent machine learning and artificial intelligence techniques. These applications are becoming more widespread thanks to increasingly powerful hardware, simpler programming languages, and more user-friendly software. However, while these aspects facilitate and stimulate using these tools, they may raise concerns regarding their inadequate usage. Given its characteristics, this activity also pertains to the field of work of the marketing scientist.

The last activity is managing or implementing the results achieved in solving the operational model. This means that the results may be used in planning, so that organizational managers can act on the marketing locus, promoting transactions, relationships, or changes in consumer beliefs, attitudes, or behaviors. In other words, they may do a better marketing job. Therefore, research materializes into actions that impact not only marketing management efficiency and effectiveness, but mostly consumer and society welfare. This the field of management professionals.

Although the model has been described from a deductive perspective, it can be also analyzed using the inductive approach. Thus, marketing knowledge is created through a continuous flow of activities interacting in a context of discovery, leading to conceptualization and implementation. A second context is that of verification – modeling and model solving. It is reasonable to assume that the more articulated or linked these four activities are, the more valid and reliable the knowledge generated will be.

3 The methodological association matrix

As already stated, the main purpose for building a methodological association matrix is to allow the researcher to critically evaluate if all the points in the research project are adequately coordinated, articulated, and linked. It is the moment to consider if he/she has properly formulated the theoretical model and research hypotheses, and duly justified the conceptual foundation. He/she must also analyze whether the survey questionnaire has adequate scales to test the research hypotheses or is missing any questions. Is the chosen metric suited for the data analysis technique? Will the expected result be useful in implementing actions concerning the phenomenon under study? And so forth.

The first element in the association matrix is the theoretical model. Puschel (2009) gives a good example of a theoretical model in his Master´s thesis, published in an academic journal (Puschel et al., 2010). The model depicted in figure 3 refers to empirical research done with bank clients who were users and non-users of mobile banking. Puschel designed two questionnaires to test a set of 13 research hypotheses drawn from the consumer behavior literature.

This study’s main purpose was to evaluate which factors influence users’ adoption of mobile banking and which factor could influence non-users’ future adoption. Figure 1 shows that diverse observable variables (rectangles) are associated with each construct (ellipses). The relationships established between the constructs correspond to the research hypotheses. As knowledge is always evolving, one can always use scales already validated in previous studies. The study´s basic conceptual model derives from the theory of innovation adoption (Rogers, 1983), whose antecedent constructs have been perfected by Moore and Benbasat (1991). The other underlying theories are perceived ease of use (Davis, 1989), subjective norms and perceived behavioral control (Taylor and Todd, 1995). The integration of all these constructs was based on Ajzen’s (1971, 1985) theories of reasoned action and planned behavior. Note that the study also considered the effect of attitudes, perceived behavioral control, and subjective norms on behavioral intention itself (Taylor and Todd, 1995).

The theoretical construction of the model proposes that seven first-order antecedent constructs, identified in the literature as benefits from using mobile banking, are positively related to the construct favorable attitudes toward this service (hypotheses H1 to H7). That is, higher levels of perceived benefits would be positively associated to more favorable attitudes toward mobile banking. A further hypothesis is that consumers’ favorable attitudes toward mobile banking would mediate (potentialize) the effect of the different benefits perceived in adopting this technological innovation in banking services. Therefore, non-users’ favorable attitudes potentialize the effect of the different perceived benefits in the probability of adopting mobile banking.

Two additional constructs are hypothesized to positively impact adopting this service: subjective norms due to friends’, colleagues’ and relatives’ influence (H11) and perceived behavioral control (H13). Thus, the theoretical model must be tested in search of evidence that this set of constructs positively impacts adoption of mobile banking. On the other hand, it is reasonable to expect that the influence of such factors on adoption is distinct for users and non-users.

After presenting Puschel’s (2009) example of a theoretical model based on a review of various studies in the literature, the next step is to explore the proposed tool named ‘methodological association matrix’. Figure 3 shows an example of this matrix, drawn from a doctoral dissertation focusing on a social marketing program to provide company workers with better nutrition (Mazzon, 1981). The Worker´S Food Program (Programa de Alimentação do Trabalhador - PAT) had been established in 1976, but three years later it still had low market penetration. Why did a program with so many benefits and evident socioeconomical impact for companies and other stakeholders (workers and their families, government, agribusiness, restaurants, among others) have such a low market penetration? Which elements should be investigated to manage this program from a social marketing perspective? How did companies and workers feel about this program? Which market segments were there? Which elements could be important in devising an adequate marketing strategy and different marketing actions targeting each one of these segments?

Whereas in the previous example the theoretical model was designed based on other existing models, the theoretical model used in this study corresponds to the traditional purchasing process model: identification of consumer needs and desires; pre-purchase activities or search for information on the program; consumer judgment and decision-making process; usage behavior; and post-purchase feelings represented by favorable or unfavorable attitudes toward the program.

Two research hypotheses were formulated regarding the first component of the theoretical model. The first one referred to the identification of latent factors related to the motivations for adoption of the program by companies. The second hypothesis proposed that these latent factors were distinct across the three types of food services available for workers. The first hypothesis was tested through the statistical techniques exploratory factor analysis and correspondence analysis using variables V6 to V12 in the questionnaire, measured along a Likert scale. Using factor and correspondence analysis with this ordinal scale was done because, at the time, there was a discussion around the possibility of assuming this scale as interval. As for the second hypothesis, the tests meant to show whether the variables and latent factors showed a significant difference for, at least, one of the three food services. In this case, two statistical tests were used - a parametric analysis of variance (F-test) and the non-parametric Kruskal-Wallis test. These tests were chosen for the same reason mentioned before. The main purpose of these tests was to identify the motivators associated with each type of food service, so that the results could help formulate and implement differentiated communication efforts, increasing their efficacy and the program’s market penetration.

Next, the theoretical model aimed to evaluate program information search by the companies and the decision variables for choosing the food program. Therefore, the third research hypothesis tested whether there was an association between the three food services and a set of segmentation variables: the type of industry (manufacturing, retail, or services); region of headquarters; company size; and segment, among others. Variables V2 to V5 and V23 to V26 were used to test this hypothesis through the chi-square test.

For hypothesis H4, the decision attributes were tested for differences regarding the food service using variables V2 and V48 to V61 and the F- and Kruskal-Wallis tests for comparison of means.

The aim of hypothesis H5 was to test for a determinant decision factor for choosing a food service using the same variables as H4, as well as factor and correspondence analysis (two dimensionality reduction techniques). Lastly, the fifth hypothesis tested which variables and factors were the most relevant in choosing the food service, based on the variables of the two previous hypotheses. In this case, discrete, multiple discriminant analysis was employed.

For which purpose could the results of different data analysis techniques be used? To segment the market according to the decision variables used to choose a food service. Also to identify selling arguments for each target segment, focusing on variables which were relevant for service judgment and choice. A further aim was to classify Brazilian companies which were non-users of the program according to their likelihood of selecting one of the three food services.

Hypothesis H7 was meant to identify food program managers’ psychographic profiles. Variables V95 to V106 were measured along a Likert scale and tested through exploratory factor analysis and correspondence analysis. The results led to building profiles that helped tailoring communication messages to engage this target audience.

The remaining hypotheses were based on a set of 29 statements (V65 to V94) related to attitudes toward the food program measured along a biphasic, Likert-type scale. Initially, agreement or disagreement (the direction) with each statement was assessed and next, the intensity (slightly, strongly, or totally). Albaum (1997) discusses the relative advantages of using this scale. The hypotheses were tested using diverse techniques, depending on their formulation and scale: mean comparison (F- and Kruskal-Wallis tests), factor and correspondence analysis, and cluster analysis. The combined results from these statistical tests aim to inform marketing strategies and actions targeting each market segment according to their attitude profiles, as well as the demographic variables of the companies and program managers.

4 Final considerations

A growing number of papers is being published in journals and conference proceedings in the applied social sciences, particularly marketing and business. New journals have appeared in diverse countries during the last years. The number of Master´s theses and doctoral dissertations showed expressive growth in Brazil in the last decade. Graduate programs have been implemented all over the country. Efforts have been made to continuously improve such programs in order to supply the country with better qualified Masters and Phds.

The aim of this paper is to add to such efforts, by showing that marketing knowledge creation may take different routes and encompass researchers with diverse profiles. Marketing philosophers work with conceptualization, marketing scientists work with modelling and model solving, and marketing managers use the results in action plans aimed to generate positive impacts on companies, consumers, and society.

Despite these efforts, it seems reasonable to admit that, especially under pressure or stress, failures may occur when conducting empirical research - surveys and experiments in particular. We believe it is useful to propose a tool that enables researchers to question several aspects before collecting the data. For example, question whether the research objectives will be in fact fulfilled, whether the research hypotheses are clear and well formulated, or the questions in the questionnaire allow testing these hypotheses. Are the scales adequate and the data analysis techniques compatible with the scales and hypothesis tests? Will the results obtained be useful to formulate relevant marketing strategies and actions?

As a final reflection we suggest including in the methodological association matrix a new column to the right of the research hypotheses column showing the references to theory that support the formulated hypotheses.

References

Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J. Beckman (Eds.), Action-control: From cognition to behavior (11-39). Heidelberg, Germany: Springer.

Ajzen, I. The Theory of Planned Behaviour. Organizational Behavior and Human Decision Processes, 50 (2), 179-211.

Albaum, G. (1997). The Likert Scale Revisited: An Alternate Version. Journal of the Market Research Society, 39 (2), 331-348.

Apostel, L. (1960). Towards the formal study of models in the non formal sciences. Synthese, 12 (2-3), 125-161.

Barberi, B. (1967). Il metodo statistico nello studio dei fenomeni observazionali. Torino, Editore Boringhieri.

Bunge, M. (1974). Teoria e realidade. São Paulo, Editora Perspectiva S.A.

Capes (2018). Catálogo de Teses e Dissertações. Retrieved August, 31, 2018, from http://catalogodeteses.capes.gov.br/catalogo-teses/#!/

Davis, F. D. (1989). Perceived usefulness, perceived easy of use, and user acceptance of information technology. MIS Quartely, 13 (3), 319-340.

Lilien, G. L. (1975). Model relativism: A situational approach to model building. Interfaces, 5 (May), 11-18.

Mazzon, J. A. (1978). Formulação de um modelo de avaliação e comparação de modelos em Marketing. Dissertação de mestrado, Universidade de São Paulo, São Paulo, SP, Brasil.

Mazzon, J. A. (1981). Análise do Programa de Alimentação do Trabalhador sob o conceito de Marketing Social. Tese de doutorado, Universidade de São Paulo, São Paulo, SP, Brasil.

Moore, G. C.; Benbasat, I. (1991). Development of an instrument to measure the perceptions of adopting an information technology innovation. Information Systems Research, 2 (3), 192-222.

Puschel, J. P. Mobile banking – proposição e teste de um modelo integrativo de adoção de inovações. Dissertação de mestrado, Universidade de São Paulo, São Paulo, SP, Brasil.

Puschel, J. P.; Mazzon, J. A.; Hernandez, J. M. C. (2010). Mobile banking: proposition of an integrated adoption intention framework. International Journal of Bank Marketing, 28 (5), 389-409.

Rogers, E. M. (1983). Diffusion of innovations (3rd ed). New York, The Free Press.

Taylor, S.; Todd, P. (1995). Decomposition and crossover effects in the theory of planned behavior: a study of consumer adoption intentions. International Journal of Research in Marketing, 12 (2), 137-156.

Enlace alternativo

https://periodicos.uninove.br/index.php?journal=remark&page=article&op=view&path%5B%5D=12424&path%5B%5D=5987 (pdf)