Reasoning and methods

How tempting would it to be to describe the research process in the following linear and sequential way: a problem was identified that is of interest to the research community, specific research questions or hypotheses were formulated relying on the theoretical resources of the research community, appropriate research strategies based on either or both deductive or inductive logic were elaborated, qualitative or quantitative measures were chosen and put to work, data compilation and analysis followed, and, with pluck and luck, plausible (or verifiable) inferences and conclusions resulted; end of story (Van Maanen et al., 2007). Van Maanen et al. (2007, p. 1146) point out that matters of how the actual research process proceeded are rarely discussed (at least in print) since in reality:

the flow on research is lengthy and uneven, is seen most clearly in hindsight, and perhaps most important, is contextually idiosyncratic, often chaotic, and always personal; how we arrive to conclusions is difficult to penetrate when publication norms do not favour the presentation of results in the manner in which they evolved and when the personal history of how the research process unfolded over time may be revised or forgotten as the project moves toward its final printed version.

26

The ironic description of the research process illustrates the overly simplified and idealized version of the interplay between theory and data which serves as the general form of representation of the results in academic writing (Van Maanen et al., 2007). Luckily, inch by inch the cognitive limitations of researchers are being incorporated into the discussion regarding the process of doing research (Alvesson & Kärreman, 2007; Van Maanen et al., 2007) and reasoning (Mantere & Ketokivi, 2013).

The interplay of theory and data does not usually follow this prescribed, almost magical, sequence (Van Maanen et al., 2007); it is more a sense‐making venture that evolves over time (Bailyn, 1977;

Weick, 1989; Alvesson & Kärreman, 2007).

Regarding the abductive reasoning logic of the dissertation entity, the following notions on the interplay of theoretical and empirical observations are in order. Understanding on the variation in ecosystem architecture was sought against the theoretical backgrounds of two‐sided markets, value co‐creation, and business models. For example, before turning to industry architecture literature, explanation about the phenomenon was sought from literature on industrial organization and two‐

sided markets (see publications I, II, V and VI). Further, in order to explain how and upon which template a firm organizes its transactions with suppliers, complementors and customers, explanation was sought from business model literature (see publications II and IV) before turning to literature on meta‐organization design. The theoretical explanations from two‐sided markets, value co‐creation, and business models literature were not seen as comprehensive enough to address the phenomenon of interest and thus they were excluded.

The research design of the overall dissertation and the individual publications is depicted in Figure 2, illustrating the different reasoning logics and related methodologies. In addition to this introductory part, the dissertation comprises six separate publications, each focusing on different aspects concerning the subject under scrutiny.

27 Figure 2: Research design of the dissertation

The essence of an argument is to proceed from premises to conclusions in a credible manner and to defend the claims made in these conclusions (Toulmin, 2003). In order to bridge premises with conclusions, scholars as scientist use various reasoning principles (Mantere & Ketokivi, 2013), namely deduction, induction and abduction. Mantere and Ketokivi (2013) explain the difference between deduction, induction and abduction in the following way. For deductive reasoning, the rule and the explanation are the premises and based on these, the observation is derived. Thus a conclusion about the particular is drawn based on the general. For inductive reasoning, the observation and the explanation are the premises and based on these, the rule is derived. Thus conclusion is drawn from particular to general. For abductive reasoning, the rule and the observation are the premises and based on these, the explanation is inferred if it accounts for the observation in light of the rule.

As depicted in the figure above, the overall research design can be characterized as interplay; the theoretical framework evolved simultaneously and interactively with empirical research (Dubois &

Gadde, 2002; Van Maanen et al., 2007; Dubois & Gibbert, 2010). As labelling the research deductive, inductive, or abductive does not by itself tell how the research process proceeded nor justify the methodological choices (Eriksson & Kovalainen, 2008), Table 1 further summarizes more precisely the research design adopted in complementary publications of the dissertation in terms of reasoning logics, methods and analysis, and the data used in the publications. The characteristics of the research phenomenon, the amount of prior research about it, and the nature of the research goals and objectives have guided the methodological and method choices of the dissertation along with the philosophical assumptions adopted. The individual publications have three different reasoning logics: abduction (publications I, II and VI), induction (publications III and IV) and

28

deduction (publication V). Two different methodological approaches were used in the publications:

qualitative, exploratory case study (publications III and IV) and quantitative method (publication V).

Table 1: Research design of the individual publications PUBLICATION REASONING

Content analysis Interview data gathered from 14 globally operating companies in 20

Content analysis Interview data gathered from 14 globally operating companies in 20

Univariate General Linear Model Simulated data with agent based modelling and Monte‐Carlo Analysis

development Existing theoretical and empirical research on two and multi‐sided platform based markets, business ecosystems and industry architecture

2.3.1 Abduction

The overall research design of the dissertation and the research design of publications I and II is abductive. Abductive reasoning is one of the primary reasoning tools we use not only in scholarly inquiry but also in mundane situations. Abduction refers to seeking explanations (Peirce, 1878) and to the selection of the best explanation from among competing explanations, also know has inference to the best explanation (Harman, 1965; Lipton, 2004). Thus abduction can be characterized as beginning with an unmet expectation and working backward to invent a plausible world or a theory that would make the surprise meaningful (Peirce, 1878; Van Maanen et al., 2007).

Regarding the overall abductive research design of the dissertation, abduction was chosen as the dominant reasoning logic given the inherent problems of pure deductive or inductive reasoning.

Deductive reasoning can be said to “sidestep the question of alternative explanations” (Ketokivi &

Mantere, 2010, p. 318), it “does not provide selection criteria for choosing between alternative explanations” (ibid, p. 318) and generation of new theory from the data with deductive reasoning is difficult. With inductive reasoning, the research will always encounter an “unavoidable logical gap

29

between empirical data and theoretical generalizations” (ibid, p. 316). Further, it has been noted that deductive and inductive reasoning actually involve abduction at some point of the process (Marcio, 2001).

The abductive mode of inquiry of the overall study comprised a continuous interplay between an empirical world and a model world in order to generate theory, also labelled as systematic combining (Dubois & Gadde, 2002). During the research process, the research issues and the analytical frameworks were reoriented whenever they were confronted with the empirical world (Dubois & Gadde, 2002). The frameworks which were not seen to address comprehensively the theoretical puzzles posed by the examined phenomenon were excluded during the process.

Regarding publications I and II, the mode of inquiry was also abductive, as the purpose of these publications was to locate the research problem, create a pre‐understanding of the phenomenon under scrutiny and to establish relevant directions for future studies and theory development. In addition, in publications I and II, an illustrative single case was chosen in order to present a unique illustration of the phenomenon on which understanding was sought (Stake, 1995). Publication VI also relied on abductive reasoning with the purpose to create a full‐fledged pre‐understanding of the phenomenon in question. To conclude, the abduction of the overall study proceeded from the pre‐understanding created in publications I and II to the active interplay between the data and the various theoretical frameworks (especially in publications III, IV and V, with the assistance of deductive and inductive reasoning) to the revised pre‐understanding created in publication VI and finally to the construction of the best explanation of the phenomenon under scrutiny.

2.3.2 Induction

Inductive reasoning in its purest form (i.e., grounded theory) begins at the intersection of a theorist’s general wonderment and raw data (Shepherd & Sutcliffe, 2011). Induction based on qualitative research focuses and seeks to understand the complexity of business‐related phenomena in context‐specific settings; the aim is to produce new knowledge about how things work in real‐life business contexts, why they work in a specific way, and how we can make sense of them in a way that they might be changed (Eriksson & Kovalainen, 2008). Qualitative research seeks instead illumination, understanding, and extrapolation to similar situations (Golafshani, 2003; Hoepfl, 1997). Typically the use of inductive qualitative research is based on the following arguments; the theory is nascent or undeveloped or researchers are employing an interpretive perspective (Graebner et al., 2012).

As it is usually the case that certain constructs or theoretical frames are drawn from existing literature before gathering data and then further built into a data collection effort, reasoning might not be purely inductive. Also, during data analysis, some definitions or frameworks can be drawn from the existing literature even though no specific constructs or theories were taken as a starting point. This can be the case when seeking for example to complement and extend previous theoretical work.

Induction and bottom‐up theorizing based on qualitative research was chosen as the research approach for publications III and IV. Due to the focus of the study and the undeveloped stage of the existing research field on the topic, an exploratory case study approach was utilized. An exploratory case study is aimed at producing new theoretical ideas, propositions or hypotheses or determining the feasibility of the desired research procedures (Eriksson & Kovalainen, 2008). The first and foremost purpose of the exploratory case studies was to produce new theoretical ideas and propositions.

30

Interviews were utilized as the data collection method. The interview data was collected from 20 informants from 14 globally operating companies with semi‐structured interviews (Table 2). In the sampling, expert opinions were utilized and the informants were selected so that they would represent the views of different relevant ecosystem players. The interview guide had five sections and during the interviews, the informants were asked to describe for example their collaboration with the ecosystem hub using an open‐ended format, prompting the informants with questions related to resource complementarities and interdependencies. Further, the informants were asked to describe their cooperation with competitors and other complementors in order to identify co‐

operation in different levels in the ecosystem architecture with related possibilities, benefits, challenges and risks.

Table 2: Gathered interview data for publications III and IV

FIRM INFORMANT informants participated in the email interview. The number of informants per company varied as during the first contact, the initially selected company representative was asked to identify other company representatives who could provide information related to the topic. In some smaller companies, the contacted company representative was the only one to represent both a technological and business perspective, and in some larger companies the contacted company representative identified one or two other suitable informants. All the face‐to‐face and phone interviews were recorded and transcribed. The duration of the interviews varied from 35 minutes to 76 minutes. In total, 170 pages of transcribed material were analysed. In general, qualitative data can be characterized as open‐ended, concrete, vivid, rich and nuanced (Weick, 2007; Graebner et al., 2012). Furthermore, the beauty of qualitative data lies in its ingrediential nature; there is no cookbook or recipe for qualitative research and analysis (Coffey & Atkinson, 1996; Maxwell, 2005;

Pratt, 2009; Bansal & Corley, 2011; Graebner et al., 2012)

The analysis began after all the data had been collected in order to preserve the integrity of replication logic (Eisenhardt, 1989; Yin, 1994; Ozcan & Eisenhardt, 2009). The analysis began by looking for similar constructs and themes from the transcribed interview data. From the emerging constructs and themes, tentative relationships between constructs were formed.

31

In order to address the quality of inductive reasoning based on qualitative inquiry, two dimensions of quality can be taken into account: the quality of the inquiry process (trustworthiness) and the quality of the inquiry product⁴. This section focuses on the quality of the inquiry process based on the four criteria proposed by Lincoln and Guba (1985): credibility, transferability, dependability and confirmability⁵. In order to ensure that the analysis conducted in this study would meet these criteria, several actions were taken. In order to meet the credibility criterion, the results being acceptable representations of the data, the interpretations and conclusions made from the data and analysis were discussed among the researchers and with the focal representatives of the informant companies. This ensures that the research is credible at least among the subjects of the study.

Transferability refers to the way of reporting the findings which allows the readers to determine to what extent the results are applicable in their situation. This study reports the results in a way which does not limit the applicability of the results only to the empirical context of this study (ICT industry). The dependability criterion has been addressed with careful recording and documentation of the data as ensuring the stability of findings was not possible with replication.

Further, the careful recording and documentation ensure that the induction based on qualitative research methods meets the confirmability criterion.

2.3.3 Deduction

Deductive reasoning based on quantitative study was utilized in publication V. The publication relied on previous theoretical literature to build testable hypotheses, generate data with a simulation model executed with agent‐based modelling and to subject the data to the Univariate Linear Modelling procedure to examine the hypothesized effects. Given the complexity and dynamic nature of the phenomenon under scrutiny and the theoretical background being rooted on the literature on complexity and complex systems, the selection of simulation is justified. Further, simulation takes into account multiple actors and the independent but interdependent decisions which they make.

The model simulates how actors, namely customers and suppliers, adopt and connect to different platforms. This complex system was analysed by explaining the dynamics, the controlling variables behind it and how it affects the competitiveness of each platform. The simulation model was executed with agent‐based modelling and Monte Carlo analysis leading to a sample size of 1.7 million observations, with the purpose of exploring entry timing strategies of the follower platform and competition between first‐mover and follower in platform‐based markets. The dependent variable was the follower’s relative financial performance.

4 This dimension can be judged upon the following criteria (Lincoln & Guba, 1990): resonance, rhetoric, empowerment, and applicability.

5 Alternative strategies to judge the quality of qualitative research are suggested by Yin (1994), Gibbert et al. (2008) and Creswell (2009).

32