1.3    Research Approach

The research questions posed above express a need for learning more about the dominant concepts of enterprise modeling as a research domain. Hence, a qualitative research approach based on observations, document studies and interviews taken from real world modeling projects is argued for in order to meet the research objectives. However, a few alternative research approaches will also be discussed.

1.3.1     Four alternative approaches

From literature and discussions on various traditions and approaches to "good" research (Denzin and Lincoln, 1994; Johannessen, 1992; Miles and Huberman, 1994; Strauss and Corbin, 1990; Walsham, 1995), four potential research strategies have been considered for the research project: Below, a brief discussion of how each strategy could have been applied is provided. The various approaches can to some degree be combined in the same research project. However, the fundamental views on human inquiry and science underlying them differ in many respects and extensive combination is not common. Figure 1.1 provides an illustration of the four approaches adapted to the needs of this project.
Figure 1.1: Four alternative research approaches

None of the research strategies are considered to be unconditionally superior or inferior to the others -- they all have qualities that make them preferable for some purposes and research problems.

Alternative A: Logical, theoretical research

By a logical theoretical research approach is meant formal deduction of logical consequences from a set of initial assumptions (axioms). If the axioms are true and the rules are logically sound, the consequences are true as well.

This mode of research may be appropriate for formal sciences as exemplified by mathematics and parts of computer science. But the concern in this project is practical enterprise modeling and the study of real world projects. A logical theoretical approach would not draw upon the benefits of empirical work. Also, according to Hirschheim et al. (1995:195, italics added),

"data modeling is first and foremost a social and organizational activity and very little, if anything (except consulting folklore) is known how data modeling is exercised in practice and what its impacts are on organizations, their information systems management, and business operations."
Enterprise modeling and data modeling are closely related activities (as will be discussed in chapter 3), and although the quite pessimistic assessment cited above must be taken with a pinch of salt, lack of knowledge about actual practice is assumed to be a valid claim within enterprise modeling as well.

Alternative B: Quantitative, experimental research

A quantitative, experimental approach to doing research is within the classical scientific paradigm of natural, "hard" sciences like physics. The scientific method implies postulating hypotheses, doing quantitative experiments, and then either sustain or reject the hypotheses based on statistical analysis of the measured data (verification or falsification of hypotheses). The scientific method may be claimed to be the "best" research approach in relatively well known areas of research and when natural laws can be assumed to exist (in the sense that phenomena are repeatable and to some degree controllable). Even if there are indefinitely many theories explaining a given set of data, experiments may be repeated and theories can be verified (or rather, the confidence in theories may increase).

To be able to propose fruitful hypotheses, one must have a well developed understanding of the research area. In addition, in order to gain statistically reliable results, the number of samples must be large (for survey studies, in the range 40 and up, according to Galtung, 1967). Both these requirements suggest looking for other research instruments: A well developed understanding of enterprise modeling practice is not widely available (at least not according to Hirschheim et al., loc. cit.), and the access to directly comparable projects is limited. There have also been some critique of adoption of the scientific method as an approach to research on social systems, at least unmodified. See e.g. (Guba and Lincoln, 1994) for a discussion.

One perspective on quantitative research is as counting. Correlation between variables are estimated using statistical devices. However, in order for counting to be meaningful, one must also know that the variables counted are meaningful in the given setting. Hence, quantitative research requires well developed understanding of a domain in advance in order to judge if variables are meaningful.

Alternative C: Qualitative, observational research

Qualitative, observational studies refer to traditions that base their research upon qualitative data (as opposed to quantitative research) and do not actively and purposely manipulate the phenomenon under investigation. Grounded theory studies (Strauss and Corbin, 1994; Strauss and Corbin, 1990) and ethnographic methods (Atkinson and Hammersley, 1994) are examples of this mode of research.

Through close contact with the research field in question for a prolonged period of time, the researcher develops a profound understanding and (as claimed for grounded theory studies) becomes able to formulate a conceptually rich theory explaining the phenomenon under investigation. Contact with the field of research may be based on interviews, observations, or analysis of documents and other artifacts. In addition, literature studies are performed to the extent required to develop sensitivity in observation and interpretation.

A qualitative research approach can be used to develop the understanding required for evaluating if a variable is relevant or not to a given problem situation. Compared to the perspective on quantitative research as counting, qualitative research can be seen as proposing which variables to count

Alternative D: Participatory action research

Participatory action research (Reason, 1994; van Meel, 1993:13ff) refers to a set of approaches to research on social systems in which the researcher actively engage in the process under investigation (the actors of the social system being studied can be considered as co-researchers). The work of van Meel (ibid.) exemplifies this approach to research: First, an initial case study is performed for identification of problems, followed by theory development and implementation of a prototype information system. Finally, the prototype is employed in another full-scale project where the researcher participates and reflects upon the use of the prototype with the actors that are studied.

This approach to validation of results is interesting, but somewhat questioned by more traditional scientists. Walsham (1995:77) points out the problems of being perceived to have a personal stake in the researched project, and reporting on one's own role within the project as particularly challenging problems of participatory action research.

A possible adaptation of an action research strategy would be to investigate a few initial enterprise modeling projects, propose a method for enterprise modeling, and apply this in another full-scale project. The principal problems of this approach would be to trace (hopefully) successful outcomes back to the use of the method and to know what would be the situation without the method (as there is no control project).
From a more pragmatic point of view, in order to follow a participatory action research strategy one needs access to real world projects being willing to try out the proposed framework for enterprise modeling in practice. This was not considered practically attainable with the resources and time frame of the current research project.

1.3.2    Outline of a research approach

The research strategy referred to as qualitative and observational in section covers a diversity of approaches that resemble each other to a more or less degree. Without getting trapped in the often heated debates on what characterizes right and wrong approaches to good research, a few principles that worked as guidelines in the research project are discussed.

Narrowing the chosen research approach

Before outlining the research principles, the research approach is narrowed a bit further. Strauss and Corbin (1990:21ff) divide qualitative approaches into three main categories: Non-interpretive, interpretive and theory building.

Non-interpretive studies focus on describing the life world that is investigated. Observations are not necessarily analyzed and made sense of by the researcher, but instead left for the reader to interpret. Miles and Huberman (1994:8) associate ethnographic methods with this leaning towards descriptions.

Interpretive studies acknowledge the importance of the analysis performed by the researcher to the meaning attributed to observations. The researcher sets out create an account of the empirical observations, consisting of descriptive as well as analytical passages. The intention is to provide the reader with a sense of the "real world".

The grounded theory approach, being the prime example of a theory building approach, is a relatively well defined and comprehensive research method that seeks to develop conceptually rich theories grounded in observations from empirical studies. The approach (as presented by Strauss and Corbin, 1990) is based on a number of rather rigorous procedures and techniques, and adhering strictly to these is claimed to provide valid theories (ibid.:27).

The research approach followed here is in accordance with the interpretive scheme outlined above 51; observations are analyzed to make sense of enterprise modeling practice; a number of principles are proposed as lessons learned from the analysis of the observations (meeting RO1), and finally, a framework for enterprise modeling is developed (meeting RO2). The choice of research approach is also consistent with Walsham's (1995:74) discussion of interpretive case studies as the preferred strategy to answering "how?"-questions like the main research question in section 1.2.2.

Research guidelines

The following guidelines apply to this specific research project:

Loose research design: The researcher is assumed to approach the empirical studies with few expectations about what to find. The research questions are allowed to emerge from observations made in the enterprise modeling projects. Miles and Huberman (1994:16) refer to this as a loose research design.

The researcher as an observer: The role of the researcher is to observe and not to purposely manipulate the projects through active participation. Although project participants must be made aware of the presence and intentions of the researcher, substantial contributions to project work is not made.

The researcher's sensitivity: The researcher's sensitivity towards observed events and situations, and the interpretation of these, will be developed by literature studies and analyses of observations. Hence, frequent alternations between literature studies and analysis of observations are considered preferable.

A note on the decision to have a loose research design: There is ample supply of warnings against following this guideline, e.g., that studies tend to take extensive amounts of time and resources, data are sampled too broadly and coarsely, and cross-case comparability is jeopardized (Miles and Huberman, 1994:17). On the other hand, a loose design may be the most pragmatically feasible when studying phenomena that are not under the control of the researcher (e.g., when the availability of projects to study is highly uncertain and unpredictable, as was the case in this research project).

The purpose of adhering to the above guidelines was to reach a practical understanding of enterprise modeling practice, enabling the provision of convincing arguments in favor of features of the proposed framework. Particularities of the chosen research process are discussed in more detail in chapter 10 when evaluating both the outcome and the conduct of the research project.

1.3.3    Outline of the accomplished research process

The main activities conducted as a part of the concrete research project are illustrated in figure 1.2. The figure is idealized in the sense that iterations and alternation between activities are not accounted for. Briefly, the activities include the following:
Figure 1.2 Outline of main activities in this concrete research project
Selection of projects has been highly pragmatic due to several reasons: The source of projects practically available was limited, and involvement in a project is also time-consuming. The main common trait of the projects is an intention of using enterprise modeling in some way in order to understand aspects of the enterprise better.

The most influential empirical study is the one referred to as the main study. At that point in time, the research focus was on modeling as a way to support human sense-making and communication. The first three projects were not focused to the same degree, as data were sampled in a more broad and coarse way.