5.4 Concluding Remarks

The survey of state-of-the-art approaches to enterprise modeling has revealed that the variety is large, in the sense that the aspects of enterprise modeling receiving most attention in the surveyed literature differ substantially. From this point of view, revisiting the research questions may be a natural closing of the survey.

Main research question: All of the approaches surveyed intend to support the purpose stated in the main research question. This was ensured by deliberate selection of frameworks (it was one of the criteria outlined in section 5.1.1). The frameworks have provided insights on how enterprise modeling should be performed (but very little on how enterprise modeling is performed, as they do not provide extensive accounts of enterprise modeling practice).

RQ1: Most of the approaches had several purposes in addition to human sense-making and communication (except SSM and SAMPO). Actual use of enterprise models cannot be assessed from the literature surveyed.

RQ2: The activities perceived to be most important differ substantially from approach to approach. While some do not discuss activities at all (Enterprise, TOVE), others prescribe a detailed list (OMT, OOram, SAMPO). Some prescribe a sequence that may be iterated (OMT, Metis), others reject a strict sequence (F3, OOram). Many approaches prescribe their activities using the language of the modelers ("identify object classes" in OMT). This is consistent with focusing on representation, not on sense-making.

RQ3: The model properties perceived to be most important varies just as much as the activities. Some prescribe formal models (Enterprise, TOVE), while others rely on highly informal ones (SSM, some OOram, some SAMPO). Some have predefined ontologies (Enterprise, TOVE), while others have none (SSM). Most of the approaches rely on graphical modeling languages.

RQ4: One relationship seems to be that the modeling approaches that focus on work dominated by replication risk also focus most on the enterprise models and less on the modeling process, and expose an attitude to enterprise modeling that is more consistent with an objectivistic worldview than with a constructivistic one. This makes perfect sense from a theoretical point of view: If reality is widely agreed upon, the need for seeing modeling as social construction of reality is reduced as compared to a situation where reality is up for grabs.

Another relationship is that the approaches that intend to use enterprise models most extensively (in the sense having the widest variety of purposes for enterprise modeling) are the ones that model work dominated by replication risk. This also makes sense, as routine work more easily lend itself to automation.

RO2 from section 1.2.2 stated that the intention of this research project is to develop yet another framework for enterprise modeling. A timely question is if this really is necessary and why one of the existing could not be adapted instead? To anticipate the rest of the thesis, the observations from the main empirical study in chapter 6 provided some lessons learned (summarized in chapter 7) concerning the challenges of conducting modeling when the sense-making part of the modeling effort is dominant. These challenges are not handled appropriately by any single framework for enterprise modeling today (with the possible exception of SSM). A direct comparison of the proposed framework and the most favorable ones from the state-of-the-art survey (F3, SSM, and OOram) is provided in chapter 10.