In reflecting on the research questions addressed, we observed several interesting research topics. These would allow us to further evaluate and expand the findings of this thesis.

For any modeling language, functionality and usability form central issues: ME languages are no exception. Future research on ME languages should concentrate on these aspects, either by extending existing ME languages or by creating new ones. With respect to the functionality aspect, the sufficiency of the metamodeling constructs could be analyzed by modeling more methods. The selected sample should include other types of methods than those modeled here. The sufficiency of the proposed metamodeling constructs can also be examined by modeling organizations’ in-house methods, rather than methods described in the literature. These examinations can confirm (or raise doubts about) the relevancy of the proposed metamodeling constructs, and most likely find new constructs.

The proposed metamodeling constructs can be used to evaluate other metamodeling approaches. They can be used as a set of requirements to develop new metamodeling languages, or extend existing ones. Research on metamodeling should be extended to cover other types of method knowledge, in addition to the conceptual structures behind modeling techniques. Candidate types of method knowledge to be modeled include processes, participation, and decision making.

When analyzing the functionality of a metamodeling language, its usability should not be forgotten. This suggests investigating the use of ME languages using different research methods. In fact, to proceed in ME research we need empirical studies about the use of metamodeling more than proposals of new metamodeling languages. Surveys and field studies must be made to analyze what metamodeling languages are used in practice; laboratory studies are needed to investigate user preferences for different visual representational paradigms (e.g. Kelly and Rossi 1997); and case studies are needed to assess the usability of metamodeling languages in a ME project.

Empirical research is also relevant to the study of the ME process. Because ME is a relatively new research field, complementary research efforts and the use of various research methods are needed to improve the quality of research conclusions (see Tolvanen et al. 1996). As pointed out in this thesis, more case studies and action research are needed to analyze local method development in detail. These research methods should be applied to examine what factors contributed to success or failure in local method development, how frequently and to what extent methods are changed, and how methods evolve. These questions presuppose longitudinal research efforts, as well as close interaction between method use and method development situations. In addition to longitudinal studies, larger scale ME efforts, in terms of the number of stakeholders and method size, should be inspected. Studies should also address methods other than business modeling, apply different metamodeling languages, and implement method-tool companionship with different metaCASE tools.

Within empirical research, other research methods, such as field studies and surveys, must be used. Although several surveys of method use and to some extent also of method development have been performed, there is still a need for new ones. One reason is that existing studies have obtained different results, and several key questions of ME remain unanswered. Surveys should analyze how common in-house methods are, and whether stakeholders are satisfied with local methods. Field studies allow the examination of the ME process in more detail. They should examine the circumstances under which local methods are developed, whether the ME process consists of “radical” or incremental changes, and how ME projects are organized and managed.

Finally, the incremental ME principles should be taken into account while developing metamodeling languages and metaCASE tools. In addition to extending metamodeling languages with the proposed constructs, they should also be applied in metaCASE technology. MetaCASE tools should offer functionality to modify and version metamodels, to update models when a method already in use is changed, to support the collection and structuring of experiences about the use of the method, and to automate the mechanisms of method evaluation. In particular, metrics for type-instance matching should be implemented into metaCASE tools. Design rationale models should also be taken into use for recording and explaining metamodeling decisions. Tool support for these functionalities would allow the proposed principles of incremental ME to be used to full advantage.