In this chapter we defined method engineering and placed our research in the context of ME research. We analyzed the current understanding of ME in terms of its process, the type of method knowledge “engineered”, and the criteria used in method construction. First, related to the ME process, the analysis shows the dominating a priori approach. Most principles are targeted toward method selection and construction, and little attention is paid to analyzing whether the constructed method is applicable in the task for which it has been promoted, or could the method be improved. Hence, information about methods and ISD contingencies is expected to be known completely in advance. Moreover, learning from method use and the evolution of methods are ignored.

Second, we surveyed the ME criteria that have been proposed to construct methods. These explain how the situational applicability of methods can be improved. Of these approaches, we discussed those based on contingencies, problems, and stakeholders’ values. Each of these approaches is limited in its ability to guide method construction in detail. They are too general to provide guidelines for constructing techniques or their parts for situational needs; they rely on existing problems and contingencies; and they do not support the creation of new knowledge originating from an organization’s own experiences. Finally and maybe most importantly, none of these approaches seems to be used systematically in an organization’s local method development efforts. This is especially important since empirical studies of local method development reveal that few organizations apply systematic customization processes (Smolander et al. 1990, Cronholm and Goldkuhl 1994), but rather follow ad-hoc practices (Hardy et al. 1996, Hughes and Reviron 1996).

Third, based on the shell model (cf. Section 2.2) we analyzed which types of method knowledge are identified and subject to ME. We focused on metamodeling languages targeted to representing the conceptual structures behind modeling techniques. These are also most widely studied in the ME literature. Since conceptual models describing methods should be based on some knowledge representation scheme, we furthermore described a set of metamodeling languages. We illustrated, through a small example, various grammatical and notational constructs applied in metamodeling languages. This example showed some differences and limitations in the metamodeling constructs, and it serves as the background for a more detailed analysis of their metamodeling support in the next chapter.