In this chapter we shall demonstrate the viability of the proposed method engineering principles by analyzing two cases of incremental ME. Our focus on real-world method development efforts means that we will face two major differences in metamodels: hereafter the metamodels developed are situation-bound, and method applicability varies as its use situations change.

So far we have modeled ISD methods as they are described in the method literature: they are “universal”, standard and largely fixed. In Chapter 4 each method was specified using a single metamodel and no situational method modifications were made. Recently, method developers have adopted metamodeling for describing meta-data models, e.g. Booch et al. (1997) present metamodels for their Unified Modeling Language, and Henderson-Sellers and Bulthuis (1996b) for their Open Modeling Language. These metamodels, however, neither suggest modifications of methods nor provide different method versions for different situational needs. Although some situational needs are identified (e.g. Booch and Rumbaugh 1995), versions that can meet these situations are not specified. In incremental ME, metamodels are made based on situational needs. At the same time we can demonstrate that the metamodeling constructs are relevant for modeling situation-bound methods, not only applicable for modeling text-book methods, as we used them in Chapter 4.

Based on the re-evaluation of method use (cf. Section 2.5) we shall focus on supporting the evolution of methods. Two cases of local method development are analyzed longitudinally and the methods constructed are evaluated a posteriori using the principles of incremental method engineering. These principles seek to externalize experiences of the methods’ use and channel them back into method improvements. This allows us to address our second research question on how to refine methods through modeling experiences. Possible method refinements resulting from a posteriori analysis demonstrate that the a priori method was not as applicable as intended. If some refinements occur, these justify our conjectures that local methods are evolutionary and need to be maintained. Alternatively, if no method refinements are needed, then a posteriori analysis can be considered unnecessary, or the evaluation mechanisms were inapplicable to improve methods.

The chapter is structured as follows. First, we describe the action research method followed. Second, two cases of local method development are discussed using the steps of incremental ME. Finally, the cases are analyzed by soliciting lessons about local method development, about method engineering principles, and about the incremental approach.