This chapter has focused on complementing existing method engineering principles by introducing an incremental approach. This approach is motivated by the limitations of a priori ME approaches. In short, a priori ME is not interested in method use, and it assumes that the constructed method is understood and applied as the method engineer intended. In contrast, we believe that method knowledge and method construction criteria can not be known completely beforehand. Moreover, we claim that an ISD environment is not stable, because method use situations change and method users learn about their methods. These method use characteristics were also emphasized in our re-evaluation of method use (Section 2.5). As a result, at some point of time a method becomes less applicable for the tasks for which it was promoted. Certainly it is possible that a method can be found to be fully applicable during the ISD project. Even in these cases it is of key importance to learn about method use. The learning aspect is also a key difference between a priori and a posteriori approaches to ME. Instead of expecting that a method engineer is responsible for all improvements to a method, the incremental approach emphasizes the role of method users and their experiences. In other words, method development should be based on stakeholders’ experience and situational needs, in contrast to selecting methods solely by using ‘universal’ ME criteria. The focus on experiences is also relevant because learning through experiences has been identified as a main way of learning about methods (Hughes and Reviron 1996).

The incremental approach is clarified through method engineering scenarios. The scenarios illustrate steps of ME in which modifications can occur. The scenarios are used to explain incremental ME principles: we are interested in experiences which arise from method use, which can be made explicit, and which can contribute to method refinements. Explicit means that method improvements are not tacit, nor individual knowledge, but can be discussed and shared in an organization. This is important because learned methods often become tacit and “invisible” (Wastell 1996) and an individual developer’s productivity (Davis et al. 1991) can be reduced by methods (Fitzgerald 1996). The aims of method refinements mean that we evaluate methods primarily for improving them in a current use situation.

To relate our incremental approach to other studies, we reviewed the approaches proposed for situational method evaluation and validation. This analysis pointed out that most of the evaluation approaches do not follow any systematic evaluation procedure for data collection or analysis. They are carried out mostly by method developers, and they do not aim to systematize the method improvement process. Moreover, unlike ME they aspire to a general situation-independent proof (or disproof). This proof has been found difficult to obtain (Fitzgerald 1991) as it necessitates that evaluations could be replicated, the variety and complexity of ISD environments reduced, and data collection limited to factors relevant to method use. Our approach is different. We aim to evaluate methods in situations in which they are applied and use an organization’s own experiences as a source for method improvement. In this sense method modifications are subjective, but generalizations can be found by iterating in cycles of incremental ME.

The incremental approach is described through the mechanisms for collecting and analyzing experiences for the purpose of method improvements. The mechanisms deal with differences between intended and actual use of a method, the modeling power of modeling techniques, and a method’s support for problem solving. In each case the experience is collected and analyzed differently and can lead to modifications of a method or a tool. For each mechanism, principles for collecting and analyzing experiences are described and alternatives for possible method refinement are explained. First, the approach collects experiences and analyzes the applicability of modeling techniques through the use of types and constraints (in a metamodel) for representing an IS (in models). Second, it focuses on mechanisms that emphasize the capability of a modeling technique to abstract relevant aspects of the IS and maintain the consistency of these models. Third, the suggested mechanisms evaluate the support of modeling techniques in problem solving. This evaluation deals with the capability to provide alternative solutions through form conversions and to support review and validation of models.