Building on CHASM: A Study of Using Counts for the Analysis of Static Models of Processes

Process modelling is gaining increasing acceptance by software engineers as a useful discipline to facilitate both process understanding and improvement activities. This position paper builds upon previous work reported at the 1997 ICSE workshop on process models and empirical studies of software engineering (Phalp and Counsell 1997). In the previous paper, we argued that simple counts could be used to support analysis of static process models. We also illustrated the idea with a coupling measure for Role Activity Diagrams, a graphical process modelling notation adapted from Petri Nets. At that time only limited empirical work had been carried out, based upon a single industrial study, where we found high levels of coupling in an inefficient process (a more thorough description may be found in (Phalp and Shepperd 1999)). We now summarise a more recent study, which uses a similar analysis of process coupling again based on simple counts. In the study, we compared ten software prototyping processes drawn from eight different organisations. We found that this approach does yield insights into process problems, which could potentially be missed by qualitative analysis alone. This is particularly so when analysing real world processes, which are frequently more complex than their text book counterparts. One notable finding was that despite differences in size and domain, role types across the organisations exhibited similar levels of coupling. Furthermore, where there were deviations in one particular role type, this led the authors to discover a relationship between project size and the coupling levels within that type of role. Given the simplicity of our approach and the complexity of many real world processes we argue that quantitative analysis of process models should be considered as a process analysis technique