An examination of a ripple effect in industrial web system change

Abstract

In this paper, we explore the characteristics of change categories in three evolving web systems. We examine those changes from the perspectives of three maintenance categories (adaptive, corrective and perfective) and the influence of a `ripple' effect as a result of those changes. All three systems were developed by a software development company based in London. Results showed that the ripple effect was a prominent feature of many changes made by developers to the systems; however, while the adaptive category most frequently caused a ripple effect, in terms of effort hours the perfective category was most effort-intensive. We provide explanations for why this might be the case supported with the specific changes made by the developers and identification of architectural `forward' and `backward' ripple effects. Finally, we explore whether an 80:20 law was evident from the effort data (both for ripple and non-ripple based data).