A metric for quantifying the ripple effects among requirements

Abstract

During software maintenance, it is often costlier to identify and understand the artifacts that need to be changed, rather than to actually apply the change. In addition to identifying the artifacts related to the change per se, one needs also to identify the artifacts that are changed due to ripple effects. In this paper, we focus on ripple effects and propose a metric for assessing the probability of one requirement to be affected by a change in another requirement (i.e., requirements ripple effect). We focus on the requirements level, since most maintenance tickets (which stem from the customer) are captured in natural language and therefore are more naturally mapped to requirements, rather than source code. The proposed metric—the requirements ripple effect measure (R2EM)—is calculated by considering the conceptual overlap between the involved requirements (through their past co-change), the parts of the code in which they are implemented (i.e., their overlapping implementations), and the underlying dependencies of the source code (i.e., ripple effects between classes). We note that despite the involvement of source code artifacts in the calculation of R2EM, this metric is considered as a requirements’ level one, since the unit of analysis is pairs of software requirements. To validate the proposed metric, we conducted an industrial case study, on two enterprise applications of an SME. The study design involved both quantitative and qualitative data, and input was given by 9 practitioners. The results suggested that R2EM is able to identify ripple effects between requirements at a satisfactory level, and those effects are mostly caused by overlapping implementations and source code ripple effects of these implementations.