Multi-response optimization based on desirability function and Taguchi method in agile software development

Abstract

Software industries are continuously striving to achieve software quality through delivery of defect free software within time and within estimated budget. Most of the quality improvement strategies in software development focusses on optimizing a single response variable such as the defect count. But in reality, many responses contribute to the quality of software developed which requires to be studied simultaneously. This paper attempts to study the number of defects in software and the effort required in correcting these defects, simultaneously using desirability function and Taguchi method. Effort is defined as the time span from when the defect was logged into the system to the time when the defect was corrected and removed from the system and is measured in hours. The work has been executed on the issue tracking system of a large telecommunication organization. The organization manufactured consumer electronic devices having embedded software developed using agile development method. The input factors studied are Change Request Priority, software development phase and severity, which were selected from a total of 24 parameters using analytic hierarchy process. Significant factors and their optimal levels which optimized the defects captured and the effort taken to eliminate them, were obtained. Confirmation tests were performed using the optimal settings of the input factors on the data from the subsequent sprint which showed a 20% improvement in the overall desirability value.