| dc.description.abstract | This paper addresses the need for a means of comparing the performance of two different software development teams. A structured literature review was performed to identify the software development metrics that exist today, which was later supplemented with the results of an expert inquiry aimed at identifying software development metrics that the review had missed. This yielded a total of 191 distinct software development metrics, identified in 44 included papers, as well as 6 additional software development metrics, identified by the experts. These results were systematically mapped in a graphing database, and analysed in focus groups with the experts in order to extrapolate tacit knowledge about what makes these metrics strong or weak. The extrapolated knowledge was encapsulated in a new model for metric strength, which was subsequently used to assess the strength of the 197 identified software development metrics. This model states that a metric should (a) be simple to explain and simple to measure, (b) be hard to optimize without increasing business value, (c) correlate strongly with increased business value when optimized, (d) be universally applicable in many different contexts, without confusing edge cases, and (e) be transparent in how it’s measured and how it’s formulae are calculated. A new model of software development team performance was then created, using a diverse set of software development metrics that were all deemed strong, targeted many different aspects of the software development process, and shared little input data-points. | |
