Proposta de modelo único para priorização de dívida técnica

Abstract

Over the years, software has become more and more present in the people’s and compani’es daily life. The software development companies aiming to understand customer needs, often investing in optimization, improving the product quality. However, software development companies have been dealing with limited amounts of time and resources, which have to be applied to generate in a short term financial gains and in addition they also have to develop features that satisfy the customers. Therefore, managers and developers focus other aspects than quality to establish a balance between the objectives, features and functionalities of the product, development can be taken shortcuts because of the short time to prioritize this over the quality. In 1992, the term technical debt was cited by Ward Cunningham to reflect the scenario in which to accelerate the development of software would require the writing of an immature code, thereby generating a debt. This metaphor gradually extended to other parts of the software, generally reflecting the immature, inadequate, or incomplete artifacts present in the development life cycle. Due to limited resources, identified technical debt items should be prioritized, seeking to classify and rank debts from technical factors or needs. Although there are some studies on technical debt prioritization, there are still many gaps on how to prioritize a technical debt item. So, it is important to elaborate new models to help developers to prioritize technical debt items, seeking to assist in decision making and aiming to clarify to entrepreneurs the real benefits linked to technical improvements. The objective of this study is to identify and categorize the technical debt prioritization approaches in the literature and, finally, a unique model of prioritization was proposed based on the characteristics derived from these studies. The approach developed is intended to prioritize classes affected by code smells, which are design problems at the source code level of a system and can indicate technical debt points. The model has six phases of classification, each phase representing specific ranking metrics. At the end of the prioritization process, the classes smelly with the highest priority of correction are obtained in relation to the considered criteria. The proposed model was validated with specialists in the area in order to verify its contribution and relevance.