Seleção de projetos na gestão de portfolio de desenvolvimento de software

Abstract

Project Portfolio Management (PPM) of software development is a complex decision-making process, surrounded by uncertainties which implicate analysis of several criteria. This decision-making process, supported by methods of multiple criteria decision, commonly, result in the selection and ordering of projects, according to criteria defined by the decision makers. One of the mainly challenges consist in making the process of selection and ordination agile and facilitate, once the PPM models are formals, complexes and require an assiduous effort from the portfolio management team. Consequently, this study aims to analyze this issue, addressing related concepts, processes, flows and conceptual basis related to the selection and ordering of projects, to propose a framework confluent with the uncertain nature of software development projects. For that purpose, a Systematic Literature Review (SLR) was provided, which aimed to identify works with connections between areas of project portfolio management and information technology; The second SLR was addressed to analyze the main methods and criteria used to select and order the software development projects. Subsequently, the results identified the main criteria and group criteria applied in the process of selection and order of the software development projects, as well their stages and adherent methods. Based on identified theorical lines, a diagnostic survey about business practices for selecting software development projects was made by semi structured interviews applied in 13 companies. A framework version was developed including processes, stages and PPM activities searched from the SLR that focused in works involving PPM in software development. The results obtained by the operational practice from the interviews led to the proposition of a new version of both the criteria and criteria groups (required and optional) and the processes, steps and activities established for the framework. Considering: i) the flow of software development, ii) the approach of how and in which moment of the process the decisions are taken, iii) the required and optional criteria, iv) the process of PPM. The definition of methods to handle uncertainties and to select and order projects were used in order to be appropriate for the context. The computational model 2-Tuple linguistic was used to handle uncertainties, and VIKOR 2-Tuple multi-criteria method for selection and ordering of projects. The framework was tested in 4 companies from the semi structured interviews. For each company, groups of required and optional criteria identified respectively from the SLR and the semi-structured interviews were organized. The analysis results show a diversity in the studied cases, indicating situations with one, two and three alternatives as viable compromise solutions. The mainly contribuition from this research is the simplification of a process which, as a rule, proved to be complex and often unfeasible, as seen in previously SLR’s. Furthermore, the research adds the computational model 2-Tuple linguistic to the methods to treatment of uncertainties, once the application of this method was not identified in other works in the SLR yet. As a sequel to this work, it is suggested that: i) add a formal process to identify and resolve risks; ii) aggregation of methods/processes which allow the acceptance of projects with no prior knowledge of their existence and iii) the application of the framework in more cases aiming the longitudinal monitoring.