Combining data analytics and developers feedback for identifying reasons of inaccurate estimations in agile software development

Background: Effort estimations are critical tasks greatly influencing the accomplishment of software projects. Despite their recognized relevance, little is yet known what indicators for inaccurate estimations exist, and which are the reasons of inaccurate estimations. Aims: In this manuscript, we aim at contributing to this existing gap. To this end, we implemented a tool that combines data analytics and developers' feedback, and we employed that tool in a study. In that study, we explored the most common reasons of inaccurate user story estimations and the possible indicators of inaccurate estimations. Method: We relied on a mixed method approach used to study reasons and indicators for the identification and prediction of inaccurate estimations in practical agile software development contexts. Results: Our results add to the existing body of knowledge in multiple ways. We elaborate causes for inaccurate estimations going beyond the borders of existing literature; for instance, we show that lack of developers' experience is the most common reason of inaccurate estimations. Further, our results suggest, for example, that the higher the complexity, the higher the uncertainty in the estimation. Conclusions: Overall, our results strengthen our confidence in the usefulness of using data analytics with human-in-the-loop mechanisms to improve effort estimations. (C) 2019 Elsevier Inc. All rights reserved.