Software Analytics needs the right data and the right questions

Software development leads to the creation of large amounts of data, such as source code changes, defects, and test executions. Software Analytics aims at uncovering patterns and actionable insights from this data to support software development and maintenance.

Selecting the right data is key to the success of Software Analytics. Unstructured software data (e.g., emails, bug descriptions, and technical forum discussions) is a valuable form of data as it opens a unique view on human factors involved in a software project; yet it is hard to harness. In the first part of the talk, I introduce automated techniques we devised, based on text search, machine learning, and island parsing, to mine this data and the initial results we obtained exploiting them.

Data alone is not enough: It has to be analyzed to answer the right questions to tackle relevant developers' needs.  In the second part of this talk, I introduce qualitative studies we conducted to uncover developers' needs. Particularly, I describe how we uncovered motivations, real outcomes, and fundamental challenges of Modern Code Review, thus opening a very promising research line to be tackled with Software Analytics.

I conclude the talk presenting the current paths I am taking to tackle Modern Code Review challenges. I also introduce the first steps we made together with the terrific graduate students I have the pleasure of supervising.

Alberto Bacchelli is assistant professor at Delft University of Technology, The Netherlands. His work involves conducting empirical studies, analyzing data generated during software development, and building tools to support managers and engineers when taking decisions and developing software.
His current research interests focus on mining unstructured software data and improving the modern code review process.
He received his Ph.D. in Informatics in 2013, from Università della Svizzera Italiana (aka University of Lugano), Switzerland.