Processes and practices used in data science projects have been reshaping especially over the last decade. These are different from their software engineering counterparts. However, to a large extent, data science relies on software, and, once taken to use, the results of a data science project are often embedded in software context. Hence, seeking synergy between software engineering and data science might open promising avenues. However, while there are various studies on data science workflows and data science project teams, there have been no attempts to combine these two very interlinked aspects. Furthermore, existing studies usually focus on practices within one company. Our study will fill these gaps with a multi-company case study, concentrating both on the roles found in data science project teams as well as the process. In this paper, we have studied a number of practicing data scientists to understand a typical process flow for a data science project. In addition, we studied the involved roles and the teamwork that would take place in the data context. Our analysis revealed three main elements of data science projects: Experimentation, Development Approach, and Multidisciplinary team(work). These key concepts are further broken down to 13 different sub-themes in total. The found themes pinpoint critical elements and challenges found in data science projects, which are still often done in an ad-hoc fashion. Finally, we compare the results with modern software development to analyse how good a match there is.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.