Abstract. Research in data science processes tends to be based on the scientific method which emphasizes on the hypothesis-testing approach; they are task-focused and deal with high-level phases for the delivery of a single and completed data product. They do not consider proper project management processes or practices nor do they tackle the problems raised in exploratory data science projects. This paper sets out to question the standardized data science process models and proposes to refine the existing data science process models into an orderly manner to guide data scientists in the execution of their projects. This study also compares the exploratory type of data science projects with software development agile practices. The proposed Data Science Project Management Framework (DSPMF) identifies management processes, tools and techniques that would be suited for each phase in the exploration of a data science project. We argue that, when working on a multiple-phase data science project, a distinction should be made between the various types of situations and phases. For each stage of the project, different types of methodologies, iterative processes and coordination tasks are needed. The proposed framework is based on four phases with proper “Data Science Project Management” principles or mindset appropriately incorporated with agile practices and as such it challenges the traditional project management models.
Keywords: Data Science; Project Management; Exploration; Framework; Agile; Cognitive coordination; Management Processes, Methodologies, Lifecycle, Phase Gates