On the Current State of Industrial Data Science: Challenges, Best Practices, and Future Directions

Data science provides organizations with the capability to draw insights from large amounts of data. Numerous sectors such as banking and marketing leverage this potential already. In comparison, industrial enterprises are slow in creating profit from their data. This originates from various factors, such as availability and quality of data, reliability of solutions, or a lack of organizational embedding. Traditionally, Industrial Data Science (IDS) is referred to as the application of data analytics within industrial settings. In this paper, we expand this definition, positioning IDS as an interdisciplinary field to encompass the holistic integration and organization of data science efforts into industrial enterprises. We provide three contributions. First, we conduct a thorough literature review on current issues and the state of the art of implementing IDS. Second, we conduct eight interviews with practitioners from different industrial enterprises to identify best practices and research gaps. Third, we condense our findings into a single framework which structures multiple layers of IDS in real-world organizations, such as projects, organizational embedding, and roles. Our findings can lead practitioners and academics in their quest to streamline their IDS efforts, guiding them into structured application and further research directions.