Guideline for Addressing Current Data Science Challenges Through Leadership Practices

Abstract

Data science has become the enabler for many business applications, but almost 90% of data science projects in businesses fail. The main causes have been identified as issues with team and project management. While other studies have analyzed the processes and methodologies used in data science projects, the effects of leadership on project success have yet to be investigated. However, the leadership could have a crucial impact on project outcomes and individual team member performance. There is, however, no evidence that leadership recommendations help to overcome the main challenges data science projects face. Therefore, this research creates a guideline for data science team leaders with recommendations regarding the main challenges in data science projects. This was performed as part of a design science study whereby leadership recommendations existing in the literature are analyzed and mapped to the main problems. This guideline is then evaluated and compared with techniques from data science practitioners as determined through interviews with seven data science team leaders. From these results, improvements to the guideline are derived. The evaluation and discussion show that the recommendations from the guideline mostly match data science practitioners' methods. However, several differences are identified due to the literature’s lack of consideration of the data science team’s organization and the leader’s role within a company.