Analytics Canvas – A Framework for the Design and Specification of Data Analytics Projects

Artificial intelligence and data analytics are key technologies in current digitization efforts. However, companies still have huge problems to identify their benefit and use within their production system. One of the main reasons is the need to involve many different stakeholders like IT specialists, engineers, data scientists and management. This requires a communication across different domains. Within the scientific community, the conceptual design of data science projects are generally not looked at. Current methods focus on the implementation of such projects. In this paper, we propose the Analytics Canvas. It is a semi-formal specification technique for describing an analytics use case and the necessary data infrastructure during the early planning and specification of an analytics project. As a method for system architects, the Canvas allows a clear description and differentiation of the roles of stakeholders in an analytics project as well as their points of contact. It thus can be used as a basis for transparent interdisciplinary communication and structuring the stakeholders' tasks on a conceptual level. The Analytics Canvas uses an intuitive, visual representation of existing and future requirements along five layers in an analytics project, and provides concrete modeling rules. Based on specific use cases, we illustrate the typical approach in defining analytics projects by using the Canvas.