Mathematical modelling is at the core of metrology as it transforms raw measured data into useful measurement results. A model captures the relationship between the measurand and all relevant quantities on which the measurand depends, and is used to design measuring systems, analyse measured data, make inferences and predictions, and is the basis for evaluating measurement uncertainties. Traditional modelling approaches are typically analytical, for example, based on principles of physics. But with the increasing use of digital technologies, large sensor networks and powerful computing hardware, these traditional approaches are being replaced more and more by data-driven methods. This paradigm shift holds true in particular for the digital future of measurement in all spheres of our lives and the environment, where data provided by large and complex interconnected systems of sensors are to be analysed. Additionally, there is a requirement for existing guidelines and standards in metrology to take the paradigm shift into account. In this paper we lay the foundation for the development from traditional to data-driven modelling approaches. We identify key aspects from traditional modelling approaches and discuss their transformation to data-driven modelling.