The era of big data research poses new challenges to the IS community. Nefarious underlying assumptions limit the potential of methods and techniques, both in practice and in theory. This is also the case for the field of process mining, a specialised area within the field of big data where data resulting from process executions is analysed to achieve real-world process insights and improvements. In this paper we first show how process mining methodologies would benefit from re-visiting two main assumptions underlying their approaches: (i) lack of attention to existing theories, thus overlooking the potential of process mining methodologies to theory building, and (ii) assuming that event data is a faithful representation of real processes. Secondly, we propose Signpost, a high-level process mining methodology, developed by applying the principles of Peircean Semiotics to existing process mining methodologies. The proposed methodology facilitates process mining research in realising its full potential in terms of (i) providing accurate insights for practitioners, and (ii) in contributing to theory building and development. An initial version of the resulting Signpost methodology is presented in this paper.