In this article, we conclude that students’ collaborative problem-solving activity can usefully be considered a complex system that is subject to unpredictable variation. Supplementing sound research methodologies with intuition-based speculations about the likelihoods of non-deterministic events suggests potentially more productive ways of researching and promoting collaborative problem-solving in mathematics. The motivation for this essay emerged when Laura became intrigued by the socio-emotional tensions between two students shown in a video excerpt. While interacting with a rich dataset, analysing it over and over, and over a long period of time, we realised that typical analytical approaches, based on deterministic lenses, can leave the interpretations of collaborative problem-solving incomplete. By discussing multiple ways of tackling the data, we tell our story of trying to establish what is optimal in collaborative problem-solving. We apply two modes of meaning making in the spirit of Bruner, elaborate validity aspects of complex data using the framework of Messick and suggest adding complexity aspects when discussing generalisability.