In clinical trials, exploratory analysis of multivariate longitudinal data is key to the identification of patient subgroups having some of the studied features, e.g., biomarkers, shared similar time-course. In the context of tumor evolution kinetics, we present a new JMP add-in Non-Negative Tensor Factorization (NTF) – which allows for the identification of such patient subgroups and simultaneous selection of parameters featuring a prototypical time-course. We show that selected features can be subsequently used to model tumor evolution. Moreover, thanks to the unsupervised nature of our approach, resulting estimates of the overall survival are robust to overfitting. The add-in provides dynamic heat map visualizations that are meaningful to the biologist and clinician. Thanks to JMP 14 Python interoperability, a fast NTF package written in the Python 3.6 language is run in the background. Finally, NTF results are compared with the ones achieved by the JMP 14 functional analysis platform and an R package performing path modeling.