Classifying objects into groups is a natural activity which is most often a prerequisite before any physical analysis of the data. Clustering and phylogenetic approaches are two different and complementary ways in this purpose: the first one relies on similarities and the second one on relationships. In this paper, we describe very simply these approaches and show how phylogenetic techniques can be used in astrophysics by using a toy example based on a sample of stars obtained from models of stellar evolution. We first perform some cladistic analyses to understand how the evolutionary behaviours of the parameters may affect the clustering process. We then show the mathematical principles that connect four different algorithms: one partitioning method, k-medoids, and three phylogenetic methods, Minimum Spanning Tree, Neighbor Joining Tree Estimation and Maximum Parsimony (cladistics). We finally use a challenging sample of stars to assess their performances. The phylogenetic methods naturally perform better than the partitioning method to retrieve the stellar lineages, and Maximum Parsimony, being more general, surpasses all approaches.