Widely used to make segmentation of customers or to define marketing typologies, the role of clustering is to group individuals of a population on homogeneous classes. It is expected that two individuals of the same class present similar characteristics. The implementation of clustering algorithms is more or less automatically done. However, the most effective procedures are the most flexible, requiring a precise configuration. Results largely depend on mathematical formulation of the concept of « similarity ». Clustering method is based on this mathematical formulation. An expert statistician should define the notion of proximity.