Many approaches have been proposed to allow customers’ segmentation in retail sector. However, very few contributions exploit the existing semantics links that may exist between objects and resulting groups. The aim of this paper is to overcome this drawback by using semantic similarity measures (SSM) in customers’ segmentation to provide clusters based on product’ topology instead of numerical indicators usually used (i.e. monetary indicators). More precisely, we intend to show the main advantage of SSM with a product taxonomy in the retail field. Usually, traditional approaches consider as similar three customers buying respectively apple, orange and beer. However, human intuition tends to group customers who buy orange and apple because both are fruits. Our approach is defined to identify this kind of grouping through SSM and abstract concepts belonging to product taxonomy. Experiments are conducted on real data from a French Retailer store and show the relevance of the proposed approach.