In order to establish similarity patterns among accessions to facilitate the identification of potential crosses, it was made an a priori clustering of the potato Colombian Central Collection in subspe- cies andigena, by features of tuber skin color and pri- mary meat color of tuber. The necessity of including other tuber variables and doing another classification based on statistic methods grew through time and with the progress of multivariate data techniques. In this work, a non-hierarchical clustering of the Collection was carried out by means of a Partition Analysis which uses the k-mean algorithm before the morphologic characterization of 435 accessions only by tuber features. When comparing both clusterings, it was found that there was no coincidence between them; it was necessary to improve the objectivity in genetic resource characterization. The characteriza- tions were carried out in the Corpoica CI Tibaitata (Mosquera, Cundinamarca), where is conserved the Colombian Central Collection of potato.