If this option is selected, the Output Means Data Set contains one observation selected from each cluster, making it a subset of the rows of the Input SAS Data Set. The observation selected is the one with the smallest distance to the cluster mean.
Check the K-Means Clustering box to activate this parameter.