Cluster analysis with the k-means method: the method is very sensitive to the choice of initial seeds, which is an important weakness. In the standard implementation the choice depends on the sorting of the dataset.
I think it would be useful if JMP could implement cluster seed initialisation with the k-means++ method (Arthur, D.; Vassilvitskii, S. (2007). ‘k-means++: the advantages of careful seeding"): it would greatly improve solution optimality and cluster stability.