One can make visualizations of the clusters, plot the F statistic, etc, but they generally always improve as more clusters are included (increase k).
Should clustering always be just considered exploratory, and if so what would you do next?
[link][6 comments]