Usually I try to tie the clusters to some kind of a bottom line metric. It's not always possible though.
To be honest I am not a big fan of segmentation as is, since it results in a set of subjective decisions when both deciding number of clusters and their interpretation.
An alternative is DBSCAN which comes up with natural clusters, but its also not 100 % foolproof as it has a group of unclusterables which at times ends up too high.
To be honest I am not a big fan of segmentation as is, since it results in a set of subjective decisions when both deciding number of clusters and their interpretation.
An alternative is DBSCAN which comes up with natural clusters, but its also not 100 % foolproof as it has a group of unclusterables which at times ends up too high.