K-means clustering is not a free lunch

On the one hand the article is saying, "Look at your data." On the other it's saying, "Visualizing data is hard in more than two dimensions." It doesn't really offer any suggestions to fix that.

I imagine using something like PCA comes with its own warts. What is a clustered to do?

FYI: I had some problems accessing the site, here's a mirror: http://www.r-bloggers.com/k-means-clustering-is-not-a-free-l...