Usage of K-means clustering

The K-means algorithm usually compares well to more refined and computationally expensive clustering algorithms concerning the quality of results.

The range of possible values of the k parameter is sufficiently small so that you can examine this range by running the algorithm several times with different values of k.

Note: Partitioning the data into more than a dozen clusters or into a few dozen clusters is not useful. An application is likely to narrow down the range to a few k values.