Most of the papers on initialization techniques for k-means clustering compare the proposed technique with random method. Why? This is but obvious that as the random name suggests it is very random and can give large iterations for k-means clustering.