The cluster centroid is essentially the data point’s arithmetic mean. When we have less cluster variation, the data points in the same cluster will become homogeneous. Here is how k-means works: