One of the assumptions in k-means clustering is that all features are equally scaled. So, we need to perform feature scaling if our data have very different scales. The data used in this tutorial are approximately in same scale. So, feature scaling is optional here.