Centroid
The mean point of a cluster. K-Means places centroids, assigns each point to the nearest one, then moves each centroid to the mean of its assigned points.
When to use
It is the object K-Means updates in the Assign, Update, Repeat loop, which stops when no point changes cluster. Inertia is the sum of squared distances from each point to its own centroid.
See Lec 5.