The elbow method is used in determining the number of clusters in a data set. Use the "elbow" as a cutoff point to find a point where diminishing returns are no longer worth the additional cost. Choose a number of clusters so that adding another cluster doesn't give much better modeling of the data.