Transfer learning is a machine learning method used to train a network using prior knowledge from a source network. This method has been applied for anomaly detection in CNC machines to detect anomalies such as chatter. Before transfer learning can be applied, the Maximum Mean Discrepancy (MMD) score between the source and target data sets should be evaluated. A low MMD score indicates that the system can be transferred, but it is unclear what the physical implication of this score is. Experiments were conducted on a variety of CNC machines and materials to determine a relationship between this score and observed physical outputs and other measures. Through experimental testing, it was found that a low MMD score will indicate that the source and target machines will respond to an input signal in a similar fashion. Furthermore, the MMD scores and cross-entropy followed the same trend, but this did not correlate with the accuracy of the system.