They feed all of this into a machine learning model, who then look at the predicted average of the next output of the next SCADA monitoring station. If it exceeds the predicted output, it means that it is leaking, so that’s that. Intuitively it’s not hard, but it actually requires gathering of many different sources of data information.