It's a great question. We certainly are taking into account all of those factors, I would say, and are looking across....
I'll give you an example. We saw a that pattern of main track derailments was sneaking up, and we were very concerned. We pulled all the data we could and started doing data analytics on it. I don't want to oversell it. It was basically detailed Excel spreadsheets looking for trends and variances, etc. We found a pattern related to ambient temperature on the day of the accident and whether or not the track was signalled—i.e., it had a system to automatically identify broken rail and speed. We put in place the higher-risk key train rule last year to get at that.
Infrastructure is a big part of it. I would say the data we're seeing and the analyses we're seeing are taking us to a broader dimension that is getting more into human factors. It's not traditionally been part of the rail safety system, where we've focused on the rolling stock and the track and the procedures, but, for example, we've changed the work/rest rules for fatigue. I think we're broadening out the number of things we look at. As data comes in, we're finding factors that we may not have been looking at in the past that we see as being significant in outcomes.
That's why LVVR, locomotive voice and video recorders, are so important. We think there are some subtle human factors that we can understand by analyzing the patterns related to cab activity before incidents.