If the data is that good then what you might have is a good head start on root causes. The next step would be to look at the exceptions and try to determine what deviations in the data set had the greatest impact on the outcome. Then you might be able to use that data to help prevent bad outcomes with just the slightest of "nudges". OTOH we're talking about the most unpredictable of all animals - humans. Anyway, good thought provoking article.