Again, I think the problem was that initially they felt they could resolve those pay requests for the 82,000 people they originally identified. They thought they could get those all resolved by October 2016, and I think then discovered that was a larger task than they expected. In the meantime, while they were trying to resolve those, additional requests were coming in and the problem got worse.
Again, I think it was very much not identifying the size of the problem early enough in the process.