It's not just not being collectible. I think the issue here is that we are likely sending workloads to people we should not be sending to, because the account will find its way to resolution. The new generation of systems have memory, they recall, and they can go back and say, if that taxpayer paid after the second letter or the first letter, then we may not need to involve a lot of people in that, because that kind of profile will tell us that.
On the reverse, we send a lot of our accounts to collectors that probably are too complex for that level of collector. We should know ahead of time which accounts are the most complex, and we could send them straight to a more expert collector because the complexity of dealing with the cases will require roadwork, it will require an agent who has more training; it's a more complicated debt, a repeat type of situation. Those are the criteria we are looking to acquire in our data, so we can actually become more strategic in the way we use our resources and adopt strategies better suited to resolve those cases.