On the testing, there was a May 2015 checkpoint, meaning that we would make a decision on whether we would go live in October and December, which were the first implementation dates. At that point in time, we'd been at it for about 70 months. It was a 79-month project.
At that point in time, we found too many critical defects in the system to risk going with either the pilot or going live. That is when we started to engage, at least at my level, with IBM to get through those defects. By September we felt we had cleared all of the defects that were critical. There were still a number of outstanding ones that we could have workarounds for, including having more people in the pay centre. We were actually using the pension centre to augment that as well.
The next effort was to try to get that number of people down and really hold IBM's feet to the fire to deliver the system without defects. By September we felt we had cleared those defects and that any of the remaining ones would not have an impact on the system, or that we had effective workarounds to deal with them. We worked all the way through until mid-January and cleared those defects progressively from September to mid-January.