Conversion 3 was completed in 4 working days, plus about a half day to run post-conversion processes. We had expected it to take about a week, so it looks like we are right on track.
Our goal for this conversion was to convert all valid data, and that goal was met. We were somewhat surprised by the number of errors that we encountered along the way - some errors that we had seen before but hadn't been fixed, but most new errors since the last conversion two months ago. We fixed these errors as we were converting, in order to meet our goal of leaving no data behind, which slowed us down a bit.
For the July test conversion and also the August live conversion, we plan to run a mini-conversion a week or two ahead of conversion, in order to find these errors and get them fixed when we aren't under pressure to get the conversion completely quickly.
Our computed columns for solving address ambiguities worked well in this conversion. For N1 (Other addresses) the modification to the computed columns that would map former addresses to former addresses and current to current worked better than expected. We were able to resolve all but 115 addresses at N1 with this technique.