Mike, Scott, Eric,

After brainstorming and discussing further on data here, we think that our final specifications for modelling data requirement need to be as follows:
We need bankrupt, default and nondefault (which covers nonbankrupt) accounts with 4 quarterly observation snapshots and 12 months performance following the latest snapshot.  Monthly performance indicator need to be available for the entire 12 months performance period.  We will need all the bankrupt and default accounts and comparable sample of good accounts with weights.
For the purpose of model validation, we will need data with above specs covering 16 months of performance.  This means that we will need rolling (4 quarterly snapshots + 12 months performance) data for 4 monthly shifts:

Input snapshots											Performance
1999 March end, 1999 June end, 1999 September end, 1999 December end			12 month end performance for Jan 2000 through Dec 2000
1999 Feb end, 1999 May end, 1999 August end, 1999 November end				12 month end performance for Dec 1999 through Nov 2000
1999 Jan end, 1999 Apr end, 1999 July end, 1999 October end					12 month end performance for Nov 1999 through Oct 2000
1998 December end, 1999 Mar end, 1999 June end, 1999 September end				12 month end performance for Oct 1999 through Sep 2000

We will need bankruptcy chapterwise indicator, if available during the performance period.  Our definition of default is either bankruptcy or 90+ days delinquency on any trade credit.
We have also discussed the cost aspect and we think considering the big picture, it makes sense to spend  the extra amount to get the right data for analysis.
Please let me know your thoughts.  This will require D&B to give us a modified quote and we could possibly move forward quickly.

Regards,
Amitava