Re: Reverse engineering a formula...

[samizdat]

devnul, you didn't tell us if the individual reports gave you any clues that you might be able to match to the dump from the same data set record.

Failing that, a combination of recursive brute force and success pruning (aka 'genetic') is the answer. I'm hoping you can assume that the data in the variables does have a constant format, so that once you find a solution that works for one set (within rounding deltas), it should work for every set. Where you'll have problems is if there are late payments with penalties and interest charges which are only stored in aggregate totals. That possibility would increase the scary factor immensely. :(

Best of luck; this looks like a case of Interesting Times^TM.