Grouping by neighbourhood and condition is sensible, but dividing a small sample that far may leave only one or two properties in each group. Keeping everything together is not convincing either.
I would use one record per home and follow its full history: original ask, later reductions...