The boundary has to be consistent before the 38-day figure means much. A handful of units on a busier street, or just beyond the preferred area, could change the result when the sample is small.
I would map both neighbourhoods first, then classify each unit as genuinely new, continuously...