The jump from 27 to 123 distinct businesses in the top three is more useful here than average rank, since changing the search depth can distort that average. I'd keep the original 11×11 coordinates and 200 m spacing fixed, then compare the share of pins in the top three within separate distance bands from each business, before and after the change. A loss mainly on the outer pins, with the nearest pins holding up, would support the proximity explanation without proving an algorithm change. Since you've already checked a dozen points while logged in, I'd use those same points to record the search pack and map results separately, with the exact query and time, rather than assume the two surfaces match. Also split the queries with a neighbourhood name from those without one, given the difference you noted in post #19.
