In this paper, we study how existing definitions and algorithms for latency optimization can be augmented with information about about the transactions themselves in order to optimize peering algorithms.