Great write up. It would be nice to see more detail on the clustering methodology for determining what you defined as an entity and what the confidence/error was on this classification. It’s hard to draw meaningful conclusions between the upper bound total addresses and the entity number you gave without an explanation of how you derived the difference.