Oh I’m sure. My point was just that the use cases least likely to be able to afford paying are clustered on historical data. And those clustered on real-time data tend to involve $$ anyways