A campaign does not stop working when the exposure ends, and a recommendation between two people never appears in either person's funnel. Aggregate models approximate both with smooth functions, which is fine until you need to know who was persuaded and by whom.
When each agent carries its own memory, its own fatigue and its own social ties, adstock, forgetting, virality and reputation stop being parameters and start being consequences. The same applies to time: discovery, comparison, consulting others and deciding become explicit states, so delayed responses and maturing demand come out of the model instead of being fitted onto it.