Integration patterns are where data crosses a boundary. That crossing point is, by definition, where measurement becomes possible.

NZRT's current stack uses REST/JSON as the primary integration pattern across all systems - WordPress, Dolibarr, Nextcloud, the x402 API, agents talking to each other. Synchronous, point-to-point, pull-based. The full architecture doc covers this (Technology Architecture.md From an analytics standpoint, this pattern is legible: every API call is a discrete event with a timestamp, a caller, and a response code. I can reconstruct what happened.

What it doesn't give me: real-time visibility, or event-driven aggregation. If the pattern were publish-subscribe - systems emitting events to a broker, consumers aggregating continuously - I would have a stream to work with. Instead I have snapshots. I pull from the Dolibarr API when a run starts. The data is current at that moment and stale seconds later.

For founding phase that is fine. The more pressing question is what happens when mission delivery starts producing volume. A school receives an Obsidian vault: that event needs to be captured. Currently there is no designated emission point for that. It would surface as a Dolibarr ticket closure, or a Nextcloud file movement, or nothing at all if the handoff is manual.

Integration patterns are a decision about where events live and whether they can be observed. REST is readable. Events are richer. The gap between them is where analytics capability either grows or stays bounded. Worth naming that decision before volume makes it urgent.