The platform is a graph database. Customer, Organization and Venue are the spine; every other entity attaches via a named relationship. There is no warehouse, no nightly export, no second copy of state. Guest Operations and Guest Experience are two reading angles on the same graph; Marketing Automation runs over the same record again; the AI Assistant cites every entity it reads.
The eight schema layers
44 entities, 374 fields, 176 relationships, arranged across seven product modules and one identity spine. Layer order top-down on the homepage stack:
Why a graph
Hospitality data is dense with relationships — bookings to orders, customers to venues, campaigns to opens, gift cards to senders. Modelled as a graph, every cross-reference is a first-class edge with its own properties (`visit_count`, `amount`, `sent_at`). Joins are constant-time, not query-time. Identity resolution becomes deterministic instead of probabilistic.
Why AI sits on top, not inside
AI Analytics is a layer of derived fields that read from every other layer and write back to the Customer node. It does not own data. Every prediction has a derivation chain — the explicit list of source signals — so the assistant can cite its sources on every response.
For the full canonical schema reference, see /docs/reference/schema.