Partner data model
The public record, response metadata and privacy semantics exposed by Partner Demand v1.
Demand record
| Field group | Fields | Meaning |
|---|---|---|
| Identity | id, source_version, lifecycle_status, changed_at | Public record identity and change state. |
| Evidence | evidence_class, evidence_count, count_suppressed | Modeled, observed or explicit evidence and its released count. |
| Need | risk_type, desired_business_outcome, harm_direction, time_horizon | The normalized protection outcome. |
| Context | business_segment, coarse_geography, at_risk_amount_band | Coarse dimensions allowed at the partner boundary. |
| Market design | closest_existing_series, reason_no_fit, proposed_contract_shape | The nearest current market, why it misses and a proposed structure. |
| Evidence window | first_seen, last_seen | Modeled dates or released cohort buckets, depending on evidence class. |
Customer evidence counts are either zero or a positive multiple of five. A customer record is never returned with a suppressed count. The record itself is omitted until it is publishable.
Metadata
meta.public_generation is present on every successful response. Resource-specific metadata can also include:
- response generation time;
- evidence freshness;
- pagination cursor;
- snapshot or incremental sync checkpoint;
- privacy rules; and
- the active rate-limit mode.
The freshness as_of value describes published evidence. It does not describe the response time. A no_data status can be a valid, coherent result.
Lifecycle status
The record lifecycle moves through bounded public states such as unmet, candidate found, mapped and resolved. Integrators should retain unknown future additive fields and use documented enum values for current behavior. A tombstone in the change feed removes a record from the local current view.