Introduction and Concepts

Introduction & Concepts

What SIGNET is

SIGNET — Secure Intelligent Governed Network for Exchange and Trade — is an open specification for procurement networks: systems in which buyers, suppliers, and their agents discover needs, run sourcing events, submit and evaluate bids, award and contract, order, deliver, and invoice — across organisational boundaries.

What distinguishes a SIGNET network from a conventional procurement platform is that it assumes three things from the start:

1. Agents act, not just people. Some participants are AI ("synthetic") agents. They must be governed, bounded, and accountable.
2. Identity is decentralised. Parties are identified by stable, resolvable identifiers (often DIDs) and make verifiable claims about themselves with Verifiable Credentials — no central registry required.
3. Trust is cryptographic and auditable. Every material change is an append-only, tamper-evident event with provenance.

The central normative artifact is the Canonical Data Model (CDM) — a common vocabulary for representing everything that happens in such a network. This repository ships the CDM as JSON Schema, plus a JSON-LD context, controlled codelists, worked examples, and conformance materials.

The schema is the source of truth. Where the prose specification and the JSON Schema disagree, the JSON Schema takes precedence.

The problem it solves

Procurement data today is fragmented across mutually unintelligible standards (OCDS, EN 16931, UBL, eForms, Peppol BIS, ePO) and a long tail of proprietary platforms. That fragmentation:

• prevents interoperability — data cannot move cleanly between systems; and
• blocks agent deployment — an AI agent cannot reason or act across organisational boundaries when every counterparty speaks a different dialect, with no shared notion of authority, provenance, or consent.

The CDM is the Rosetta Stone: a single coherent model that bridges the established standards and adds the structures needed for agent-native, governed, decentralised commerce. It does not replace OCDS or EN 16931 — it profiles them and fills the gaps they leave.

Who it is for

• Procurement platforms & marketplaces that want interoperability with public and private procurement standards out of the box.
• Public bodies that must publish OCDS releases, issue EN 16931 / Peppol-compliant invoices, and keep auditable decision records (e.g. UK Procurement Act 2023, EU AI Act).
• Builders of procurement agents who need a model where an agent's authority, actions, and rationale are first-class, governed data.
• Auditors and regulators who need provenance and event integrity as native features, not bolt-ons.

Design principles

The model is governed by eight design principles (specification §1). They explain almost every modelling decision in the CDM.

#	Principle	What it means in practice
1.1	Reuse before invention	Where a concept is well defined by an open standard (OCDS, EN 16931, UBL, ePO), the CDM adopts and aligns to it rather than reinventing it.
1.2	Standards as profiles, not forks	A SIGNET document can always be projected down to a conforming OCDS release, Peppol BIS invoice, or eForms notice, and data in those formats can be lifted into the CDM. Mapping tables are normative.
1.3	Agent-native by construction	The model records not only what was decided but which agent decided it, under what authority, from which inputs, and with what rationale.
1.4	Identity is decentralised	Parties are identified by stable resolvable identifiers that may be DIDs; claims about them are W3C Verifiable Credentials, verifiable without a central registry.
1.5	Data sovereignty is explicit	The model separates an assertion from the right to access the data behind it. Consent and access grants are data (the Solid pattern). No central data warehouse is assumed.
1.6	Policy is data	Eligibility thresholds, evaluation weightings, approval routing, and agent mandates are machine-readable, human-auditable Policy objects — "rules as code".
1.7	Events are immutable	Every material change is an append-only Event with provenance. Current state is a projection over event history. Audit is native; tampering is evident.
1.8	Linked-data foundation	The canonical serialisation is JSON-LD: every object and property has a global URI, while remaining ordinary JSON to consumers that ignore the semantics. See Serialisation.

Key concepts in one paragraph each

• Canonical Data Model (CDM). The shared vocabulary. Four layers (Foundation, Process, Agent, Trust). Everything a SIGNET network reads or writes is expressible in, and losslessly mappable to, the CDM.
• Profile-and-bridge. SIGNET is a coherent superset that profiles existing standards and bridges between them. SIGNET-original structures (the whole Agent layer) are always cleanly separable from third-party standards.
• Synthetic agent. An AI agent operating as a first-class Party. It has declared capabilities, is bound by a Mandate, and its actions produce accountable Decisions.
• Mandate. The structural guarantee that an agent cannot exceed its remit — what it may do, within what limits, and where human approval is required.
• Decision. The accountability backbone: a record of what was decided, by whom, under which mandate, from which inputs, under which policy, with what rationale, and with cryptographic provenance.
• Event sourcing. State is derived from an append-only, hash-chained stream of events, making the history tamper-evident.
• Consent. A data-sovereignty access grant — the right of a named party, for a stated purpose, for a bounded time, to access data held by another party.

Where to go next

• Architecture Overview — the four-layer model in detail.
• Standards Mapping — exactly how SIGNET bridges OCDS, EN 16931, UBL/Peppol, VC/DID, and ePO.
• Worked Examples — see the model on real instances.