๐Ÿ“ฆ

Business Data Exchange

Orders, ship notices, invoices and remittances as Governed Data Records, composed from open standards. The meaning travels with the record. Whatever format a partner mandates is a projection of it.

Scheduled for the first quarter of 2027. The business documents library, composed from the OASIS Universal Business Language and GS1 identifiers on the published Default and ProvGov libraries, and a clone-and-run demonstration: a supplier and a retailer exchanging a year of orders, ship notices and invoices, and a deduction that survives a question. The components and the demonstration are published when they ship.

This is the data layer. Business Data Exchange represents your B2B transactions as SDC data. When an exchange needs a record both sides can verify, see the Verifiable Settlement Layer, which settles a change of state against the model and issues a receipt anyone can check offline.

The Challenge

Moving a business document between two companies has been nearly free for twenty years. Agreeing what it means is paid for again with every partner and every change.

๐Ÿ“„

The Meaning Lives Outside the Document

A document format defines the fields. What a field means in a given relationship is written in the partner's guide, and the supplier keeps one map per partner, renegotiated whenever either side changes anything.

๐Ÿ”„

Version Steps Recur, and Cost More Each Time

The last mandated version step of the US healthcare transaction standards was costed by the regulator at 25 to 50 percent of the original build, with testing at 60 to 65 percent of that. Industry assessed the next one at two to four times the effort of the last.

โ“

Identifiers Without Their Issuer

A product arrives as a buyer part number, a GTIN or a marketplace code depending on the partner, and nothing in the document says which system issued it. The supplier's cross-reference spreadsheet is the only place they meet.

What the open standards already give, and what they leave out

The document vocabulary is not the problem, and it does not need inventing. The OASIS Universal Business Language defines the order, the order response, the despatch advice, the receipt advice, the invoice, the credit note and the remittance advice as an OASIS Standard, royalty-free and freely available; it is the basis of European e-invoicing and the Peppol network. GS1 defines the identifiers of products, locations and shipments. What those standards leave outside the document is the same thing every format leaves outside it:

A value bound to its definition:

A schema says a field is a string of up to eighty characters. It does not say, in a form a machine can resolve, which concept the field is an instance of, what it may hold, and where that definition came from.

Provenance, validation and a receipt, in the record:

Who produced the document, from what, under which version of the partner's rule, whether it conformed before it was sent, and a settlement record both parties can verify afterward. Today those live in logs, emails and a dispute.

For a worked account of where that cost lives in the standard most US supply chains still run on, read the implementation guide: where EDI's recurring cost lives.

The Governed Data Record

Model the business documents once from open content. Keep the record. Emit whatever a partner mandates as a projection of it.

๐Ÿ“š

Business Documents from Open Standards

The order, order response, despatch advice, receipt advice, invoice, credit note and remittance advice, modeled from the OASIS Universal Business Language and composed by identifier slot from the published Default library (organizations, addresses, money, quantities, dates) and ProvGov library (provenance, audit, the order's state). Product, location and shipment identifiers as GS1 issues them, each carrying the system that issued it.

Order → Default organization + address + money, ProvGov activity + audit + order workflow

๐ŸŽฏ

The Meaning Is the Component

Every value in a record is an instance of a published component: a permanent identifier, a definition, its constraints and units, and links to the vocabulary that defines the concept. A partner who receives the record resolves the definition instead of reading a guide. Two partners who compose the same component agree because they reference one definition, not because two guides happen to align this year.

Requested Delivery Date → one component, one definition, every partner

๐Ÿงพ

Provenance, Validation and the Receipt

Beside its data, every record carries the activity that produced it, the agent, the document it came from, and an audit event, bound at the source. It is validated against its published model before it is sent, and each change of the order's state settles against the model's workflow with a receipt that names the model version and can be verified offline by a party who runs none of our software.

deduction → which rule version, what it required, did the notice conform, when

๐Ÿ”Œ

Your Translator Keeps Its Job

Nothing here replaces the network, the transport or the format a partner mandates. The Governed Data Record becomes the system of record; the translator you already run writes the partner's format from it and reads the partner's documents into it, inside your own environment. Move one partner at a time. Stop emitting a format when the partner accepts the record.

record → table, XML, graph, and the partner's wire format, all projections

How It Works

1๏ธโƒฃ

Model Once, from Open Content

UBL documents on the Default and ProvGov libraries

2๏ธโƒฃ

Keep the Record

Validated once, projected as table, XML and graph

3๏ธโƒฃ

Project the Partner's Format

Your translator, your environment, one partner at a time

4๏ธโƒฃ

Validate, then Settle

Refused before transmission; receipted on every state change

Read Next

Where the cost lives today, and the demonstration that shows the alternative.

Analysis

The implementation guide: where EDI's recurring cost lives

The meaning of a document lives in the partner's guide, so agreement is rebuilt per partner and per change. Where that happens, with the three cost figures that survive sourcing.

โฑ๏ธ 10 min read ๐ŸŽฏ Business/Technical
Demonstration, Q1 2027

Halvorsen Foods: a verdict that survives a question

A fictitious food manufacturer takes a three percent deduction for a ship notice that both systems processed correctly, because a rule and an identifier changed on dates the notice did not carry. Two clone-and-run stacks, a supplier and a retailer, a year of orders as Governed Data Records, and the deduction settled against the model with a receipt. Published when it ships.

๐Ÿงช Clone and run ๐ŸŽฏ The person who signs, and the people who run it

Use Cases

B2B scenarios where the meaning travels with the record

๐Ÿญ

Order to Settlement

Order, order response, despatch advice, invoice and remittance as one chain of Governed Data Records, each composing the same components, each settled with a receipt.

๐Ÿ“‹

Many Retailers, One Model

One purchase order model across every retailer a supplier sells to; each retailer's mandated format a projection written by the supplier's own translator.

๐Ÿ’ณ

Three-Way Match

Order, receipt advice and invoice matched on the components they share rather than on fields reconciled by hand, with the match itself a settled state.

๐Ÿšš

Chargebacks That Can Be Examined

A despatch advice validated against the retailer's published rule before it is sent, and a deduction that names the rule version, what it required and whether the notice conformed.

๐Ÿ“Š

Catalog and Product Data

One product record carrying every identifier it has been issued, GTIN, SKU, buyer and vendor part numbers, each with its issuing system named.

๐Ÿ”

Audit and Provenance

Every record states who produced it, from what and when, projected into a knowledge graph a third party can query without a reconstruction project.

What Changes for a Supplier

The structural claim, for a company integrating with many trading partners

N
Bindings, Not N²

Each party binds once to the shared components instead of negotiating a private interpretation with every other partner.

0
Forced Migrations

A published component is immutable, so a new version is an addition. Documents already exchanged stay valid against the version that governed them.

1
Place the Meaning Lives

Bound to the record, not described in a guide that each partner reads differently.

A Supplier With 15 Trading Partners

One map per partner, today

  • Read 15 implementation guides, each defining the same fields differently
  • Build and test 15 separate maps for the same document
  • Maintain all 15 for as long as each relationship lasts
  • Redo the affected ones at every partner-side change and every version step
  • Pay a hub or network to translate between versions, per partner, on an ongoing basis

The work scales with the number of relationships, not with the number of documents.

The Governed Data Record

  • Each party binds once to shared, published components
  • The meaning is in the record rather than in a document describing the record
  • A new version adds components; published ones are immutable and keep their identifiers
  • Documents already exchanged stay valid and stay readable
  • Adding partner 16 does not require renegotiating with partners 1 through 15; their formats stay projections

The work scales with the number of concepts, which is finite and shared.

Why there are no dollar figures here

This page used to carry a five-year model for this exact scenario, with per-line annual costs on both sides, a total saved, a percentage and a payback period. We removed all of it, because we could not source any of it. What a map costs depends on the documents in scope, the partner's guide, and what your hub charges, and those vary far too much for one modeled supplier to stand in for yours.

The structural claim does not need the model, and it is the one worth testing: the cost of integration scales with how many partners you have, because the meaning lives in a document each of them interprets privately. Bind the meaning to the record and it scales with how many concepts you exchange instead. You already have the numbers to check that. Count your active maps, and ask what you paid to change one last year.

๐Ÿ’ก What also follows

  • โ€ข Onboarding: a new partner binds to components that already exist, and its format is one more projection
  • โ€ข Fewer disputes: validation before transmission catches the document that would fail the partner's rule; a receipt records what was sent under which rule
  • โ€ข Cross-partner questions: consistent components make the knowledge graph answer across partners without a mapping table
  • โ€ข Settlement-ready: the same record settles value or state on the Verifiable Settlement Layer