2 Reconciliation After the Fact
Suppose a buyer receives fewer cartons than the carrier’s invoice charges for. Each side rebuilds the delivery from its own records, and the two versions first meet at the invoice. Days go into working out which one is right. This chapter shows why that work exists, why it costs more than it did five years ago, who has kept the history so far, and what a history must have before the parties can stop rebuilding it.
2.1 Each Party Keeps Its Own Records
Today a company can check a document. The actions around it, who issued it, who relied on it, and what happened next, sit in emails and in other companies’ systems. Each party keeps its own records of a shared process, and the whole history exists nowhere, so whoever needs it rebuilds it from the parts it can reach.
Even two parties who trade every day lack a history of their shared work that both hold. In 2020 FreightWaves reported that up to 70% of Walmart Canada’s carrier freight invoices had “some sort of issue”. Staff spent their time sorting out the charges, and carriers waited for payment. Then the retailer and its carriers worked from one shared history of each shipment. After that, the carriers disputed fewer than 2% of more than 150,000 invoices (Tabak 2020). The invoices did not disagree because anyone lied. They disagreed because each side reconciled the shipment from its own records and met the other side’s version only at the invoice.
Where reconstruction falls short, organisations fill the gap with trust, as when a manufacturer trusts a supplier’s database. The supplier can change a record in it later, or show different parties different copies, and nobody outside would notice.
2.2 Why Now
Four changes make a history that outsiders can check worth more than it was five years ago: more value rests on digital claims, faking evidence has become cheap, the burden of proof is moving onto the company, and the tools to sign at source already sit in people’s hands.
Governments are moving transactions that once rested on paper onto digital claims. Dubai Land Department forecasts that tokenised property will reach AED 60 billion, “7% of Dubai’s total real estate transactions”, by 2033; the figure is the department’s own forecast for a pilot launched in March 2025 (Dubai Land Department 2025). Trade documents are moving the same way. Under the UK’s Electronic Trade Documents Act 2023, “An electronic trade document has the same effect as an equivalent paper trade document” (s.3(2)) (UK Parliament 2023), and 13 jurisdictions have adopted UNCITRAL’s Model Law on Electronic Transferable Records, including Singapore and Abu Dhabi Global Market, several of them in part (United Nations Commission on International Trade Law 2025). A digital title or an electronic bill of lading is worth what the evidence behind it is worth: who checked the identity, who valued the asset, who approved the transfer, and in what order.
Faking the evidence has become cheap. In 2024 an employee in the Hong Kong office of the engineering firm Arup made 15 transfers totalling HK$200 million, about US$26 million, after a video call on which everyone except the employee was a deepfake (Hong Kong Free Press 2024; CNN 2024). The call carried every signal the employee normally relied on: familiar faces, familiar voices, a senior instruction. A check on who signed, with a key held in the signer’s own device, still holds, because a deepfake cannot produce that signature.
The burden of proof is moving onto the company. Since 21 June 2022, US Customs and Border Protection has presumed that forced labour made any goods produced wholly or in part in Xinjiang, and the importer must rebut the presumption (U.S. Customs and Border Protection 2026). The evidence that rebuts it originates with suppliers, their suppliers, and the facilities that made each input, each keeping its own records. An importer that cannot show that history in an order customs can check carries the loss.
The signing tools are already in place. The FIDO Alliance, the industry body behind passkeys, reported 5 billion passkeys in use worldwide in May 2026 (FIDO Alliance 2026), and Windows 11 requires “Trusted Platform Module (TPM) version 2.0” (Microsoft 2026). People already carry the credential that can sign a statement, and computers already hold hardware that can protect a signing key.
2.3 Who Keeps the History
Each generation of trust infrastructure changed whom an outside party has to trust, and only the last generation hands that party the whole history (Table 2.1).
| Generation | Capability gained | Outside party still trusts |
|---|---|---|
| Relationships | Action through known counterparties | The counterparty’s memory and good faith |
| Documents | Claims travel beyond their authors | The issuer and every custodian since |
| Databases | Structured coordination at scale | The operator’s account of its own records |
| Blockchains | Shared state without one controlling operator | The consortium’s members, consensus, and governance |
| Portable signed evidence | Every interested party holds the full history and reconciles it as it happens | The signers and the enrolment evidence behind their keys; the heartbeat hosts (parties that hold the Hub’s heartbeats and that the Hub does not control); the Hub’s availability to accept new entries |
Each earlier generation made the history easier to share, but each left the outsider holding only its own records, which it reconciles against the others’ after the fact. In the last generation every interested party holds the whole history and reconciles it as it happens. The Hub, the one service that orders and distributes entries, publishes a signed statement of the newest entry at an agreed interval (a heartbeat). Each holder checks that every entry carries its producer’s signature over everything before it, and that its copy matches the latest heartbeat (Chapter 3).
2.3.1 Why Blockchains Stalled
Several enterprise blockchain consortia closed in 2022 and 2023. Their principals cited funding, participation, and commercial viability (A.P. Moller - Maersk 2022; Wragg 2022b, 2023a; McGee 2022). Participants named what consumed the money. One, quoted in Global Trade Review, called it “a business model problem more than anything … people underestimate the time and money required to reach critical mass” (Wragg 2022a). Swiss Re’s chief executive said of B3i, a consortium of insurers of which Swiss Re was a member: “It would need all insurance companies to basically create smart contracts at the beginning, at the origin” (Howard 2022). We read this as the architecture’s cost surfacing as a funding problem: a network that is useful only once most participants have joined needs financing until they do.
TradeLens reached ten ocean carriers and connectivity to about two-thirds of global containerised freight (IBM 2021), and still closed. In 2018, Hapag-Lloyd’s chief executive, Rolf Habben Jansen, named the objection: “each individual project claims to offer an industry platform that they themselves control. This is self-contradictory” (Hill 2018). The UN’s trade-digitalisation database records that “the shipping industry didn’t have the appetite for sharing confidential data with a competitor” (United Nations ESCAP 2025; Najati 2025). Price and the industry’s reliance on paper documents weighed as well; only 1.2% of bills of lading were electronic (PierNext 2023; Wragg 2023b). Neither cause belonged to the protocol, and both have moved since: electronic trade documents now carry the legal effect of paper (Section 2.2), and a history that costs one write per event has no per-container price to charge (Section 3.2).
Where one accountable operator already existed, a shared blockchain added little. ASX’s independent review found “little value to processing all the business logic on-ledger as ASX maintain data integrity as the market operator”, and that consensus contributed to round-trip latency; settlement runs of more than 100,000 transactions failed under load (Accenture 2022). ASX wrote off A$245–255 million (ASX Limited 2022).
The blockchain projects that lasted solved one business problem, treated the blockchain as an implementation detail, and had a party able to bring participants in: a regulatory mandate, a party with workflow authority, or an existing orchestrator (Gottlieb 2026). Where no such party exists, portable signed evidence needs no one to bring participants in, because each outside party checks the evidence it holds without joining anything.
2.4 Five Requirements
Portable signed evidence needs five properties at once.
| Requirement | Practical Meaning | Failure Without It |
|---|---|---|
| Tamper-evidence | Any holder detects alteration or insertion; a check against the Hub’s latest heartbeat detects truncation and divergent copies | Whoever keeps the history can change it after the fact, and others cannot detect the change |
| Attribution | A key signs each entry, and enrolment evidence identifies the key’s owner | Responsibility diffuses across departments and service accounts |
| Independent verifiability | Any authorised holder can check the history without relying on the system that produced it, once the holder has the evidence each check needs | Auditors must trust the operator they are auditing |
| Every event, every party | The history receives every reading, revision and approval from every party, large or small, at whatever volume the work produces | High-volume sources and smaller parties fall outside the history, which then has gaps where disputes start |
| Continuity | One ordered history across parties and systems, in which physical checks, human decisions, and automated outputs all sit | Nobody can place a piece of evidence in sequence with what happened around it |