The criteria set proposed in The Carbon Accounting Landscape Review (ICC, 2026), §4. Every approach in that document — including PACT — is assessed against them, so the wiki needs them as a page rather than as a paragraph.
Provenance. Seven are “the guiding principles of the Panel’s Mission,
Opportunity, and Principles statement, adopted here in that statement’s
order and wording”; the opening definitional sentence of each restates that
document and the elaboration is the Secretariat’s own. The eighth,
Operability, is the Secretariat’s addition, “included because practical
implementation is critical to adoption, even though operability is not
itself an accounting principle.” The Mission/Principles statement itself is
not yet in sources/ — see the
ICC page.
The criteria are “stated neutrally and at the level of principle” and deliberately prescribe no allocation formulas, boundary cut-offs or assurance thresholds.
The eight
A. Accuracy — “correctly reflect the underlying flows of net emissions based on actual measurement or scientifically grounded calculation, under defined boundaries, reporting periods, proper classification, and consistent uncertainty assessment.” Direct measurement where practicable, with its share expected to grow. Materiality thresholds must “focus effort on meaningful contributions without allowing producers to omit items (such as capital goods or overhead) large enough to bias product comparisons.”
B. Verifiability — “allows a qualified independent party to attest that accounts were recorded faithfully […] to a reasonable assurance level.” Reasonable assurance, “the level financial accounts are held to, rather than self-verification”; every number traceable to origin through documentation and clear audit trails. This criterion also carries the confidentiality requirement — permissioned access, aggregation, data minimisation — which the Review then does not carry forward. See the confidentiality gap.
C. Comparability — “produced such that they can be compared across defined boundaries and reporting periods.” Requires “common units, common reporting periods, common boundaries, and common allocation rules”. Named as “the criterion most directly tied to the differentiation purpose, and the one most easily defeated by methodological flexibility”.
D. Representational faithfulness — “represent the reality of what they purport to represent […] free from material misstatement.” Anchored in physical law: “specifically mass and energy balance”. Data recorded neutrally, without deliberate or systematic bias.
E. Fungibility — “expressed in standardised and consistent measurement units, such that they can be aggregated, compared, transferred, and, where appropriate, netted or allocated across entities and products without loss of meaning or integrity.” Embedded emissions “pass by legal transfer in parallel with the physical aspect of a transaction”, and a figure from one supplier is interchangeable with the same quantity from another.
F. Conservation across entities — “each unit of emissions is assigned to one and only one entity at any point in time and retains that integrity as each unit […] moves through the value chain over time.” Mutual exclusivity: counted exactly once regardless of where in the chain it occurs; allocated completely and consistently to outputs; passing with title from seller to buyer.
G. Completeness of scope — “all emissions within the defined system boundary are fully counted across entities, such that the aggregation of all entities’ net emissions represents the total of the system.”
H. Operability — the Secretariat’s own. Implementable directly in enterprise systems “rather than reconstructed annually”; records emissions for transactions aligned to financial periods, “tending towards near-real- time where feasible”; scalable across products, facilities and firms; and “improves recursively as data quality and coverage grow.” With an important proviso against a purity trap:
“A decision-grade system does not require perfect primary data for every input on day one. It requires a rule-governed pathway that distinguishes measured data, supplier-specific data, modelled data, and conservative defaults; discloses or flags the quality of each; and creates incentives to improve over time.”
⟨Inference, added 2026-08-17 after Martin Pompéry asked what Operability actually means (2026-08-17: “Die Definition ist mir noch unklar”): H is four separate demands wearing one name — (1) native, implementable in ERP rather than reconstructed annually; (2) transactional, recorded per transaction and aligned to financial periods; (3) scalable across products, facilities and firms; (4) improvable, the rule-governed data-quality pathway above. Parts 1–3 are architecture claims and follow from the ledger; part 4 is a data-quality-tiering claim, and it is the part PACT already implements via PDS and DQR. Naming them together lets one criterion carry both an argument for a new architecture and a concession to the incumbent one.
Part 2 (transactional) is where the criterion set contradicts itself: it demands near-real-time transaction records while Criterion C demands comparability, and period-averaging is how the LCA tradition delivers C. That objection, and the absence of any treatment of services in a framework addressed to Scope 3, are worked out on their own page — temporal resolution and the service case.⟩
⟨Note on proof cost, 2026-08-17: an earlier version of this page read part 3 as a cost argument and cited the two papers’ verification figures against each other. Martin Pompéry’s objection, accepted: verification cost is a property of a deployment (Heiss’s Gas figure is a fact about Ethereum in 2023), not a fundamental limit, and it does not bear on the ICC document at all. The figures stay on the two paper pages. What survives is narrower — Man & Kleppmann (2026) is a first indication of what H costs in an operationalised environment, which is more than the criterion has had from anywhere else.⟩
Which criteria are new
⟨Inference: this grouping is mine.⟩ A–D and G are recognisable restatements of the qualitative characteristics already in FASB’s Conceptual Framework, IFRS S1 and ESRS 1 — relevance, faithful representation, comparability, verifiability. E (Fungibility), F (Conservation) and H (Operability) are the ones that do the work: they are the criteria an inventory cannot satisfy in principle, and they are precisely where Table 4 records PACT, ISO 14067 and the GHGP Product Standard as open. The criteria set is therefore not neutral between architectures — it is engineered to be satisfiable only by a ledger.
Where they come from (§4.1)
Eight existing criteria sets, which the Review says “overlap substantially”:
| Source | Criteria |
|---|---|
| IPCC national-inventory guidelines | transparency, accuracy, consistency, completeness, comparability |
| GHG Protocol (corporate) | relevance, completeness, consistency, transparency, accuracy |
| ISO 14067 | adds life-cycle perspective, prioritisation of science, explicit avoidance of double counting |
| PACT | builds on ISO 14067 + GHGP, adds harmonisation and scalability |
| FASB Conceptual Framework (2024) | fundamentals: relevance, faithful representation; enhancers: comparability, verifiability, timeliness, understandability |
| IFRS S1 (ISSB, 2023) | the same characteristics carried into sustainability disclosure |
| ESRS 1 (Reg. (EU) 2023/2772) | the same structure again |
| E-Ledgers proto-standard | record at source; receive on legal transfer; allocate by causal logic; verify to a reasonableness standard; never remove except by transfer |
| EFI Foundation | relevance, materiality, accuracy, unbiasedness, completeness, transparency, comparability, verifiability |
The Review’s own gloss: “There is more agreement in the field than the plurality of frameworks suggests.”
⟨Inference: eight overlapping principle lists, substantial agreement, and no common architecture is a textbook coordination failure — aligned interests, missing focal point. See coordination vs. bargaining failure. The Review does not name it as such, and adds a ninth list.⟩
Mapped to regulatory demand (Table 3, p. 24)
The point of the mapping is stated openly: it “explains why a framework satisfying the criteria will be well positioned to serve as compliance infrastructure rather than yet another voluntary overlay.”
| Criterion | Corresponding regulatory demand |
|---|---|
| A. Accuracy | CBAM and LCFS require scientifically grounded, installation- or pathway-level quantification, not unverified self-reported estimates; compliance value attaches to figures that determine tariffs, credits and penalties |
| B. Verifiability | CBAM, LCFS and EU ETS all require accredited third-party verification; China is building traceable, auditable data |
| C. Comparability | LCFS mandates a standardised model framework (CA-GREET); CBAM prescribes a harmonised methodology with CN-code-based reporting categories |
| D. Representational faithfulness | CBAM’s mass-based attribution and LCFS’s substantiated pathway records are checks that reported accounts match actual physical production |
| E. Fungibility | CBAM penalises default values; LCFS requires substantiated operating data; all push towards verified primary data in common units |
| F. Conservation across entities | CBAM attributes emissions to specific goods and avoids double pricing; “the differentiation purpose fails if an emission is counted twice or lost” |
| G. Completeness of scope | Missing emissions reduce comparability, weaken integrity, and can affect compliance obligations, tariffs or carbon prices |
| H. Operability | CBAM quarterly/annual cycles and LCFS pathway reporting reward systems that produce data continuously, not annually by hand |
Note what is absent from this table: confidentiality, despite being stated as part of Criterion B’s design.
Objections on record
The grid is redundant: A, C, D and G name almost the same thing. — Martin Pompéry, 2026-08-17, on first reading.
⟨The following assessment is mine. I think the objection is right about the taxonomy and needs one correction.⟩
Where it plainly holds. The Review derives its criteria partly from the FASB Conceptual Framework and IFRS S1 (§4.1), where faithful representation is a single fundamental characteristic with three components: complete, neutral, and free from error. The Review takes that parent concept and lists it (D) alongside two of its own components as siblings — free from error becomes A, complete becomes G. That is a category error on the face of it: a genus and two of its species presented as three coordinate items. Anyone applying the grid will score the same defect three times.
Where I would resist the collapse — two distinctions that survive.
- Comparability (C) is a different logical type. A, D and G are properties of a single figure — is this number right, unbiased, complete? Comparability is a property of a set of figures. Two PCFs can both be accurate, faithful and complete and still be incomparable, if each used a different but individually defensible allocation rule; and two PCFs can be perfectly comparable while both being wrong, if both applied the same biased default factor. Neither implication runs, so C is independent, not a synonym. The Review says as much when it calls C “the criterion most directly tied to the differentiation purpose, and the one most easily defeated by methodological flexibility” — the defeating mechanism is methodological choice, which is invisible to A, D and G.
- D’s neutrality half is not covered by A. A figure can sit inside its stated uncertainty band — accurate — and still be systematically chosen at the favourable end of every permitted option. That is precisely the methodology arbitrage the Review itself identifies in §6.2, and only D’s “recorded neutrally, without deliberate or systematic bias” catches it. D’s other half — the mass-and-energy-balance anchor — is a physical plausibility test, which is also not the same claim as accuracy.
So the honest count is not four-collapse-to-one. It is closer to: A and G should be folded into D as its components (as FASB has them), C stands alone, and D should be split into its physics test and its bias test, which are doing unrelated jobs under one name.
Why the redundancy matters beyond tidiness. ⟨Inference, and speculative as to intent — flagged as such.⟩ Padding the consensus half of the list changes how the list reads. Four of eight criteria restating long-settled accounting characteristics makes the grid look like an inventory of established principle, which softens the fact that the load is carried entirely by E, F and H — the three no inventory-based approach can satisfy, and the three on which Table 4 records every non-ledger approach as open. A shorter, non-redundant list would make the architectural pre-commitment obvious at a glance. I do not claim this was designed; I do claim the redundancy is not neutral in effect.
⟨Added 2026-08-17: with the second objection below in hand — H, the one criterion the Secretariat authored, contradicting C — this reading hardens into a claim that the grid was built backwards from the preferred architecture. Martin Pompéry directed that it be carried as a position; it is open task 1, point 8, with an explicit warning that it is the one point on that list which cannot be conceded gracefully by its recipient, and that the sharp version depends on first reading the Panel’s own principles statement.⟩
One defence of the Secretariat. It may not have been free to redraft.
The Review states that seven criteria are adopted from the Panel’s Mission,
Opportunity, and Principles statement “in that statement’s order and
wording” — so the overlap may be inherited from a document the wiki has not
read. That document is
not yet in sources/, and until it is, the wiki cannot
tell the Panel’s drafting from the Secretariat’s gloss. This should be
settled before the objection is put to ICC in SINE’s name.
Second objection: H demands what C forbids, and services are missing entirely. — Martin Pompéry, 2026-08-17, on the Operability definition.
Checked against the Review’s own text and upheld on both counts. Criterion H’s second demand — transaction-level, “tending towards near-real-time” — runs against Criterion C, because period-averaging is the LCA tradition’s mechanism for removing temporary variance and making figures comparable; the Review treats averaging only as a defect and never states what it is for. And the word “service” appears exactly once in the entire document, in the ICC boilerplate — in a framework addressed to embedded emissions across value chains, i.e. mostly Scope 3, i.e. mostly services. Worked out in full, with the verbatim passages, on temporal resolution and the service case.
⟨Inference: note the pattern with the first objection. Both defects are in the grid, and this one is in the criterion the Secretariat wrote itself rather than inherited from the Panel — so the “not free to redraft” defence above does not cover it.⟩
The criteria against an actual protocol ⟨2026-11-06⟩
The Review assesses approaches. With the Technical Specifications for PCF Data Exchange v2.2.0 now in the wiki, five of the criteria can be checked against a specification that exists and is implemented, rather than against a programme’s self-description. This is the exchange layer of Framework v2 (April 2024), not PACT’s current state — but it is what the “40+ conformant providers” figure was earned on.
- E. Fungibility — met, and this is the protocol’s strength.
declaredUnitis a closed enumeration grounded in SI (liter,kilogram,cubic meter,kilowatt hour, …); every emissions property is specified askgCO2e / declaredUnit; geography uses ISO 3166 and UN regions, product category UN CPC 2.1, characterisation factors a named IPCC AR version. A figure from one supplier really is expressed in the same terms as the next one’s. - C. Comparability — syntax yes, method no. §7.2.1 places footprint
calculation out of scope of the protocol. A host can be fully
conformant while exchanging numbers computed under any methodology; the
wire format records which standards were claimed
(
crossSectoralStandardsUsed,productOrSectorSpecificRules) but nothing checks them. Comparability is delegated to the Framework, and conformance to the two is certified separately. - B. Verifiability — not met, by design. Assurance travels as an
unverifiable claim (
providerName,level,completedAt, …) with no signature over the payload, and logging — “access logs and audit trails” — is declared out of scope in the same section. “Every number traceable to origin through documentation and clear audit trails” has no protocol-level expression. Criterion B’s confidentiality half is handled by access control that hosts merely SHOULD implement. See the confidentiality gap. - F. Conservation across entities — not addressed, and not addressable
here. The protocol moves documents, not quantities. A
ProductFootprintis fetched and copied; there is no transfer, no balance, no mutual exclusivity check, nothing that could detect the same emissions being claimed twice. ⟨Inference: this is the precise technical content of the ICC’s “exchanges data rather than accountable quantities”, and it follows from the 2022 architectural exclusion — no storage, no central intermediary — rather than from an oversight. Conservation needs state that persists and reconciles; PACT ruled that out before the protocol was written.⟩ - H. Operability — partly, and the near-real-time half is the weak one. The Events mechanism (CloudEvents, mandatory since 2.2.0-20240312) provides push and request/fulfil flows, which is the machinery near-real-time exchange would need. But §7.8.4.1 says the specification “does not yet specify the processing requirements of a host system upon receiving such an event”, filtering is best-effort and may be ignored, and partial results (HTTP 202) are legal. Under Postel’s law that tolerance is deliberate and is what lets heterogeneous hosts interoperate — but it means latency and completeness are properties of each implementation, not of the network.
⟨Inference: the pattern is consistent. Everything the criteria ask of the representation — units, boundaries, vocabularies, lineage — the protocol delivers well, because representation is what an interoperability profile can fix. Everything the criteria ask of accountability — verification, conservation, audit trail — is out of scope by explicit statement. That boundary is where SINE’s instruments would attach, and it is also the boundary the ICC proposes to move.⟩
Cross-references
- The Carbon Accounting Landscape Review (ICC, 2026)
- Ledger-based carbon accounting
- Product-level carbon regulation
- Pathfinder Network — assessed against these criteria
- PACT Data Quality & Verification — PDS/DQR against criteria A and E
- Tech Specs for PCF Data Exchange v2.2 (2024) — the criteria checked against a shipped protocol ⟨2026-11-06⟩
- Interoperability profile — why the representation/accountability split falls where it does ⟨2026-11-06⟩
- The confidentiality gap — Criterion B’s unfinished half
- Verifiable Carbon Accounting — all eight criteria assessed against what cryptography can and cannot prove
- Temporal resolution and the service case — the second objection, in full
- Coordination failure vs. bargaining failure — the diagnostic applied to the criteria set
- Carbon Measures
- Open tasks — SINE’s response is owed
Sources
sources/whitepapers/2026-08-17-icc-carbon-accounting-landscape-review/2026-08-17-icc-carbon-accounting-landscape-review.md(§4.1, §4.2, §4.3)sources/whitepapers/2026-08-17-icc-carbon-accounting-landscape-review/tables-recovered-from-pdf.md(Table 3)sources/standards/2026-10-29-pact-data-exchange-protocol-v2.2.md(§5.2, §5.4, §5.16DeclaredUnit, §7.2.1, §7.8.4.1 — the 2026-11-06 section)