SINE Strategy Wiki

Coordination failure vs. bargaining failure — and what SINE actually does

24 min read

A synthesis page, not a source summary. It came out of the strategy discussion between Martin Pompéry and Elinor on 2026-08-15, in which the question “which thematic fields are SINE-relevant?” was pushed back to “which market failure is SINE in the business of fixing?”. Claims here are analytical, so they carry ⟨Inference⟩ markers rather than source citations; the two empirical anchors are PACT and CCTI, which do have sources.

The distinction

Coordination failureBargaining failure (Coase)
Interestsaligned — everyone prefers the other equilibriumopposed — there is a rent to divide
What failsexpectations: I move only if I believe you movethe trade itself: strategic misrepresentation, transaction costs
Scarce resourcecommon knowledge — a focal point, a credible signal of others’ intentcredible commitment — believable revelation of one’s own type
Missing information is aboutothers’ intentions (will you move?)others’ types (what is it worth to you? how good is your product really?)
Instrumentsconvening, standards, focal points, critical mass, sequencingmechanism design, verifiable disclosure, cryptography, escrow, enforcement
Hard limitnone in principle — it is a belief problemMyerson–Satterthwaite: with two-sided private information no mechanism is efficient, voluntary and budget-balanced at once

In one line: coordination lacks a focal point; bargaining lacks a credible commitment.

Information asymmetry runs through both, but it is different information — which is why it is not a third item on the same list. The cleaner layering:

Domain: commons — many parties, no enforcer Two failure shapes within it: coordination and bargaining Common mechanism: private information that cannot be made credible Instruments: convening/standard/focal point vs. mechanism design/cryptography/proof

The diagnostic test

One question decides which shape a case has:

Would the parties act if they simply knew what the others would do?

  • Yes → coordination. Convening and a standard are the instrument; cryptography would be expensive over-fulfilment.
  • No, because disclosure would hurt them → bargaining. Confidentiality and credible commitment are the instrument; convening alone runs empty.

⟨Inference: the practical value is in the misdiagnosis cost. Treating a coordination problem as a bargaining problem buys cryptography where an announcement and a convener would have done — slow and expensive. The reverse convenes endlessly among parties who will not move without a credible commitment.⟩

Where SINE’s own work falls

⟨Inference throughout this section — the classification is analytical, the underlying facts are sourced on the linked pages.⟩

  • PACT — predominantly coordination. Everyone wanted comparable Scope 3 data, no single company could impose a method, several candidate methods existed. SINE built the focal point: methodology plus network, until “everyone does it this way” became believable. The cryptography was not load-bearing here.
  • iLEAP — same shape, and since 2026-10-01 the classification rests on a primary source rather than on PACT’s shape. The specification builds a focal point for logistics emissions (ISO 14083 + GLEC, a data model, a wire protocol, a conformance gate) and contains no cryptographic mechanism at all. ⟨Lint 2026-10-05: that specification is a draft — v1.1.0-draft.2, a Living Document, not a Release. Here the draft status cuts less than it does on PACT, because the claim rests on an absence across the entire document rather than on any single provision, and a draft that intended to add cryptography would normally show the seam. But it is still one unreleased document, and this is the bullet that weakens if a confidentiality mechanism appears before Release.⟩ Sharper than PACT as evidence, because here the confidentiality constraint is explicitly named — Data Transaction 4 exists so carriers can withhold leg-level data “due to privacy or competition concerns” — and the instrument still is not reached for. The chosen answer is aggregation, which destroys granularity irreversibly for everyone downstream. ⟨Inference: a coordination problem with a bargaining residue inside it, settled by conceding the residue rather than solving it. See the confidentiality gap, addendum.⟩
  • Secure benchmarking / Polytune / Garble / Tandem — bargaining. Parties with opposed interests who will not reveal.
  • CCTI — entirely bargaining. Transaction costs, credible commitment, and the EIC as an attempt to make private beliefs bindingly revealable without exposing them — a direct assault on the Myerson–Satterthwaite obstacle.

The consequence. The split between SINE’s two centres of gravity is not organisational, it is theoretical: the adopted, best-evidenced work solves coordination; the distinctive technical asset solves bargaining. This explains H1 of the Hypothesenpapier (infrastructure role vs. end-product language) better than that paper does itself.

A worked case from outside: the carbon-accounting landscape (2026)

The first test of the diagnostic against material that was not written by SINE and does not use its vocabulary: The Carbon Accounting Landscape Review (ICC, 2026). It supplies one clean instance of each shape, and — less comfortably — external evidence for the second objection below.

Coordination failure, textbook. §4.1 inventories eight existing criteria sets for carbon data (IPCC, GHG Protocol, ISO 14067, PACT, FASB, IFRS S1, ESRS 1, the E-Ledgers proto-standard, plus EFI’s), observes that they “overlap substantially”, and concludes: “There is more agreement in the field than the plurality of frameworks suggests.” Aligned interests, substantial agreement on content, and no common architecture — the missing resource is a focal point, not a credible commitment. ⟨Inference: this is exactly the shape PACT was successful against, and it is why convening — the ICC’s own instrument, a coalition of coalitions — is the right instrument here. The Review does not name the failure, and responds to it by publishing a ninth list. See the eight criteria.⟩

Bargaining failure, equally clean. §4.2’s Criterion B requires that “authorised parties […] verify the number while limiting unnecessary disclosure of underlying operational data”, because product-level embedded emissions “can reveal energy intensity, process efficiency, and cost structure”. Opposed interests, private information about one’s own type, no reason to reveal voluntarily. The Review states it once and then treats it nowhere — see the confidentiality gap. ⟨Inference: a document written almost entirely in the coordination register proves the diagnostic’s practical point by omission. Its one bargaining sentence is the one it has no instrument for, because convening produces none.⟩

Evidence for objection 2, from an unsympathetic source. The Review’s §6.2 says of allocation: “Allocation is not merely a methodological detail. It influences how emissions are assigned to products, and therefore which products receive a commercial, regulatory, or reputational advantage” — and names methodology arbitrage, where “companies may be rewarded for selecting favourable calculation methods rather than improving underlying operations”. ⟨Inference: this is a standards body stating plainly that the content of the standard is distributive — that agreeing the allocation rule is not a neutral coordination act but a decision about who wins. That is the “aligned interests is often a polite name for a distributive fight” objection, made independently and about the exact domain in which this wiki’s flagship coordination success sits. It does not refute the coordination reading of PACT, but it does mean the reading cannot be assumed; it has to be argued case by case, and the allocation rules are where to look.⟩

A second worked case: biodiversity monitoring (2026) — the near-pure control

Added 2026-09-24 from Kim et al., From Data to Decision. Useful precisely because it is not a mixed case: 23 pages on standardising biodiversity and ecosystem-service data across two UN reporting regimes, in which commercial confidentiality is never named and cryptography is never once reached for.

⟨Lint 2026-09-28 — what this evidence is, since the whole section rests on an absence. Kim et al. is a position paper with a review’s apparatus, written by the community that would operate the layer it proposes (see the paper page). That status is carried on Data Commons, which discounts the paper’s silence — “its silence is weak evidence either way” — and was not carried here, where the silence does more work than anywhere else in the wiki. The two pages were written the same day and read, side by side, as a contradiction.

They are not, but the reason has to be stated rather than assumed: the two silences are differently motivated. Data Commons reads a silence about governance — about whether the authors’ own federation holds up without Ostrom’s machinery. The authors are advocates for that federation, so a self-serving omission is exactly what one would expect, and the discount is right. This page reads a silence about confidentiality and cryptography — an instrument set the authors have no stake in either promoting or avoiding. An unmotivated silence is better evidence than a motivated one. The asymmetry is the reason both readings stand.

What the silence still does not license: one article by one community is not the biodiversity literature. The claim this section can carry is that a well-resourced standardisation effort in this domain reached for none of the cryptographic instrument set; it is not that no such effort ever would. The market-bounding consequence below is therefore a hypothesis with one observation behind it, and the second observation to look for is a biodiversity data effort in which a firm is the reporting entity — which is precisely open task 10.⟩

Run the diagnostic. Would the parties act if they simply knew what the others would do? Yes — and the paper says so in its own vocabulary. The actors are states reporting to a convention and national statistical offices producing accounts; there is no rent to divide, nobody’s competitive position is exposed by a species-abundance time series, and the paper’s entire proposal is a focal point: a shared intermediate layer of essential variables with common vocabularies, self-describing file formats, and two published crosswalks. The constraint it does name is data sovereignty — states not wanting data to leave their jurisdiction — for which its answer is federation, not encryption.

⟨Inference: this is the control case the distinction has been missing. The ICC Review showed both shapes inside one document, which demonstrates the diagnostic but does not isolate either half. Here the coordination half runs on its own for a whole literature, and the observable consequence is exactly what the table predicts: the instrument set is convening, standards, vocabularies, capacity-building networks, and a data-cube specification. Nothing else. A community with a genuine multi-party data-sharing problem and no bargaining component arrives at none of SINE’s distinctive technology — which is confirmation the split is real and not an artefact of SINE’s own product line.

Two consequences worth carrying:

The diagnostic also works in the negative. Absence of the cryptographic instrument is evidence about the problem shape, not a deficiency in the community. Conversely, wherever the wiki finds confidentiality machinery in use, that is a signal a bargaining component exists even if the actors describe their problem as coordination.

It bounds SINE’s addressable market more sharply than the positive cases do. Pure-coordination domains are large, well-funded and institutionally active — and SINE’s sharpest instrument is irrelevant in all of them. What remains transferable there is the convening-and-focal-point half, which is the half that is not distinctive. The line only re-enters the bargaining register when the reporting entity becomes a firm — see nature-related reporting frameworks, where TNFD and the NGFS sit, and where that claim is still unverified.⟩

A third worked case: the reinsurance model layer (2026)

Added 2026-10-14 from Sounding the Alarm, an anonymous memo by senior UK/European (re)insurance professionals. It is the first case where the diagnostic is run forwards — the previous two classified documents after reading them; this one was predicted from an observable feature before the content was weighed.

The prediction. The section above commits the page to a rule: “wherever the wiki finds confidentiality machinery in use, that is a signal a bargaining component exists even if the actors describe their problem as coordination.” This memo is published by a group that is anonymous to the outside world and to its own members, recruiting over Signal, under a disclaimer that it shares nothing commercially sensitive (see Inside Track). Confidentiality machinery, visible before a word of the content is read. The rule predicts a bargaining component.

It holds. Run the diagnostic on the memo’s own stated problem — nobody can see aggregated or concentration risk across the industry. Would the parties act if they simply knew what the others would do? No, and not because they lack a focal point: exposure data is commercially sensitive, a firm that reveals a sharper view of its own risk reveals its pricing and its weakness, and the memo documents the consequence in its most self-incriminating passage — modellers asked to “give a ‘haircut’ to future projections […] to keep getting deals done”. Private information about one’s own type, strategically misrepresented, with a rent to protect. Textbook bargaining.

What is new: the instrument was built, and it is bad. The previous two cases are about an instrument not reached for. Here the actors reached for one — and the cheapest one available, which buys confidentiality by destroying attribution. The memo therefore cannot be checked, which is why the source page has to discount most of it. ⟨Inference: this is the clearest demonstration in the wiki of what SINE’s claim is for. The trade-off the group faced — credibility or confidentiality, pick one — is precisely the trade-off the core description below says SINE dissolves. An industry with a serious thing to say chose confidentiality and paid full price in believability, in public, in a document that states its own unverifiability. Demand demonstrated by substitution, which is better evidence than demand asserted.⟩

And a limit on the distinction, from the same source. The memo’s other problem — the three-step withdrawal from coverage, see the protection gap — does not fit either column, and the page is better for saying so. Nobody is failing to coordinate and nobody is failing to bargain: each firm reprices, restricts and withdraws correctly, the contracts conclude, and the cost lands on parties who were never at the table. That is an ordinary externality. Neither a focal point nor a credible commitment touches it, because the harmed parties are not counterparties.

⟨Inference: this cuts against the stance taken on 2026-08-15, where the externality frame was contested in five points and the two failure shapes were preferred to it. That preference still looks right for the cases this wiki works on — but the correct reason is narrower than “externalities are the wrong frame”. It is that coordination and bargaining failures are the ones with an information remedy, and therefore the ones SINE has an instrument for. Genuine externalities need a price or a rule, which is the state’s instrument and not SINE’s. The useful version of the diagnostic has three outcomes, and the third one is “not our problem” — a result worth being able to reach deliberately rather than by failing to notice.⟩

A fourth worked case: CVP in a value network (2026) — the cleanest bargaining instance

⟨2026-10-16, from the CE Association white paper.⟩

The forward test fires again, and earlier than before: the study group that assessed CVP was constituted as a “(confidentiality and consultation) committee” — so by this page’s own rule, a bargaining component should be present before a word of content is read. It is, and the source states it outright.

Applying CVP to a value network requires each participating organization to contribute its sales, intermediate-goods costs, COGS and SG&A for circular activity. The study group records the obstacle — “the disclosure of numerator and denominator data within VNs is difficult” — and then names the stake:

“Disclosing numerator and denominator values within a VN can be expected to have an effect in maintaining equity among the stakeholders that comprise the VN.”

⟨Inference: that is a distributive claim with no coordination content at all. There is no missing focal point here — the formula is published, the committee exists, everyone agrees what should be measured. What is missing is any party’s willingness to show the others a cost base that would be used, by its own counterparties, to argue about shares of a surplus. The other three cases in this page each mixed the two shapes; this one is close to pure.⟩

What makes it the most useful instance the page holds:

  • The parties want the disclosure. It is wanted for fairness, by the same actors who refuse it. That rules out the reading that confidentiality demands are merely obstructive.
  • The remedies attempted are visible and all institutional. A trustee (“data management by financial and other institutions as third parties”), an auditor (certification of the final value only), or coarsening (year-on-year instead of levels). The cryptographic instrument is not reached for — contrast the reinsurance case, where it was reached for and the instrument chosen (anonymity) destroyed attribution.
  • The quantity is arithmetically friendly. Eq. (3) is additively decomposable across parties, which is the shape MPC handles well.

⟨Inference: of the four worked cases, this is the one where the diagnostic most directly implies an action rather than a classification — bargaining failure, information remedy available, instrument named, venue open (ISO/TC 323 WG3 and the ISO 59011 annex).⟩

Strengthened by the primary source ⟨2026-10-16, same day⟩. The H-AIST research report adds three things the white paper did not.

First, it puts a number on the stake: the network-level figure is 1.27 with intra-network purchase costs eliminated and 0.49 without, against standalone values of 0.47, 0.50 and 0.49. So the bargain is not over a rounding difference — the whole case for cooperating is a factor of 2.6 that only exists if the parties reconcile each other’s figures.

Second, it makes the distributive content explicit as a purpose. A network CVP is useful, the authors write, because even when it is high “the distribution of value added and costs among participating companies can be highly skewed”, and a third-party calculation is “expected to help prevent large imbalances in benefits and burdens within the VN”. The indicator is being designed, in part, as an instrument for arguing about shares — which is why nobody wants to supply its inputs.

Third, it specifies the remedy as a role: “a trusted, impartial third-party organization” that collects the figures and computes the ratio, which “can exist outside the VN, or … within and become part of the VN’s formation.”

⟨Inference: that last point changes the shape of the opportunity. A bargaining failure whose remedy has been specified as an institution to be created is one where the institution’s design is still open — and the substitution on offer is not “trust a different party” but “need no party”. The venue is open and so is the slot.⟩

Proposed core description

Not “SINE is a verification institution” — that framing was rejected in the same discussion, on the grounds that verification at volume is a scale business that audit firms and cloud vendors absorb once it becomes mandatory, and SINE has no scale.

Instead:

SINE makes claims believable to others without those claims having to be disclosed — claims about intentions in coordination problems, claims about types in bargaining problems.

⟨Inference: this covers convening work and cryptography as two instruments of one thing rather than two businesses. Verification is only the technical half; the other half is who accepts the claim as believable, which is an institutional question, not a cryptographic one.⟩

Objections on record

  • Most real cases are both, sequentially. First coordination on the method, then bargaining over the terms. PACT itself: agreeing the methodology was coordination; how much primary data a company exposes is bargaining. So the diagnosis is per case, not per organisation.
  • “Aligned interests” is often a polite name for a distributive fight. Whoever’s existing process becomes the standard wins. Standards bodies are distributive arenas. If that holds, the flattering reading of PACT as a coordination success is partly wrong, and SINE was in fact refereeing a distributive conflict — a role with quite different risks. Supported from outside by the ICC Review’s treatment of allocation as market design — see the worked case above.
  • Ostrom’s monitoring is mutual and visible, and works partly because it builds trust socially. SINE builds monitoring that shows nothing. It could therefore satisfy Ostrom’s fourth design principle technically while destroying its social function. See Ostrom’s commons design principles.

A worked classification: Upright Oy

Added 2026-09-10, as an exercise in applying the split to an outside organisation. Upright sells quantified company “net impact” data. ⟨Inference: this is squarely a coordination failure — nobody is withholding information in a negotiation; there is simply no shared measure, so capital cannot be directed. Their fix (build the missing measure centrally and sell it) is a private-sector analogue of what PACT does collectively. They sit on the PACT/iLEAP side of the split, not the CCTI side.

The consequence is worth being precise about before anyone frames them as adjacent to CCTI: the real overlap is at the level of primitives (attested provenance, epistemic-integrity claims — see Coherence as a truth substitute) and not at the level of use cases. A shared primitive is not a shared problem.⟩

A fifth worked case: the data-space field itself (2026) — and an outside restatement of the distinction

⟨2026-10-27, from OECD (2026).⟩

The four cases above apply the distinction to a domain. This one applies it to the mechanism we would use in any domain — and arrives at the distinction independently, in different words.

The restatement

§3.1 of the report separates two failure modes that must both be present for a data space to be the right answer:

  • Price formation fails. Data value is context-dependent, so no common valuation exists across participants; total social value exceeds the sum of bilateral values (Frischmann); and Arrow’s information paradox means the act of negotiating a price requires disclosing the asset. This is our coordination failure — no one can compute the number that would let a bargain be struck.
  • Asymmetric cost-value distribution. Tier-2 suppliers pay for traceability data whose value lands on OEMs and regulators; farmers pay digitisation costs captured downstream. “Not merely distributional; it is a structural participation barrier.” This is our bargaining failure — the parties know what it is worth and to whom, and the one holding the cost is not the one holding the gain.

⟨Inference: the mapping is close but not identical, and the difference is instructive. Our version asks whether the parties could agree if they could talk. The OECD’s second mode is sharper: even a willing bargain fails when the party bearing the cost has no channel to the party capturing the value, because the value is diffuse across a chain rather than concentrated in a counterparty. That is bargaining failure with the counterparty missing rather than unwilling — worth adding to the diagnostic test as a third question: is there anyone to bargain with?⟩

The classification

The data-space field as a whole classifies as coordination failure — 235 initiatives, 25 operational, none scaling, and the report’s own diagnosis is that the binding constraint is governance, not technology, and that funding instruments “were simply not designed for the problem they are now expected, or may need, to solve”. Nobody is refusing; the mechanism for agreeing is absent.

But the failures within it are bargaining failures, and this is where the distinction earns its keep:

  • Covisint (2000–2004) — suppliers understood the proposal perfectly and declined, reading it as buyer leverage over pricing and inventory. Pure bargaining failure: the parties could have agreed, the terms were unacceptable. “The failure was not technological but rather institutional.”
  • BoostAeroSpace — anchors’ returns justified their investment, smaller suppliers free-rode. Bargaining failure within an otherwise functioning coordination arrangement.
  • Agdatahub (bankrupt Nov 2024) — neither: the coordination problem was real, the governance was built, and nobody’s private return cleared their own threshold. A provision failure, which is a third category this page does not currently have. See the bootstrapping problem.

⟨Inference: that third category is the finding. Coordination failure and bargaining failure both presume the arrangement would be worth having once reached. Agdatahub is the case where it is worth having collectively and not worth paying for individually — Frischmann’s condition, “when no subset of participants can justify the investment from their own expected returns alone”. A diagnostic that only separates coordination from bargaining will classify Agdatahub as a coordination failure and recommend building the coordination mechanism, which is what was built, and it went bankrupt. The test needs a third question after the first two: once coordinated, does any participant’s own return cover their own cost?⟩

For SINE’s own cases

The report supplies the sector-level answer to the diagnostic in five domains it identifies as structurally advantaged: supply-chain visibility across competing firms, health across jurisdictions, multimodal mobility, energy with distributed prosumers, smart city. Our PACT work sits in the first. CCTI’s reinsurance case sits in none of them — which is a prompt to run the test explicitly rather than a verdict. See data space boundary conditions.

A sixth worked case: plastics digitalisation (2023) — and a misdiagnosis worth naming ⟨2026-11-06⟩

Source: POLYPROBLEM 06 — Der Circularity Code.

The report reaches the coordination diagnosis on its own, from twenty practitioner interviews across the plastics chain, and states it twice — “Nicht die Technologie stellt die große Hürde dar. Es ist eine Frage der Haltung.” (p. 9) and, in the Zwischenfazit (p. 53), “Die ‚gute’ Digitalisierung … ist keine Frage der Technologie. Umso mehr geht es um Haltung.” Its recurring demand is “eine gemeinsame Architektur anstelle von Insellösungen”.

Applying the diagnostic test above gives a mixed case rather than a clean one, and the mix is the interesting part:

  • Coordination. Eight actors touch a shower-gel bottle before filling, “alle mit verschiedenen technischen Interessen und Sprachen” (p. 8); more than 20 waste-collection apps in Indonesia alone, from which GIZ derives a harmonisation need (p. 41); the repeated call for a common language and standards (Schiller p. 13, Floehr p. 50). All of these are failures to converge on one of several acceptable equilibria.
  • Bargaining. Who bears the cost of re-equipping sorting plants, of taking containers back, of continuous documentation (Helftewes p. 13, pp. 24–25, Tasche p. 40). These are distributive, and no amount of convening settles them.

⟨Inference: the report classifies the whole field as coordination — “Haltung” — and therefore recommends convening. Its own financing material shows that the second set is distributive, which is why three years of convening by the same foundation produced report 08 asking the finance question again. Naming the split is the most useful thing SINE can do with this material: the data-model and semantics layer is genuinely a coordination failure and responds to a commons; the cost-allocation layer is a bargaining failure and responds only to an instrument that moves someone’s expected return. Treating the second as the first is the characteristic error of a convening organisation, and this is the wiki’s clearest documented instance of it.⟩

One mechanism from the same source cuts specifically at the coordination half: federated learning (Kugler, SKZ, pp. 10–11) lets competitors train a shared model without exposing inputs — a coordination device that works precisely because it does not require anyone to bargain. See the confidentiality gap.

Graph View

Backlinks