The second-order collective-action problem that sits between a qualified use case and a working one: the infrastructure must exist before the network effects that justify it, and each participant’s onboarding investment must be justified by returns that only arrive at scale.
Primary source: §3.4 of OECD, Data Spaces between Co-ordination Challenges, Financial Viability and Public Policy, 2026, with the case material from Boxes 3.3–3.5 and 4.1.
Two levels, not one
- Collective investment in shared infrastructure — reference architectures, connectors, trust frameworks, semantic models — must precede the network effects that justify it.
- Individual participation: each participant’s own investment in onboarding, data preparation and semantic mapping must be justified by returns that may not materialise until the ecosystem reaches scale.
Most practitioner discussion collapses these; the distinction matters because the two levels have different fixes. Level 1 is a club-good financing problem; level 2 is an expected-return problem for a specific firm on a specific date.
Why the infrastructure layer is a club good
Brousseau, Toledano and Eustache (2024): data-space infrastructure has high fixed development costs, low marginal cost per additional deployment, is non-rivalrous within the ecosystem and excludable to non-participants — the definition of a club good. Olson’s free-riding logic therefore applies directly: each potential contributor prefers that others build it. Bisière et al. (2025) formalise the chicken-and-egg: participants will not join without sufficient data, and data will not be contributed without sufficient participants.
Platforms normally break this with cross-subsidisation — subsidise one side, charge the other (Rochet & Tirole, 2003; Caillaud & Jullien, 2003; Evans & Schmalensee, 2010). Data spaces structurally cannot, and in Europe may be legally barred from doing so: the DGA’s neutrality requirement constrains the operator from capturing value on one side to subsidise the other. ⟨Inference: this is the report’s most under-advertised finding — the governance property that makes a data space trustworthy is the same property that disables the standard bootstrapping instrument. The Digital Omnibus’s proposed relaxation of separation from structural to functional would loosen exactly this constraint, which is why the OECD treats it as a double-edged proposal.⟩
Frischmann’s refinement — the test that actually bites
The report declines to treat free riding as automatically fatal, citing Frischmann’s (2012) infrastructure argument:
free riding does not necessarily reduce contributors’ investment incentives, because contributors may capture sufficient private value independent of others’ contributions.
Which yields the operative condition:
bootstrapping is a binding constraint specifically when the expected private returns to anchor participants fall below their investment threshold despite positive collective returns — when no subset of participants can justify the investment from their own expected returns alone.
That is a test a specific initiative can be run against, not a general lament. The case evidence below is organised by whether an anchor existed whose own returns cleared its own threshold.
The cases
Anchor with an indefinite horizon: the Finnish Maritime Data Space
Operated by Fintraffic, a state-owned company — which resolves both levels at once: the anchor has a public mandate, an indefinite time horizon, and no requirement that the infrastructure pay for itself from transaction revenue. Bootstrapping proceeded in five sequential steps over roughly eight months, each step delivering value before the next was attempted, so no participant had to fund the whole trajectory up front.
The flagship result — Virtual Port Arrivals on the Luleå–Oxelösund route — is the closest the report comes to an outcome number: an average 24 % CO₂ reduction and 20 % energy reduction per voyage, extrapolated to roughly EUR 45 million per year in fuel and ETS savings and 135 000 tonnes of CO₂ avoided against a 2022 baseline.
How hard that number is ⟨corrected 2026-10-29⟩. This page previously called it “the report’s one hard outcome number”. The OECD’s own endnote 22 disclaims almost every property that would make it hard: the figures are “based on business calculations from real ESL Shipping voyage data”, the EUR 45 million is “a counterfactual projection (what would have been saved had the system been operational across Finnish maritime traffic in 2022), not measured aggregate savings”, the 24 % and 20 % reductions come from one shipping company’s operational data on one route and are “not independently verified”, and “no peer-reviewed publication was identified at time of writing”. ⟨Inference: the correction does not dissolve the case — the revenue-sharing arrangement below is the structural finding and does not depend on the magnitude — but it does mean the field’s best-evidenced success story rests on a participant’s own unaudited numbers. That is itself a data point about the state of the evidence, and it is the same shortfall the wiki records for the EUR 1.7 bn Gaia-X portfolio figure: the quantities circulating in this field are produced by the parties with an interest in them.⟩ Crucially, the gains were shared across shipping company, charterer, port and system operator through an explicit revenue-sharing arrangement — the asymmetric cost-value distribution was designed out rather than left to settle.
Anchor coalition with regulatory forcing: ABtC
Japan’s Automotive Battery Traceability Consortium (February 2024): 14 Japanese automakers plus two industry associations (Battery Association for Supply Chain; JAPIA). Federated storage with blockchain-based verification and smart contracts; funded by member fees plus transaction fees; accredited by IPA as a “public interest digital platform operator”; built on ODS-RAM rather than IDS-RAM. The bootstrapping move is a deliberate sequencing of services: collaborative (non-competitive) services first — regulatory compliance data that every member needs and none competes on — then competitive services once the infrastructure is already in place and trusted.
The external forcing function is the EU Battery Regulation: Japanese suppliers must produce carbon-footprint declarations to keep EU market access. ⟨Inference: regulation does not remove the bootstrapping problem, it raises every anchor’s private expected return simultaneously, which is exactly what Frischmann’s test requires. This is the cleanest available argument for why compliance-driven data spaces start faster than value-driven ones.⟩
Anchor whose returns were real but slow: BoostAeroSpace
Founded 2011 by European aerospace OEMs; reached financial viability only around 2019 — roughly eight years. The anchor OEMs’ own returns justified their investment, so level 1 cleared; smaller suppliers largely free-rode, so level 2 never fully cleared for them. The report’s point is about funders, not about the consortium: the time horizon to viability is “far longer than the two-to-four-year grant cycles that typically fund data space initiatives”. A programme evaluating this case at year three would have recorded a failure.
No anchor, public funding instead: Agdatahub
French agricultural data intermediary, founded 2017 (Agritrust platform from around 2020), funded under France 2030, raised EUR 4.8 million in 2023 with Avril among investors, targeting 380 000 French farms via the API-Agro platform. Entered bankruptcy proceedings in November 2024. The OECD’s verdict: “the business model proved unsustainable without continued government funding”.
Read against Frischmann’s test, this is the predicted outcome rather than a surprise: no participant group’s private expected return cleared its own threshold, so the collective return — which was plausibly positive — never got financed by anyone but the state, and the state’s financing was finite. See also data space business models.
No neutral anchor at all: Covisint
GM, Ford and DaimlerChrysler, 2000, later joined by Renault and Nissan. Suppliers read it as buyer leverage over pricing and inventory rather than neutral infrastructure and resisted participation; the auction business was sold, the OEMs withdrew, and the exchange model was abandoned by 2004. “The failure was not technological but rather institutional.” This is simultaneously a bootstrapping failure and the report’s demonstration of boundary condition 2 — see data space boundary conditions.
Summary
| Case | Anchor | Private returns clear threshold? | Outcome |
|---|---|---|---|
| Finnish MDS | State-owned (Fintraffic) | Mandate substitutes for threshold | Operational, measured outcomes |
| ABtC | 14-firm coalition + regulation | Yes — EU market access | Operational, fee-funded |
| BoostAeroSpace | OEMs | Yes, over ~8 years | Viable 2019; suppliers free-ride |
| Agdatahub | None; public funding | No | Bankrupt, Nov 2024 |
| Covisint | OEMs, not neutral | Anchors yes, suppliers no | Abandoned by 2004 |
The architecture response: detachability
ODS-RAM’s three loosely coupled, independently replaceable pillars — ontology/semantic interoperability, addressability/discoverability, identity and usage control — allow participants to adopt one pillar without the rest (“detachability”, Tsuda 2026). The report ties this directly to bootstrapping: lowering the minimum viable commitment lowers each participant’s level-2 threshold. The contrast is IDS-RAM/DSP, which mandates the full stack including a contract-negotiation layer even where the policy is fully permissive — “binding usage control to a single, institutionally derived protocol can create a bottleneck for market adoption”.
⟨Inference: modularity is here being used as an economic instrument, not an engineering preference. That reframing is worth carrying into any architecture argument SINE makes — the question “must a participant adopt the whole stack to get any value?” is a bootstrapping question.⟩
The same structure outside data spaces: circular infrastructure
The club-good argument above is made about data-space infrastructure by the OECD. The EEA’s circular-economy financing report makes the same argument about physical circular infrastructure, in its own vocabulary and without reference to data (§3.2.2):
“Circular infrastructure, such as recycling, displays characteristics of public goods. These developments require collective investment but generate dispersed benefits, diluting private incentives to invest.”
“the knowledge generated by pioneering circular economy firms often spills over to competitors, reducing first-mover incentives.”
Two things follow. First, the structure is not an artefact of data: wherever the asset is shared and the benefit is dispersed, the same provision problem appears. Second, the EEA’s version is the Frischmann test failing in the circular-economy case — no subset of participants captures enough private value to justify the investment alone, which is exactly the condition under which bootstrapping binds. The report states the condition and never names it as such.
The EEA’s own answer is public money used catalytically: blended finance, InvestEU, subordinated finance, EIB de-risking (153 projects, EUR 5.1 bn, 2020–2024, screened against the EU Taxonomy). ⟨Inference: in this wiki’s case taxonomy that is the Agdatahub pattern — no anchor whose own returns clear its threshold, public funding substituted for one. The OECD’s reading of that pattern is that it postpones the problem rather than solving it, because the provision question returns when the subsidy ends. Worth saying out loud when the EU’s circular-finance enabler list is presented as a solution.⟩
⟨Inference: the EEA’s regulatory lever — EPR fee modulation, taxonomy extension, GPP criteria — is closer to the ABtC pattern (regulation moves each participant’s expected return) than to a subsidy, if the obligation attaches to an outcome each participant must deliver. The distinction matters: an obligation merely to offer something, as in the German Mehrweg-Angebotspflicht, moves no participant’s expected return and did not shift the market.⟩
Source backing ⟨2026-11-06⟩. That last sentence was an inference when it was written. POLYPROBLEM 06 (2023, p. 24) states it as a finding. On roughly 800 t/day of single-use take-away packaging waste in Germany:
daran hat die zum Jahresbeginn 2023 gesetzlich verordnete Mehrweg-Angebotspflicht nichts geändert
The report draws the same conclusion about what would be needed instead — politics “muss also einen Schritt weiter gehen” — and the instrument its interviewees converge on is a noticeably large single-use levy whose proceeds subsidise return and washing logistics. ⟨Inference: that is an ABtC instrument in the strict sense: it does not fund the infrastructure out of goodwill, it changes the relative price each participant faces, which is the only thing that moves an expected return. The same report supplies a second ABtC-shaped instrument — Markus Helftewes (Der Grüne Punkt, p. 13) asking for “langfristige Abnahmegarantien der Industrie” alongside better state funding, i.e. a demand-side guarantee that makes the recycler’s re-equipment investment bankable. Neither is a subsidy to the shared asset; both work on the cost-bearer’s expected return.⟩
A third observation from the same source sharpens why the provision problem binds here. Both sides of the reuse-infrastructure argument — Markus Urff (ReFrastructure), who wants a commons, and Fabian Barthel (Vytal), who wants interfaces only — share one constraint (p. 25): “Weder für eine gemeinnützige digitale Infrastruktur noch für ein Netz von Rücknahmestationen … existiert ein tragfähiges Geschäftsmodell. Niemand verdient mit der Rücknahme.” ⟨Inference: the financing objection does not discriminate between the two architectures. Choosing interoperability over a commons lowers the cost of agreeing, not the cost of running — so it is not an answer to bootstrapping, only to negotiation.⟩
The meta-level: bootstrapping the standards themselves
Annex A (Expert Workshop, Chatham House Rule) records a “reinventing the wheel” problem: consortia repeatedly re-debating technical components from scratch. Baron, Ménière and Pohlmann (2014) on standardisation consortia supply the mechanism — duplicative R&D and delay from vested interests. The report then names the instance directly: IDSA, Gaia-X, the DSSC and sector consortia each developing partially overlapping building blocks face “analogous risks”. The DSSC is flagged as illustrating the same tension in its own financing — a co-ordination body funded through project grants.
Cross-references
- OECD, Data Spaces between Co-ordination Challenges, Financial Viability and Public Policy (2026)
- Data space boundary conditions — the prior test
- Data space business models and viability — the downstream one
- Coordination failure vs. bargaining failure
- Data Commons · Ostrom’s commons design principles — free-riding and provision problems in the Ostrom frame
- Catena-X · IDSA · Gaia-X
- Digital Product Passport — regulation as the forcing function in the ABtC case
- Unlocking the circular economy (EEA, 2026) — the same provision problem for physical circular infrastructure
- Circularity valuation gap · Ecosystem financing
Sources
sources/research/2026-10-27-oecd-data-spaces-co-ordination-viability-public-policy/2026-10-27-oecd-data-spaces-co-ordination-viability-public-policy.md
(§3.4, §3.3.2, Boxes 3.3, 3.4, 3.5, 4.1, Annex A).
sources/circular-finance/2026-11-05-eea-unlocking-the-circular-economy/2026-11-05-eea-unlocking-the-circular-economy.md
(§3.2.2, ch. 4, Box 4.1). EEA Report 02/2026, doi:10.2800/5422329.