# Global DPP Atlas research protocol

**Protocol 0.1.0 · 23 October 2026 · Initial evidence cut-off 23 October 2026.**

The study asks how fragmented Digital Product Passport requirements converge through shared semantics, verifiable evidence and business identity, and where interoperability remains unproven. The unit of evidence is a qualified assertion about a specific instrument, requirement, specification version, activity or result. Similar requirements, compatible exchange and recognised authority are separate outcomes.

## Design and status

Use a structured scoping review, comparative provision coding and subsequent artifact evaluation. This protocol is frozen alongside the first seed release. The existing dossier was collected before this protocol; this is not a preregistered systematic review. The initial dataset is a selected, machine-assisted research seed. Computational validation does not constitute independent legal review or replication of interoperability experiments.

Four research questions guide inclusion: where requirements align or differ; how concepts and specifications connect them; which evidence supports interoperability by dimension; and whether a versioned graph preserves traceability and historical answers. The paper will report measured results only after the corresponding evaluation is executed.

## Corpus and geographic scope

The initial corpus is the four evidence registers and cited-source index in this folder. The index is a discovery inventory, not proof that every linked source was fully inspected. Import selected claims with their recorded limitations, reopen priority authoritative sources and preserve the distinction between inherited dossier material and new retrievals. Record each source candidate as extracted or pending in the release screening log. Pending means not yet extracted, not excluded as irrelevant.

Maintain three evidence streams: official legal texts and authority records; specifications, standards and implementation evidence; and academic literature. Academic sources inform the framework and later related-work review; this seed primarily populates the first two streams. Subsequent search expansions must log the exact query, platform, date, language, filters, results screened and exclusion reasons. Do not invent a historical search log from the final bibliography.

Include selected countries and territories in Europe, Asia, the Americas, Africa and Oceania where the dossier identifies relevant material. Distinguish supranational scope, national law, subnational activity and EU-market exposure. EU member-state map joins represent the geographic scope of EU instruments, not a completed review of each national legal system. Do not automatically extend EU law to the EEA, Switzerland, the UK or exporters' domestic markets. Great Britain and Northern Ireland remain separate scope objects.

## Eligibility and source hierarchy

Include explicit DPP rules and specifications; adjacent digital identification, traceability or labelling instruments with a stated functional relationship; relevant semantic, credential, organisational-identity and authority mechanisms; and identifiable programmes or tests. A record must identify its subject, evidentiary purpose and source. Keep project geography distinct from institutional headquarters.

Exclude unsupported vendor generalisations, duplicate announcements that add no evidence, unrelated uses of the phrase product passport, and forecast dates presented as completed events. Preserve significant conflicting accounts for adjudication. A primary announcement can support what an organisation announced; it is not independent proof of deployment, universal compatibility or legal sufficiency.

Prefer gazetted or consolidated law for provisions; the issuing body's catalogue and normative artifact for standards; a pinned specification or repository release for technical details; and published methods, payloads and results for tests. Authority guidance can explain law but must be identified as guidance. Source access is recorded as opened, dossier-only, search snippet or unavailable. An opened catalogue is not full normative-text review.

## Extraction and coding

Assign stable IDs to entities and assertions. Legal extraction should retain instrument and provision, territory, product scope, responsible actor, required action, conditions, exceptions and date dependencies. In this seed, some of these details remain in descriptions and scope notes; do not infer field-level legal equivalence from those summaries.

Record adoption, publication, entry into force, applicability, technical availability, citation and planned milestones as different date kinds. Preserve day, month or year precision. A conditional start has no computed calendar date unless its trigger is established. Observation time is distinct from validity time. The seed stores events and point-in-time observations; it does not claim a complete reconstructed legal history.

Differentiate legal instruments, provisions, requirements, policies, standards, standardisation projects, committees, organisations, specifications, ontologies, initiatives, implementations, pilots, reports and certification schemes. Record status as a dated assertion or qualified entity summary. Published, OJ-cited, applicable and operational are not interchangeable statuses.

Every substantive assertion requires source IDs, a locator, a statement, claim type, evidence type, review state and limitations. Structural vocabulary entities such as sectors are researcher-defined classifications. Source access and claim strength are different variables: reopening a source does not verify every interpretation attached to it.

## Review and disagreements

The first release uses `machine_assisted_unreviewed` throughout. Parallel automated extraction is not independent human coding. Before paper findings are finalised, obtain a second qualified review of all material legal status, scope and deadline claims, plus a preselected stratified sample of remaining assertions. Record reviewer identity, date, decisions and reasons; report the actual reviewed denominator and agreement, not a planned sample as completed work.

Maintain a review queue prioritising legal dates, conditional triggers, translation uncertainty, unavailable full texts, draft-to-final transitions and reported interoperability. Resolve disagreements by retaining both interpretations until an adjudicated change explains the choice. A correction creates a new release and a change record; do not silently overwrite a cited release.

## Interoperability coding and later experiments

Assess product identification, syntax, semantic meaning, provenance and integrity, issuer identity, delegation and authority, access, lifecycle continuity and legal acceptance separately. Record exact systems and versions, operation, environment, date, evidence type, reporting independence and result. Explicitly state unspecified versions.

Use separate results for reported success, measured pass, partial, fail, inconclusive and not tested. A supplied announcement remains a reported result. Lack of accessible evidence is not a failed test. Alignment announcements, shared vocabularies and individual conformance do not establish successful exchange. No aggregate country or technology readiness score is calculated.

The planned cases are selected battery requirements across regimes, meaning preservation across chosen public schemas, and identity/authority checks against a declared policy. Later tests should include incompatible units, product granularity, missing authority, expired status and unlinked evidence. This seed contains documentary evidence, not newly executed ecosystem exchanges or legal mutual-recognition tests.

## Graph and geography rules

Use explicit predicates and qualified assertion nodes. Distinguish `uses_as_input`, `references`, `normatively_references`, `implements_version`, `tested_with` and `aligned_with_claim`. A relationship supported only as a research classification must carry that assertion type. Keep the regulatory evidence graph separate from product-instance data.

World-map exports contain geographic join codes and records, not invented borders or facility coordinates. Unlisted countries are not assessed. A country can contain several legal statuses and sectors simultaneously. Regional or subnational rows are not automatically propagated as countrywide duties. Global standards initiatives remain global records, not laws assigned to the country hosting an institution.

## Reproducibility and releases

Each release includes the dataset, generated schema, vocabulary and context, inputs, protocol, screening log, geographic join rules, map/timeline/graph/evidence exports, query results, validation report and checksums. Use stable entity IDs and new assertion IDs when substantive claim meaning changes. Keep edition/version in the entity; create a new entity for a materially distinct specification edition rather than replacing its version string.

The build must fail on duplicate IDs, missing references, unsupported enum values, unsourced substantive records, impossible dates, invalid date precision and mismatched geographical references. Regression cases must cover missing evidence, conditional-date coercion, proposal/standard confusion, absent versus failed interoperability evidence, and future-event timelines.

Version 0.1.0 is an immutable local seed once sealed. Subsequent changes use a new release directory and a change manifest linking added, removed and revised records. Current pointers can move, but the paper cites an exact release. Preserve source hashes only for actual retained content; URL hashes identify sources and do not claim archived source bytes. Restricted source texts are not redistributed. Original data licensing should be settled before external publication; this local seed makes no new licence grant over third-party material.

## Amendment and completion log

2026-10-23: protocol 0.1.0 adopted for initial extraction from the pre-existing dossier. This is a local research protocol, not an externally registered study. Completion for this stage means a populated, versioned and computationally validated seed with explicit review limits. Human legal review, full-text comparative standard analysis, schema-mapping experiments and external publication remain later work.
