{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"layer_decay_and_expiration_management__data_science","archetype_slug":"layer_decay_and_expiration_management","domain_slug":"data_science","title":"Governed lifecycle management for stale analytical artifacts","opportunity_summary":"Evaluate a lifecycle service that inventories analytical artifacts, labels suspected staleness, applies retention and dependency gates, moves eligible artifacts out of active discovery through reversible disposition, and verifies restoration. The proposal addresses a coherent failure mode, but its prevalence, stakeholder demand, incremental effects, and differentiation from existing systems are unestablished.","adopter_authorizer":"Data or ML platform leadership could adopt the service; artifact owners could propose refresh or demotion, while data governance, legal, privacy, audit, and model-risk authorities would authorize contested, irreversible, held, or sensitive dispositions.","scores":{"meaningful_impact":{"score":4,"rationale":"If superseded artifacts materially remain in ordinary discovery, reducing obsolete reuse while preserving lineage and rollback could improve analytical decisions and control storage and maintenance burdens. The candidate provides no evidence establishing the frequency or magnitude of those harms."},"stakeholder_pull":{"score":3,"rationale":"The proposal identifies plausible beneficiaries and burden holders, including analysts, platform operators, and control functions, but supplies no interviews, incidents, workflow measurements, budget commitment, or adoption request."},"incremental_advantage":{"score":4,"rationale":"Relative to owner-by-owner cleanup and a catalog limited to badges and ranking, the proposed combination adds retention authority, dependency-gated disposition, reversible archival, restore drills, tombstones, and audit records. Whether these additions improve outcomes enough to justify their burden remains untested."},"distinctiveness_plausibility":{"score":2,"rationale":"The lifecycle composition is specifically differentiated from the stated nearest rival, but prior art is explicitly unsearched and its components resemble several named categories of lifecycle, catalog, registry, retention, and reproducibility systems. World distinctiveness cannot be inferred closed-book."},"technical_implementability":{"score":3,"rationale":"A non-production shadow pilot using inventory, labels, simulated decisions, archival, and restoration is technically bounded and plausible. Broader implementation depends on materially complete identity, lineage, dependency, policy, and ownership metadata that the candidate says may be missing."},"adoption_authority_feasibility":{"score":3,"rationale":"The candidate partitions authority by reversibility and sensitivity and names responsible functions. Actual policy authority, cross-functional agreement, owner responsiveness, and exception-handling capacity in an adopting organization are not demonstrated."},"evidence_readiness":{"score":4,"rationale":"The candidate supplies separate problem and intervention falsifiers, a 30-day reversible pilot, baseline comparisons, measurable failure conditions, and explicit halt criteria. Sampling rules, outcome thresholds, and current registry quality still require specification."},"safety_net_benefit":{"score":4,"rationale":"Quarantine, restore testing, holds, dependency blocks, tombstones, audit records, and rollback directly reduce the danger of premature retirement or deletion. Incomplete lineage, unreliable archives, and conflicts between quarantine and deletion duties remain material residual risks."},"scalability":{"score":3,"rationale":"A platform lifecycle service could apply common controls across datasets, features, models, notebooks, and reports, but semantic staleness varies by artifact class and lifecycle metadata, exceptions, dependency graphs, and restore obligations could themselves become costly at scale."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"Design and run the proposed 30-day shadow pilot for one non-production artifact class, including inventory, stratified sampling, stale-label review, simulated policy decisions, archive-and-restore tests, baseline comparison, security review, and evaluation.","confidence":"MODERATE","assumptions":["Existing registry and archive interfaces can be used without major platform construction.","The sample contains no production-critical actions and permits no hard deletion.","Data science, platform, governance, privacy or compliance, and evaluation labor are included.","No unusually costly external data licensing or regulated-environment validation is required."]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Productionize reversible lifecycle controls for one artifact class within one organization, including policy encoding, registry integration, dependency checks, discovery-label changes, audit records, access controls, monitoring, and restore procedures.","confidence":"LOW","assumptions":["A usable metadata catalog, identity system, lineage source, and archival tier already exist.","Deployment initially excludes automatic irreversible deletion and unusually sensitive artifact classes.","Material remediation of legacy ownership and lineage gaps is limited in scope."]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Launch governed lifecycle operations across several analytical artifact classes and production teams, including integrations, metadata remediation, policy and exception workflows, training, compliance validation, restore drills, rollout support, and outcome evaluation.","confidence":"LOW","assumptions":["The organization has multiple artifact systems requiring integration but not wholesale replacement.","Cross-functional governance and operational support are required.","Artifact volume, dependency completeness, regulatory obligations, and platform heterogeneity are unknown."]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Operate and maintain the lifecycle service, including policy updates, owner review, exception revalidation, metadata quality work, monitoring, archival storage and retrieval, restore drills, incident response, audits, and periodic effectiveness evaluation.","confidence":"LOW","assumptions":["One medium-to-large organizational deployment is in scope.","Recurring work includes both platform engineering and governance operations.","Exception volume and cold-storage retrieval demand remain manageable.","Major platform replacement and large-scale legal review are excluded."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"YES","reason":"The candidate specifies observable indicators—superseded artifacts marked active, uncertain ownership and dependencies, inconsistent retention, stale search results, and absent restore evidence—and provides a representative-audit falsifier. External evidence is still required to establish material presence in a target setting."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"Data and ML platform operators are identifiable adopters, artifact owners can propose reversible actions, and governance, legal, privacy, audit, and model-risk functions are assigned authority over sensitive or irreversible decisions."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The proposal claims that retention authority, dependency-gated reversible disposition, and tested restoration add value beyond owner-by-owner cleanup or catalog badges and ranking; the shadow pilot can compare search exposure, triage effort, false retirements, dependency issues, restoration, and exceptions."},"bounded_next_evidence_step":{"status":"YES","reason":"The sealed candidate authorizes a 30-day shadow pilot on one non-production artifact class with simulated decisions and stratified archive-and-restore testing, expressly excluding hard deletion."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"For the bounded evidence step, actions are reversible and non-production, hard deletion is excluded, authority is partitioned, and unresolved dependencies, policy conflicts, failed restores, audit gaps, or sensitive-data exposure require halt and rollback. This does not establish safety for broader deployment."},"implementation_cost_scope_and_range":{"status":"UNCERTAIN","reason":"The candidate defines the mechanisms and pilot boundary, but lacks artifact volumes, platform architecture, metadata completeness, integration count, regulatory context, staffing, and exception rates needed to support an implementation resource range beyond broad assumption-dependent bands."}},"blocking_evidence":["No representative evidence shows that superseded-but-active artifacts, stale reuse, or cleanup avoidance are materially present in a prospective adopter's environment.","No baseline measures establish stale-artifact search exposure, owner triage effort, dependency coverage, restoration reliability, exception volume, or storage and maintenance burden.","No pilot results establish stale-label precision, safe disposition rates, search improvements, restore success, or compliance performance relative to owner-by-owner cleanup.","No stakeholder evidence establishes willingness among platform leadership, artifact owners, or control functions to fund and operate the governance workflow.","Prior art is unsearched, so incremental distinctiveness from existing catalogs, registries, retention systems, lifecycle tools, and reproducibility platforms is unresolved.","Implementation feasibility and cost depend on unknown lineage completeness, durable metadata ownership, archival capabilities, policy integration, and artifact-estate scale."],"next_evidence_step":"Run the authorized 30-day shadow pilot on one stratified, non-production artifact class: first audit whether superseded-but-active artifacts and stale discovery are materially present, then compare simulated lifecycle labels and dispositions with ordinary owner-by-owner cleanup on stale search exposure, owner triage time, false-retirement rate, unresolved dependencies, policy exceptions, and archive restoration success. Falsify progression if the problem audit finds no meaningful accumulation or reuse, or if the intervention produces no improvement or unacceptable classification, dependency, restoration, or compliance failures; perform no hard deletion.","research_questions":["What proportion of a representative artifact class is superseded but still active or ordinarily discoverable, and how often is it reused as current evidence?","How much cleanup avoidance is attributable specifically to uncertain lineage, restoration, audit, retention, or rollback requirements?","Against owner-by-owner cleanup and catalog-only labeling, do lifecycle controls reduce stale search exposure or triage effort without unacceptable false retirements?","What fraction of artifacts has sufficiently resolved ownership, dependencies, retention status, and exceptions to permit reversible disposition?","Can stratified archived artifacts be restored within predefined completeness, integrity, and latency criteria?","How much ongoing labor is required to maintain lifecycle metadata, resolve exceptions, revalidate holds, and conduct restore drills?","Which existing lifecycle, catalog, model-registry, records-retention, and reproducibility offerings already implement the proposed composition, and what testable increment remains?","Do prospective adopters and authorizers accept the proposed division of authority, operational burden, and halt conditions?","Could quarantine conflict with privacy or contractual hard-deletion duties for any included artifact class?","At what artifact volumes, integration counts, and metadata-remediation levels do deployment and recurring costs change resource bands?"],"recommendation":"VALIDATE_PROBLEM_FIRST","uncertainty_constraints":["Closed-book assessment with no external verification.","Problem prevalence, harm frequency, stakeholder demand, market size, and realized impact are unmeasured.","World novelty and prior-art differentiation are unmeasured because prior art is explicitly unsearched.","Cost bands are resource-equivalent planning ranges, not point estimates, and depend heavily on estate scale, metadata quality, integration complexity, and regulatory context.","The favorable safety assessment applies only to the specified reversible, non-production evidence step and does not authorize live disposition or hard deletion.","Semantic staleness is artifact-specific and cannot be inferred from age alone."],"closed_book_prior_art_boundary":"No conclusion is made about novelty, prevalence, incumbent capabilities, or competitive differentiation. The candidate itself identifies metadata catalogs, model registries, records-retention systems, data lifecycle management, and reproducibility platforms as comparison categories, but supplies no search results; distinctiveness therefore remains an external-research question."}