{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"deadweight_loss_reduction__data_science","archetype_slug":"deadweight_loss_reduction","domain_slug":"data_science","title":"Guardrailed borrowing of stranded team compute quotas","opportunity_summary":"Temporarily lend compatible, unprotected accelerator or cluster capacity from underused team quotas to eligible queued workloads, with minimum guarantees, recall rights, eligibility controls, monitoring, and rollback. The opportunity is conditional on first demonstrating meaningful overlap between idle compatible capacity and runnable demand blocked specifically by quota ownership.","adopter_authorizer":"The institutional compute-governance body is the authorizer; platform or cluster operators would implement and halt the mechanism, while security, privacy, data-residency, and infrastructure-risk owners retain mandate-specific vetoes.","scores":{"meaningful_impact":{"score":4,"rationale":"If the stated simultaneous overlap exists, the mechanism could convert otherwise idle compatible compute into completed priority-weighted analytical work and reduce queue delay without purchasing additional hardware. Magnitude remains unknown because overlap prevalence and workload value are unsupported."},"stakeholder_pull":{"score":3,"rationale":"The proposal identifies data scientists, research teams, platform operators, governance bodies, and downstream users with plausible interests in shorter queues and higher utilization, but supplies no evidence of experienced demand, adoption requests, or willingness to accept interruption and governance tradeoffs."},"incremental_advantage":{"score":4,"rationale":"Relative to fixed non-borrowable quotas, revocable borrowing directly addresses stranded capacity; relative to an unrestricted work-conserving priority scheduler, it adds minimum guarantees, eligibility restrictions, recall rights, incidence review, and halt conditions. Actual incremental performance has not been measured."},"distinctiveness_plausibility":{"score":3,"rationale":"The guardrail composition is clearly distinguished from the stated baseline and nearest rival inside the packet, but prior-art status is explicitly unsearched, so external distinctiveness cannot be affirmed."},"technical_implementability":{"score":4,"rationale":"The intervention is bounded to compatible, checkpointable, non-production jobs in one pool and uses identifiable scheduler controls such as caps, guarantees, recall, monitoring, and disablement. Heterogeneous hardware, preemption loss, locality, isolation, and instrumentation create material integration risk."},"adoption_authority_feasibility":{"score":4,"rationale":"The packet assigns authorization, enforcement, veto, review, and rollback roles to recognizable institutional bodies. Feasibility is reduced by the need for cross-owner approval and locally accepted service, concentration, interruption, and security thresholds."},"evidence_readiness":{"score":4,"rationale":"The candidate supplies separate problem and intervention falsifiers, eligible pilot boundaries, comparison periods, unaffected pools, outcomes, and rollback conditions. Matching variables, estimand, spillover controls, and quantitative decision thresholds remain unspecified."},"safety_net_benefit":{"score":3,"rationale":"Minimum guarantees and access-concentration limits could protect quota owners while allowing teams with blocked workloads to use slack, but the distribution of benefits is unknown and strategic demand or capture by advantaged teams could worsen access incidence."},"scalability":{"score":3,"rationale":"The mechanism could in principle be repeated across pools, but compatibility, checkpointability, data-residency, isolation, workload priorities, and reservation purposes vary by pool and institution, limiting straightforward replication."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"A bounded retrospective workload-log study and offline scheduler replay for one pool, including data extraction, compatibility classification, quota-block attribution, matched comparisons, outcome definition, and governance review of falsifiers.","confidence":"MODERATE","assumptions":["Existing workload, quota, utilization, queue, and failure logs are accessible and sufficiently linkable.","The step remains offline and does not alter scheduling or expose production workloads.","A platform engineer, data analyst, and governance or security reviewers contribute limited time.","No major new telemetry system or paid external data access is required."]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Engineering and validation of capped revocable borrowing in one pool, including scheduler integration, eligibility enforcement, guarantee and recall logic, observability, security review, testing, and rollback controls.","confidence":"LOW","assumptions":["The existing scheduler exposes extensible quota, priority, and preemption controls.","Only compatible, checkpointable, non-production jobs are eligible.","No hardware purchase or estate-wide scheduler replacement is required.","Local isolation and data-residency controls can be reused rather than rebuilt."]},"operational_launch":{"band_2026_usd":"50K_TO_250K","scope":"Preparation and operation of a time-bounded one-pool evaluation, including preregistration, operator coverage, incident response, user coordination, contemporaneous comparison, outcome analysis, governance review, and automatic sunset.","confidence":"LOW","assumptions":["Startup engineering has already passed offline and staging validation.","The launch remains capped and excludes production and emergency-reserved capacity.","Existing operations staff and monitoring platforms support the evaluation.","Threshold definition and approvals do not require prolonged institutional process redesign."]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Ongoing operation for a limited number of pools, including scheduler maintenance, monitoring, policy and eligibility review, access-incidence audits, security and reliability review, user support, and periodic causal reassessment.","confidence":"LOW","assumptions":["Deployment remains within one institution and uses existing infrastructure.","Recurring cost excludes hardware acquisition and major scheduler replacement.","Eligibility rules and priority weights require periodic human governance.","Expansion to heterogeneous or regulated pools would increase the band."]}},"research_burden":"MODERATE","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"YES","reason":"The candidate states a recognizable operational problem: compatible unprotected capacity can be idle while runnable workloads remain queued specifically because team quotas are non-borrowable. Whether this occurs meaningfully at a target institution still requires measurement."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"Institutional compute governance is identified as authorizer, platform operators as implementers and halt owners, and security, privacy, data, and infrastructure-risk owners as mandate-specific veto holders."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The proposal claims that capped, revocable, guardrailed borrowing will improve wait time or completed priority-weighted work over fixed quotas while preserving predeclared guarantees, a claim separable from both the baseline and an unrestricted work-conserving scheduler."},"bounded_next_evidence_step":{"status":"YES","reason":"One pool and a defined historical window can be examined offline for compatible idle capacity and quota-blocked runnable demand, with fixed quotas compared against a replayed guardrailed-borrowing policy and the stated absence-of-overlap falsifier."},"no_unresolved_safety_or_authority_stop":{"status":"UNCERTAIN","reason":"The packet supplies exclusions, veto roles, halt conditions, restoration, and sunset, but local approval and measurable thresholds for security, reliability, interruption, protected service, and concentration have not been established."},"implementation_cost_scope_and_range":{"status":"UNCERTAIN","reason":"The functional scope is bounded, but the sealed candidate provides no local scheduler architecture, telemetry quality, integration effort, compliance workload, or staffing evidence sufficient to validate a resource range."}},"blocking_evidence":["Workload-level evidence that meaningful simultaneous overlap exists between idle compatible unprotected capacity and runnable jobs delayed specifically by quota ownership.","Operational definitions and locally approved thresholds for material improvement, minimum access, protected-workload service, interruption, reliability, security, and access concentration.","Offline evidence that eligible jobs can be safely checkpointed, recalled, isolated, and matched to slack capacity without violating residency or reservation constraints.","A preregistered estimand, matching or adjustment variables, outcome windows, contamination and spillover treatment, and decision rule for the proposed comparison.","Evidence that modeled gains concern completed priority-weighted work or wait time rather than nominal utilization alone.","External prior-art research before making any distinctiveness or novelty claim."],"next_evidence_step":"Using four recent weeks from one candidate pool, conduct an offline workload-log audit and scheduler replay comparing observed fixed-quota operation with the proposed capped borrowing rules. Classify compatibility, protection, checkpointability, and the reason each job waited; proceed toward an authorized pilot only if the audit finds a predeclared meaningful overlap of idle compatible unprotected capacity with quota-blocked runnable demand and replay predicts material improvement in wait time or completed priority-weighted work without breaching proposed service, interruption, security, or concentration bounds. Falsify the problem if the overlap is absent or delay is explained by scarcity, incompatibility, data movement, maintenance, or dependencies.","research_questions":["How often, for how long, and at what scale do compatible unprotected resources sit idle while runnable jobs are blocked specifically by quota ownership?","Which workload, hardware, locality, residency, isolation, reservation, and checkpointing attributes determine safe borrowing eligibility?","What estimand and priority-weight definition can governance authorize, and how sensitive are conclusions to contested workload-value weights?","What minimum guarantees, recall latency, interruption, reliability, security, and concentration thresholds are acceptable to affected parties and veto owners?","Does offline replay predict improvements in completed work and wait time rather than utilization alone, and under what demand patterns?","How much checkpoint loss, request inflation, rebound demand, or strategic hoarding would offset the expected benefit?","Can matched periods or unaffected pools provide a credible comparison without contamination or spillovers?","What existing scheduling practices resemble this composition, and does the proposal retain any externally supportable incremental distinction?","What local engineering, monitoring, compliance, coordination, and recurring governance resources would implementation require?","Are the selected non-production workloads representative enough to support any later expansion decision?"] ,"recommendation":"VALIDATE_PROBLEM_FIRST","uncertainty_constraints":["The packet provides no external evidence of problem prevalence, stakeholder demand, realized effect size, or market scale.","Prior art is unsearched, so novelty, prevalence of similar mechanisms, and external distinctiveness are unmeasured.","Cost bands are resource-equivalent planning ranges based only on the stated scope; local scheduler architecture, data quality, compliance effort, and staffing are unknown.","The proposed success and halt concepts lack locally validated numerical thresholds.","The value and incidence of completed priority-weighted work depend on governance choices that may be contested.","Results from one compatible non-production pool may not generalize to heterogeneous, production, emergency-reserved, or regulated pools."],"closed_book_prior_art_boundary":"No conclusion is made about novelty, duplication, prevalence, market position, or existing implementations. The assessment recognizes only the packet's internal distinction among fixed non-borrowable quotas, unrestricted work-conserving priority scheduling, and the proposed guardrailed composition; external comparison requires separate prior-art research."}