{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","cell_id":"backpressure__accounting_auditing","arm":"BREADTH_PROBE_ONE_SHOT","candidate_id":"backpressure__accounting_auditing__P1","proposal_index":1,"version":0,"title":"Capacity-signaled release of audit evidence requests","problem":"During an audit, multiple audit workstreams can issue evidence requests faster than the audited organization's evidence coordinators and control owners can retrieve, validate, redact, approve, and explain the requested material. Because requesters do not see this downstream handling constraint when creating new requests, unresolved requests accumulate, clarifications and duplicate submissions multiply, and aging high-risk requests can become obscured within the backlog.","actors":["Audit engagement lead","Audit workstream leads and staff who create evidence requests","Client audit liaison or evidence coordinator","Control owners and accounting staff who prepare responses","Client legal, privacy, or records reviewers where applicable"],"observable_state":"The evidence-request register shows rising open-request count, oldest-request age, clarification rate, duplicate-request rate, and work awaiting client validation, while the measured completion capacity of evidence coordinators and control owners remains below the incoming request rate. The pressure state is computed by priority class from bounded queue occupancy and queue-age bands, not from an individual's informal claim of being busy.","consequence":"The engagement can lose a reliable view of which evidence is genuinely outstanding; client staff may rush or repeatedly rework submissions; and material or deadline-critical evidence may wait behind lower-priority requests, threatening timely collection and evaluation of sufficient appropriate audit evidence.","affected_objective":"Maintain an orderly, traceable, and timely evidence-acquisition process that preserves attention for material and deadline-critical audit procedures without changing auditors' judgments about what evidence is required.","intervention":"Create priority-specific evidence-request windows controlled by downstream handling capacity. The client evidence coordinator publishes a regularly refreshed pressure state derived from queue occupancy, age, and recent validated-response throughput. When a priority class reaches its high-pressure threshold, audit workstreams must hold newly drafted requests for that class in their own visible staging lists and may release another request only when the coordinator acknowledges available capacity. Material, fraud-related, regulatory-deadline, and other engagement-lead-designated critical requests retain a reserved path. A lower release threshold and gradual recovery probe prevent rapid stop-start oscillation; every held request remains attributable, visible, and subject to an explicit latest-release date.","structural_mapping":[{"archetype_element":"Upstream producers","domain_realization":"Audit workstreams that formulate and release evidence requests."},{"archetype_element":"Downstream capacity-constrained receiver","domain_realization":"The client evidence coordination and control-owner network that must retrieve, validate, approve, and explain responsive material."},{"archetype_element":"Flow","domain_realization":"New audit evidence requests moving from audit workstreams into the client's response process."},{"archetype_element":"Observable downstream pressure","domain_realization":"Priority-specific queue occupancy, oldest-request age, recent validated-response throughput, and clarification workload."},{"archetype_element":"Upstream signaling path","domain_realization":"A pressure state in the shared request register that governs whether each workstream may release another noncritical request."},{"archetype_element":"Bounded accumulation","domain_realization":"A maximum number of released, unresolved requests per priority class, with excess drafts retained in visible workstream staging lists rather than silently entering another queue."},{"archetype_element":"Critical-flow preservation","domain_realization":"Reserved release capacity for material, fraud-related, statutory, and imminent-deadline evidence requests designated under engagement policy."},{"archetype_element":"Recoverability","domain_realization":"Hysteresis and limited recovery probes gradually restore ordinary release rates after the downstream queue drains."}],"mechanism_mapping":[{"mechanism_slug":"kanban_pull_systems_and_wip_limits","role":"The evidence coordinator acknowledges available response slots, so workstreams release noncritical requests only when downstream capacity is available; priority-specific work-in-progress limits bound the released backlog.","counterfactual_removal":"Without pull acknowledgments and work-in-progress limits, the register becomes a passive list while workstreams continue releasing requests independently of response capacity."},{"mechanism_slug":"bounded_producer_consumer_queues","role":"Each priority class has a bounded released-request queue; requests beyond the bound remain visible with the producing workstream and must be released, explicitly deferred, merged, or withdrawn.","counterfactual_removal":"Without a bounded queue, overload can be hidden in an ever-growing central register or displaced into untracked email and spreadsheet queues."},{"mechanism_slug":"reactive_streams_demand_signaling","role":"Downstream acknowledgments function as demand signals for additional noncritical requests, coupling upstream release to current evidence-handling capacity.","counterfactual_removal":"Without a capacity signal traveling upstream, any reduction in requests would be a static cap or discretionary pause rather than backpressure."}],"causal_chain":["Audit workstreams independently release requests faster than client evidence-handling capacity.","Released-request occupancy and age cross a priority-specific pressure threshold.","The shared register communicates that downstream pressure to the workstreams that produce requests.","Those workstreams stop releasing additional noncritical requests and retain drafts in attributable, visible staging lists, while reserved critical requests continue.","The inflow of released requests falls below or approaches the rate at which validated responses leave the queue.","Released backlog growth becomes bounded, allowing coordinators and control owners to complete and clarify admitted work.","After occupancy and age remain below the lower recovery threshold, limited release probes increase inflow gradually without reconstructing request state."],"baseline":"A shared request list records owners, due dates, and status, but each audit workstream may add requests whenever it chooses. Congestion is handled through meetings, reminders, deadline changes, and individual escalation after the backlog has formed; downstream load does not directly govern upstream request release.","nearest_rivals":["A static per-workstream request quota, which caps releases without responding to changing downstream capacity.","A larger evidence-response team, which expands capacity but does not couple request creation to observed congestion.","A dashboard or aging report, which exposes backlog but does not constrain upstream release behavior.","A centralized request buffer, which relocates or conceals excess work unless its state changes producer behavior.","Engagement-lead triage or cancellation after overload, which sheds or reprioritizes work after requests have already entered the response process."],"remaining_contrastive_claim":"The defining claim is limited to mechanism: priority-specific, observed downstream handling pressure changes the release behavior of the audit workstreams generating new requests. If the signal merely informs discussion, the cap is fixed in advance, or requests continue accumulating in another hidden queue, the intervention is not this candidate.","authority_safety":{"decision_authority":"The audit engagement lead retains authority over audit scope, evidence sufficiency, materiality, risk classification, deadlines, and exceptions. The client evidence coordinator may report capacity and administer acknowledged slots but may not veto, downgrade, or conceal an audit request.","authorized_first_step":"The engagement lead may authorize a shadow-mode replay using timestamps and classifications from one completed audit workstream to compare the proposed release rule with the actual request sequence; no live request, audit procedure, conclusion, or client obligation is changed.","excluded_actions":["Automatically suppressing, canceling, or downgrading evidence requests","Allowing the audited organization to decide which audit evidence is necessary","Deferring a statutory, material, fraud-related, or imminent-deadline request without engagement-lead approval","Changing audit scope, materiality, sampling, conclusions, or documentation requirements","Moving held requests into untracked email, personal notes, or undisclosed spreadsheets","Using pressure status to evaluate or penalize individual client employees"],"halt_rollback":"For any later live pilot, halt new throttling immediately and return to ordinary engagement-lead-controlled release if a critical request is delayed by the rule, a staged request becomes untraceable, a deadline is endangered, priority classification is disputed without timely resolution, or the pressure signal is shown to be stale or manipulable. Preserve the complete register and staging history so rollback requires no reconstruction."},"negative_tests":{"strongest_counterevidence":"A timestamped reconstruction shows that response delay is driven primarily by request ambiguity, unavailable records, approval dependencies, or a few intrinsically long procedures rather than by incoming request rate or concurrent released workload; lowering releases therefore would not relieve the constrained step.","problem_falsifier":"Across the selected workstream, incoming request rate and released queue occupancy do not exceed measured validation capacity before aging or rework appears, or request aging does not worsen when occupancy rises.","intervention_falsifier":"In shadow replay, the pressure state either fails to slow request release before the released queue crosses its bound or merely transfers an equal or larger unbounded backlog into workstream staging lists, while critical-flow and traceability safeguards cannot be maintained.","risks":["Signal lag could allow the released queue to overload before workstreams react.","Poor threshold or recovery settings could cause oscillation between request freezes and bursts.","Large staging allowances could create hidden queue growth and buffer bloat upstream.","Reserved critical capacity could be consumed by overclassification, producing priority inversion.","Low-priority but necessary requests could starve until too near the reporting deadline.","Conservative signals could over-throttle the audit and leave available response capacity unused.","Client-reported capacity could be inaccurate or strategically influenced, compromising signal integrity.","Protecting the evidence coordinator could shift overload to workstream staff who must maintain large staged inventories.","Reduced request concurrency could lengthen elapsed audit time even if downstream handling becomes more stable."]},"next_evidence_step":"Using only one completed audit workstream, reconstruct daily request releases, validated completions, queue occupancy, oldest-request age, priority, clarifications, and deadlines. Predefine one high threshold, one lower recovery threshold, a maximum staged-request count, and a reserved critical class; then replay the sequence once to determine whether the rule would have signaled before the observed backlog peak, kept both released and staged queues bounded, and avoided delaying any critical or deadline-bound request. Treat missing timestamps, unverifiable priorities, or any simulated critical delay as a failed first test rather than tuning the rule after seeing results.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed because this is an isolated one-shot breadth probe and no other proposals or experiment cells were inspected.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}