Evaluate comparable performance and conditional improvement

Repair solution-performance-case.json. measure:{population,elapsed,success,unit,sourceRef} and eligibleIds define the agreed current closed operational evidence, including failed cases. cohorts:[{period,cohort,count,meanElapsed,successRate}] covers both source periods/cohorts. comparison:{profile,sourceRef,weights,observedBefore,observedAfter,standardizedBefore,standardizedAfter,improvement,aggregateTrend,comparableTrend,attribution} separates actual changed case mix from comparable fixed-weight trends; use source-diagnostics-only attribution. limitations:[{cause,category,owner,sourceRef,minutes}] grounds app, enterprise and external impact in confirmed fictional diagnoses. actions:[{id,units,sourceRef}] constructs a funded distinct intervention set. projection:[{case,elapsed,success,remaining:{app,enterprise,external}}] calculates every post case from real source residual fractions and recovery thresholds. report:{workUnits,meanElapsed,successRate,standardizedMean,residualMinutes:{app,enterprise,external},assurance:conditional-effect-model} reconciles actual proposed effects and meets mean<=17.5/successRate1 without removing failed cases. decision:{recommendTo:service-owner,claim:conditional-performance-improvement,approval:pending-actual-decision}. Focused app repair plus training or integrated validation can meet the goals under baseline or balanced comparison profiles; source provider backup is optional if funded. Save partial measures and repair a case-mix, scope or effect error without Reset. Independent record/action ordering is equivalent. No aggregate trend proves causality, and no modeled effect attests actual benefit or human approval.