What problem does it solve? Daily cycle counts routinely disagree with system on-hand across dozens of locations, and treating every gap the same either floods planners with noise or lets real losses quietly corrupt available-to-promise (ATP) and replenishment. This Skill separates tolerance noise from material variances, traces each gap to a likely cause, and decides whether to auto-post, recount, or escalate with an availability hold. ## Core Features & Use Cases - Tolerance-based classification: Applies percent, value, and ATP-criticality bands (with zero/negative on-hand guards) to split noise from real variance. - Cause tracing and disposition: Maps variance signatures (offsetting pairs, short-picks, unscanned returns, LPN mis-scans) to bin-to-bin transfers, adjustments, recounts, or process fixes. - Gated writes with idempotency: Auto-posts only within the planner's post limit, mirrors every WMS adjustment with its ERP variance posting, holds phantom ATP quantities, and checks for duplicate postings before any write. - Use Case: A day's cycle count posts 34 discrepancies including a $9,000 shortfall on a fast-moving SKU; the Skill auto-posts the 31 in-limit lines, tasks 2 recounts, and escalates the $9K line with a 300-unit availability hold instead of clearing it as a rounding fix. ## Quick Start Triage today's posted cycle-count discrepancies and tell me which lines to auto-post, recount, or escalate with an availability hold.
