Investigate cycle-count variance before adjustment

Cycle counting examines part of the inventory during normal operations, so its most damaging error is comparing snapshots from different times. Before sending a counter to the shelf, specify item, location, unit of measure and cut-off, and define who may move pieces during counting and how emergency movement is recorded. In a blind count the counter does not see the book balance first, reducing the temptation to round toward it. A blind count does not remove the need for a second count when variance is material. Record the physical quantity with location, time and counter identity, then freeze movements at that moment or reconstruct the ledger balance at count time. Calculate variance as counted minus book balance adjusted to the cut-off; write its sign and unit because 'five difference' does not say whether there is a loss or surplus. Do not aggregate two locations if the control concerns one bin; an unfinished internal transfer may disappear in the combined total. When variance appears, use a short investigation ladder. First verify item, pack and location, and check whether staging or quarantine pieces were omitted. Second review recent receiving, shipping and transfer events for an action before cut-off posted after it, or one posted twice. Third have another person recount when warranted, retaining the first result as evidence rather than deleting it. Fourth classify the cause: count error, late posting, unit conversion, location mix-up, damage, or unresolved cause. If a missing posting is supported by an event record, correct the event through the authorized workflow and recalculate the variance. Do not add a separate adjustment that repeats the same effect. If the cause remains unknown, an inventory adjustment should not be used to hide uncertainty. Protect disputed quantity from new promises under local operating rules and escalate to the person authorized to approve an adjustment. Inventory accuracy itself needs a precise definition: percent of exact-match locations, value variance and unit variance are different metrics. An improved percentage immediately after adjustment does not prove that the cause was removed. At closure connect cause and preventive action to a later test, such as retraining dock transfers and recounting that location a week later. Record who approved an adjustment and when, while separating that approval from the specialist's investigation role. The purpose of counting is not to make a screen look tidy but to make replenishment and promises rest on stock that can actually be found.