Cycle counting without closing the warehouse

Stock records go wrong one small event at a time: a unit put away in the neighboring bin, a damaged item thrown out without a record, a short pick nobody reported. A yearly physical count corrects the totals once, but the errors start building again the next morning, and in the meantime orders allocate against stock that isn't there.

Cycle counting spreads the work across the year. You count a slice of locations every day, fix what's wrong while the cause is still fresh, and keep inventory accuracy high all year rather than for one day. The hard part is doing it without stopping picking, and without letting counters simply confirm the number on the screen.

Counts are location-first. You choose which locations to count, and NextStock creates one count task per location, sorted in walk sequence so the counter follows the physical route instead of zigzagging across the building.

• The counter scans the location on the floor app, then scans and counts what is there. Tracked stock is counted by lot or LPN where that applies.
• The expected quantity is never shown. The counter reports only what is on the shelf.
• If the first count differs by more than your tolerance, NextStock creates a recount task, given to a different counter when one is available.
• Small differences inside your tolerance post on submit. A larger difference that survives the recount waits for a manager.
• Approved variances post to the inventory ledger as adjustments with a reason, so the item's history shows exactly what changed and why.

Most warehouses can't stop picking to count. NextStock uses a soft freeze instead of a lock. While a bin is under count, allocation skips it, so no new orders are promised from it. A picker who still needs that bin sees a warning and can proceed.

The count stays correct because the expected quantity isn't a snapshot of on-hand at the moment you post. It is the quantity when the task started plus every ledger movement since. For example, bin A-01-02 holds 24 units of OIL-008 when the task opens. A picker takes 4 during the count, and the counter finds 20. Expected is 24 − 4 = 20, so there is no variance, and the pick is never mistaken for shrinkage.

Count problems usually come from the process, not the arithmetic. These are the ones to design out from the start.

• Showing counters the expected quantity. They will find it, whether it's there or not.
• Adjusting on the first mismatch. A recount catches miscounts and stock that was moved a minute ago.
• Counting only by SKU total. A SKU can be right in total and wrong in every bin, which still breaks picking.
• Skipping the reason. A variance approved with no cause recorded will happen again.
• Counting in random order. Following the walk route lets each counter cover more locations per hour.

A common approach is ABC classification: count fast-moving, high-value items more often than slow movers. Many teams also count a location whenever something suggests a problem, such as a short pick, a bin that shows zero, or an unexpected negative adjustment. The cycle counting guide walks through setting frequencies and sizing daily count volume.

NextStock tracks count accuracy as a KPI alongside fill rate and dock-to-stock time in reports and alerts. Watch the trend rather than a single day. A falling number concentrated in one aisle usually points to a putaway or labeling problem, not a counting one.

Cycle counting suits any warehouse that can't afford to close for a count: brands shipping every day, 3PLs whose clients expect accurate stock reports, and distributors with thousands of bins. For a 3PL, every counted line carries its client, so variances are reported against the right stock owner. If your stock is lot- or expiry-tracked, counting by lot catches the case where the total is right but the wrong batch is on the shelf.

Before your first count, label every location and set its walk sequence. Both make counting faster from day one, and both are needed for a blind count to be quick enough that people keep doing it.