How to cycle count without shutting the warehouse

Sort SKUs by annual movement multiplied by unit cost. The top band is usually a small share of lines and most of the value; the tail is most of the lines and very little value. That distribution is what makes rolling counts affordable.

• A workable starting split: top band counted monthly, middle quarterly, tail annually.
• Add anything with a history of shrinkage or repeated variance to the top band regardless of value.
• Count new lines in their first month; early errors compound.

Convert the ranking into a count-per-day figure and check it against the time your team really has. A schedule that needs three hours a day will be abandoned in week three; twenty minutes a day survives.

• Work in locations per day rather than SKUs per day — that is the unit of work the counter experiences.
• Put counts at the quietest point of the day, usually before the first dispatch wave.
• Name the person, not the role. Unowned counts do not happen.

Send the counter to a location and ask what is in it, without showing the expected quantity. Blind counting is the difference between a count and a confirmation — if the expected figure is on screen, a counter who finds nearly that many will record exactly that many.

• Scan the location barcode to start, scan each product, enter what you counted.
• Record zero explicitly when a location is empty; an unrecorded empty location is indistinguishable from one not yet counted.
• Flag stock found in a location the system did not expect rather than moving it silently.

Stock moves while you count, and that is fine as long as you account for it. Compare the count against the location's movements between the count start and the submission, and adjust for anything picked or put away in that window.

• Short counting windows make this easy: a location counted in two minutes rarely has a concurrent pick.
• If a location is being actively picked, skip it and return later rather than fighting it.

A net variance near zero can hide two large offsetting errors. Group variances by cause — mis-pick, mis-receipt, unrecorded damage, case-versus-unit confusion, theft — and fix the process that produced the biggest group.

• Repeated variance on one SKU is usually a pack-size or barcode problem, not shrinkage.
• Variance concentrated in one zone is usually a labelling or layout problem.
• Variance concentrated on one shift is a training conversation, held privately.

Track the percentage of counted locations that matched on the first count. That single number tells you whether the process is improving, in a way that a total adjustment value does not.

• Report it weekly, by zone, so improvements are attributable.
• Expect it to fall when you start — you are now measuring errors that were previously invisible.
• Once it is consistently high in a zone, you can safely reduce that zone's count frequency.