How often should you cycle count?

ABC classification

ABC classification ranks stock keeping units by the value they move through the warehouse, usually annual usage value, and sorts them into three bands. Class A is the small set of high value or high velocity lines that the business cannot afford to be wrong about. Class B sits in the middle, moving less value or less often. Class C is everything else: low value, slow moving, or both, but often the largest share of SKUs on the racking by count.

The point of the split is not the label, it is that a discrepancy in a class A line and a discrepancy in a class C line are not the same problem. One threatens a stock out on something that drives revenue or keeps a production line running, the other is a rounding error you can live with for months. Counting effort should follow that difference, not be spread evenly across everything on the shelf.

Count frequency by class

Class A stock earns the tightest cycle: counted often enough that a discrepancy is caught within days, not months, because the cost of being wrong is highest here. Class B can tolerate a longer interval, counted often enough to catch drift before it compounds but without dedicating the same attention as class A. Class C can be counted least often under a manual regime, simply because the labour cost of counting it frequently is rarely justified by what is at stake if it drifts for a while.

That logic was built around a fixed, expensive counting method. Once counting stops being expensive, the argument for leaving class C alone for months gets a lot weaker.

A worked frequency table

Manual intervals are general industry practice and vary by sector and site. The automated column reflects that a recording unit reads every pallet label and every location it passes, class A or C, so frequency becomes a scheduling choice rather than a labour budget constraint.

What regulators and auditors expect

Auditors are generally less interested in how often you count than in whether you can show a consistent, documented method applied on a defined schedule, with discrepancies investigated and adjustments approved rather than posted automatically. A perpetual inventory system, where cycle counts continuously verify a subset of stock rather than relying on one annual wall to wall count, is widely accepted as meeting that bar, provided the schedule is actually followed and the records exist to prove it.

Regulated sectors add their own layer. Pharma and food sites in particular are expected to produce an audit trail behind every count, not just a final number. That is one reason evidence-backed counting, where every discrepancy line traces back to the record that produced it, sits well with compliance teams even when it is adopted for cost reasons first.

The trade-off between count cost and drift

Every counting schedule is a trade-off between two costs. Counting more often costs labour and disrupts operations. Counting less often lets drift accumulate between counts, the small errors from receiving, picking and put away that pile up unnoticed until a customer order fails or an audit turns up a number nobody can explain. The traditional answer is to count class A often enough that drift cannot go far, and let class C drift for longer because the downside is smaller.

That trade-off only holds while counting is expensive. If the cost of a count falls close to zero, there is very little reason to leave any class uncounted for months at a time, because the labour side of the trade-off has stopped being the constraint.

When a count costs 20 minutes instead of a day

This is where automated counting changes the arithmetic rather than just speeding it up. A hand scanner moves 50 to 100 units per hour, which is why a full wall to wall count needs three to ten people over two to three days and why most sites can only justify running one once or twice a year. A recording unit driven by a forklift the site already owns covers up to 3,000 units per hour, and a 20 minute run, driven during a normal break, covers 500 to 1,000 pallets.

Once a meaningful count fits inside a driver's break rather than a team's day, the question stops being "which classes can we afford to count often" and becomes "is there any reason not to count everything, every day". Sites running this way hold 99%+ stock accuracy year round instead of restoring it once a year. ABC classification stops being a rationing tool for scarce counting capacity and becomes what it was always meant to be: a way of deciding where to look first when a discrepancy turns up. The reasons discrepancies happen in the first place are in causes of stock discrepancies, the procedure a full count still follows is in how to run a warehouse stocktake, and continuous cycle counting covers how the runs are scheduled.