Why cycle counting never sticks — and what makes it last

The problem is never the method. It is always what no one remembered to decide before starting.

The purpose of cycle counting is not to count inventory. It is to detect the causes of discrepancies early enough to correct them.

That distinction changes everything. If counting is the goal, success is measured by the number of items counted, and the exercise becomes a volume chore that is sacrificed as soon as workload rises. If cause detection is the goal, success is measured by the number of discrepancies explained—and then the exercise becomes interesting to the people doing it.

The question to ask after every count is not ‘how many are there?’ but ‘why is the quantity wrong?’. That is a different conversation, involving different people.

What must be decided before the first count

Who counts—and who does not

Asking a picker to count the inventory they pick themselves is a poor idea—not because they are suspected of anything, but because their brain already knows the expected answer. They will not see what is wrong.

That does not mean a dedicated team is needed. It means rotation: everyone counts an area other than their own. In a medium-sized warehouse, this simple rule changes the quality of the discrepancies detected.

What happens when a discrepancy is found

This is the decision everyone postpones, and the one that kills the process.

There are three possible choices, and one must be selected explicitly:

Adjust the system inventory immediately to match the physical count. It is quick, but loses the trail to the cause and turns counting into a sticking plaster.

Block the location and investigate before adjusting. It is rigorous, but impractical if there are many discrepancies at the start.

Adjust above a threshold and investigate below it—or the reverse, depending on item value.

There is no universally correct answer. There is an answer that is written down, known to everyone and applied consistently by everyone. Without it, every counter improvises and the discrepancy data becomes unusable.

Who is authorised to adjust

Cycle counting without control over adjustment rights quietly becomes a machine for making problems disappear. Authorised people must be named, every adjustment must be recorded with its reason, and those reasons must be reviewed once a month. Not to police the teams, but to spot recurring causes.

Traditional ABC classification—by value—is a good starting point and a poor finishing point.

What creates discrepancies is not value; it is how often an item is touched. A low-cost item picked fifty times a day, loose, in a poorly lit location, will create infinitely more discrepancies than a valuable pallet that moves only twice a year.

Combine three criteria:

movement frequency—the criterion that best predicts discrepancies;

ease of confusion—similar references, identical packaging, items with no readable code;

value, last—to arbitrate, not to classify.

The items that rise to the top when these criteria are combined deserve frequent counts. Count the rest rarely. You may be surprised to find that your most troublesome items are not the ones you monitor.

Consistency beats intensity. Always.

In practical terms, reserve a short, fixed slot in the schedule outside peak picking hours. Tuesday and Thursday at 7 a.m. before the docks open, for example. A slot that can be maintained on a high-workload day is one that will survive the season.

And accept that you may cover fewer items than planned. A programme that covers only part of the inventory each quarter but has run for two years is worth more than an exhaustive programme abandoned after one quarter.

A WMS does not perform cycle counts. It makes them paperless and usable.

What the tool should do in practical terms:

propose locations to count according to the rules you have defined, without a team leader spending an hour preparing lists;

allow blind counting—the counter cannot see the expected quantity, otherwise they simply confirm it;

record the discrepancy together with its reason, date and author;

and above all, report history by cause so that after three months you can say: ‘most of our discrepancies come from the ground-level picking area, on items without individual packaging’.

That last function is what justifies the tool. Without it, a spreadsheet is enough—and spreadsheets get abandoned.

The real measure of success

It is not the inventory accuracy rate. It is the time between a discrepancy arising and its detection.

A warehouse that detects discrepancies within fifteen days on average can still act: find the delivery concerned, review the procedure and correct the labelling. A warehouse that discovers in January discrepancies caused by movements in June can do nothing but record the fact.

Measure that time. It is the only number that tells you whether the process is serving a purpose.