Cycle counting without stopping the warehouse

Once a year you shut the doors, everybody counts everything, and the number that comes out is treated as truth until next year. It is the most expensive count you will ever do and the least useful, for a reason that has nothing to do with effort: by the time you find a discrepancy, the trail that would explain it is months cold.

Cycle counting is the same work spread thin — a few bays a day, every day, forever, without closing. It is not a new idea and we did not invent it. What follows is the version that survives contact with a small operation, because the textbook version assumes a team you don't have.

Why the annual count doesn't fix anything

In the 482 posts we tagged as describing a real, unresolved problem, 13 were about counting. The complaint is almost never "counting is hard". It's that the count corrects the number without ever explaining it, so the same drift starts again the next morning.

A count that produces a corrected quantity and no cause is a receipt for a problem you're going to buy again.

The version that works: one bay a day

• Pick a bay. Not a SKU, not a category — a physical bay, the smallest unit somebody can stand in front of.
• Count what's in it, before looking at what the system says. Looking first turns a count into a confirmation.
• Then compare. Write both numbers down, not just the difference.
• If they disagree, spend five minutes on why — and stop at five whether or not you found it.
• Correct the system, and write the cause next to the correction even when the cause is "unknown".

Twenty minutes a day gets a small operation through its whole rack in a few weeks, which means every bay gets looked at several times a year instead of once. That frequency is the entire point: a discrepancy found within a fortnight still has a trail.

Count the bays that move, more often

Not every bay deserves the same attention. The ones that get picked from every day drift fastest and are worth counting monthly; the slow ones can wait a quarter. If you have no data on which is which, the person who picks knows off the top of their head, and that answer is good enough to start.

The part everyone skips, and it's the part that works

Writing down the cause. It feels like bureaucracy and it's the only step that changes anything, because after a month of causes you stop having a counting problem and start having a list of four specific things that go wrong.

In the accounts we tagged, the same cause comes up more than any other: the assigned bay was full, the box went somewhere else, and there was nowhere to record that which anyone would read.

That is not a counting failure. It's a put-away failure that a count discovers three months late — which is exactly why counting more often is worth doing even though counting isn't the problem.

Never punish a variance

The moment a discrepancy gets someone told off, discrepancies stop appearing. They don't stop happening — the bin is still full, the truck still arrives — you just go blind, and the count becomes a ritual that produces agreeable numbers.

If you take one thing from this: the person who reports the variance is doing the job correctly. That has to be true out loud, more than once, before anyone believes it.

What you can't fix by counting

If the system doesn't record which shelf something is on, counting tells you the quantity is right and nothing else. You'll count a bay, find what you expected, and still not be able to find the item that afternoon — which is a different problem, and we've written about how to measure it.

And a caveat about all of the above: our data is 4,146 tagged posts from people who went online to complain. It skews small, skews US, and skews toward unresolved problems, so it's a record of what goes wrong rather than a measure of how often. 59% of the people giving advice in those threads were vendors or consultants rather than operators — including, to be fair, us.