Why your stock count never matches the system

Your stock count does not match the system because two places accept stock movements and neither of them owns the number. The shelf and the paper note record one version, the system records whatever someone remembered to type later, and both are treated as records. The gap widens quietly until the count becomes a monthly argument instead of a control. It closes when one ledger records every movement at the moment it happens and refuses a movement that has no source and no destination.

Where does the difference actually come from?

Rarely from theft. Theft is the explanation people reach for because it is dramatic, and it hides four ordinary causes underneath.

The unit of measure changes hands. You buy in sacks and sell in kilos. You receive boxes of twelve and sell single bottles. Someone converts at the shelf, in their head, with a different rounding rule each time. The system holds one unit and the paperwork holds another, so the difference is created at the moment of conversion, not at the count.

Returns go back to the shelf without going back into the system. A customer rejects a delivery, the driver puts the carton back in the van, and the carton is on the shelf by the evening. Nobody writes it down because no money moved. The system still believes it was sold.

Transfers are recorded once, or twice. Between two shops, or between a shop and the depot, the outgoing side is confirmed and the incoming side is not. The item then exists in neither place, or in both, and nobody notices until a customer asks for it.

Damage is absorbed silently. A broken bottle, an expired carton, a sample given to a good customer. Each is too small to justify a form. Together they are usually the largest single line in the gap.

Name which of the four produced last month's difference and you have a fixable problem. Until then you have a feeling.

Why does a physical count not fix it?

Because a count is an event and the drift is a process. A full count tells you the truth on one day, and the truth stops being true the moment the first movement of the following morning goes unrecorded. Teams that count once a year spend the next eleven months working from a number that is decaying, and they know it, so they stop trusting the system and keep a parallel sheet. The parallel sheet is where the next discrepancy is created.

What is cycle counting, and when is it worth the trouble?

Cycle counting is counting a small, rotating part of the stock often, instead of counting everything at once. You count fast movers and high-value lines weekly, the rest on a slower rotation, and you treat each difference as a question rather than an adjustment.

It is worth it under one condition: somebody has the authority to stop and investigate a difference instead of writing it off. Without that authority, cycle counting is a full count spread thinly, and it produces the same argument in smaller pieces.

Approach	What it catches	What it costs
Annual full count	The size of the gap, once	A stopped business, no cause
Cycle counting	The cause, while it is fresh	A weekly routine
Counting when it hurts	Nothing you can act on	The same loss, later

There is a fourth row nobody writes down: no count at all, and reports running on numbers no one believes. It is the most common approach, and the most expensive.

What must the system record for a count to mean anything?

Four things have to exist before the count is worth taking.

1. One base unit per item, and a written conversion. If you buy by the sack and sell by the kilo, the conversion factor is a field, not a habit. The count is entered in the base unit, always.
2. A type on every movement. Receipt, sale, transfer out, transfer in, return, damage, adjustment. A movement with no type is a number with no meaning, and a month later nobody can explain it.
3. A source and a destination on every movement. A location, a document, a person. "Warehouse" is not a destination when you have three of them.
4. An adjustment that is an entry, not an overwrite. If correcting the stock destroys the old figure, you lose the only evidence of why it was wrong.

The fourth point is the one people resist, because typing the right number over the wrong one is faster. It is also the one that turns a count into a control, because the reasons accumulate and the pattern shows.

What do you fix first?

Do these in order. Each makes the next cheaper.

1. Pick the ten items that cause most of the argument. Not the most valuable — the most disputed. Record every movement on those ten for two weeks, on paper if you must.
2. Compare the handwritten movements with the system entries. What is missing from the system is the process you have to fix. It will cluster: one shift, one person, one kind of transaction.
3. Give each missing movement a home. A screen, a field, or a printed slip entered the same day. If the fix needs a new screen, say so: that is a software requirement, not a discipline problem.
4. Count, but only after that. Count the ten items weekly, in the base unit, and investigate every difference the same day.
5. Extend the rotation only when the ten hold. If they do not, the process is still leaking, and a wider count produces a wider version of the same argument.

Where this goes wrong

• The count is used to punish. The first time a difference is deducted from a salary, every future count is negotiated in advance. People make the shelf match the paper before you arrive, and the count becomes fiction.
• Adjustments are made to close the month. One large correction that balances the ledger hides the leak for good.
• The conversion lives in one person's head. When that person takes leave, the count is wrong by a factor and everyone spends a week discovering why.
• Two systems both accept movements. Then you have two numbers and no owner — a separate problem that no counting routine solves.

What this does not cover

This is about the mechanics of a count agreeing with a ledger. It does not cover valuation or costing, or how your accounting rules treat stock — those change what the number means, not whether it is right. It does not cover warehouse layout or picking accuracy, which produce the same symptom from a different cause. And it assumes you can change a process. If you cannot, the honest first step is to write the discrepancy down and take it to whoever can.

We build the ledger side — one record of every movement, with a source, a destination and a reason — so that the count has something true to agree with.