For online sellers the picture has an extra layer: e-commerce and ERP integration.
The bad news first: it is rarely theft
When stock does not reconcile, the first theory is always theft. Occasionally it is right. Far more often the cause is duller: the record and the shelf diverge because of a process, not because of anyone acting in bad faith.
That is bad news, because theft can be stopped by dismissing someone and a process has to be rebuilt. It is also good news, because a process gets rebuilt once and theft recurs.
Below are the causes in the order they actually occur.
Six causes, in order of frequency
1. The document is raised at a different time from the movement
The most common by a wide margin. Goods left in the morning, the despatch note was posted that evening. In between, the system shows stock that physically is not there, and a salesperson sells it a second time.
The symptom is easy to recognise: discrepancies cluster at month end and on busy days. If your stock problems have a seasonality to them, this is why.
2. One product exists under several item codes
The same cable is "Cable 3x2.5", "cable 3×2,5" and "3x2.5 cable (Manchester)". The total is right; every individual line is wrong. The picker takes from one, the salesperson sells from another.
This one almost always accompanies companies that grew out of spreadsheets, where items were entered by hand with no duplicate control.
3. No distinction between shipped and reserved
Goods are promised to a customer but still on the shelf. If the system does not separate those two states, available stock is overstated by the total of every promise outstanding. A company with a dozen live orders is continually selling what it has already committed.
4. Returns and mis-picks are never documented
A customer returns an item, the warehouse puts it back on the shelf, no paperwork exists. Or the wrong SKU was shipped and both sides agreed to "call it even" without recording anything. Each instance is a permanent discrepancy that will never resolve itself.
5. Write-offs are posted late and in bulk
Damage, spoilage, samples and internal use are written off once a month in a single adjustment. Until then the records show goods that do not exist. Worse, by the time it is posted nobody remembers the reasons, so the losses cannot be analysed.
6. Units of measure do not match
Received in cases, sold in units, and the conversion factor lives in the warehouse manager's head. Classic for anyone importing and repacking.
Finding your cause in a day
A full stocktake is not required. A sample check is enough.
Step 1. Take the ten items with the largest discrepancies. Not random ones — the problem lines. The cause is usually the same across all of them.
Step 2. Check each for duplicates. Search by part of the description and by code. If similar items appear, it is cause number two and you can stop looking.
Step 3. Pull the movement history for the last month. Compare document dates against the dates goods physically moved. A gap of a day or more is cause number one.
Step 4. Total what has been promised but not shipped. If that figure is comparable to the discrepancy, it is cause number three.
Step 5. Ask the warehouse team what they do with returns and damages. Not as an audit — as "walk me through how it works now." The answer is usually complete.
One day is enough to narrow six causes to one or two, and then it becomes clear what to fix.
Why the annual stocktake does not help
A full count once a year aligns the numbers for one day and changes nothing about the process. Within a week the discrepancies start accumulating again, at the same rate and for the same reason.
What works instead is cycle counting: a small group of items checked every week on a schedule, with high-value and fast-moving lines counted more often. The warehouse does not stop, discrepancies surface within weeks rather than after a year, and the cause can still be traced while the trail is warm.
This is an organisational decision and it can start tomorrow, even while you are still on spreadsheets.
What a system actually fixes
Honestly: a system does not fix discipline. If documents are raised the following day, they will be raised the following day in the new system too. But four of the six causes are closed by configuration.
| Cause | What closes it |
|---|---|
| Documents posted late | Scanning on the floor: the movement is recorded as it happens, not at the end of the day |
| Duplicate item codes | Uniqueness control on codes, and cleaning the master data during migration |
| Reservations not accounted for | Reservation logic: available stock is calculated net of what is promised |
| Returns and mis-picks | Dedicated document types with a mandatory reason code |
| Bulk write-offs | Write-offs with reason codes, and loss analysis by category |
| Units of measure | Conversion factors on the item card rather than in someone's head |
The decisive one is scanning, because it removes the gap between action and document — the most frequent cause of the six. How warehouse processes of different complexity are structured is covered in warehouse logistics, and cleaning the item master before a move is in the data preparation checklist.
Frequently asked questions
What size of discrepancy is normal?
It depends on the goods. For high-value discrete items the norm is zero: every discrepancy should have an explanation. For small fixings, fasteners or bulk materials a small variance is unavoidable. The real indicator is not the percentage but the trend — if the gap grows from count to count, the cause is systemic.
Would CCTV help?
It would, if the cause really is theft — but that is the rarest of the six. Process causes are worth ruling out first, because cameras cost money and do not fix duplicate item codes or documents posted a day late.
What does warehouse scanning cost?
The hardware starts at a few hundred dollars for a handheld terminal, and a smartphone with the mobile app costs nothing if you already have one. The real cost is in configuring warehouse processes and training people, not in the device.
Can we fix this without new software?
Partly. Cycle counting, cleaning duplicate item codes and the rule that documents are raised as goods move all work in any system, spreadsheets included. What does not work without a system is automatic reservation control and recording movement by scan.
Where should we start if the problem is old and has several causes?
With the item master. Duplicates distort every subsequent diagnosis: while the same product exists under three codes, you will not see the real picture and will look for the cause in the wrong place.