How stocktaking and inventory sync actually work
Inventory mismatches are not accidents — they come from an undefined counting method and an unclear sync direction. Here is the daily order: who moves stock, how to count, how to handle variance, with a cause-and-action table.
The mismatch is usually a definition problem
Before fixing the numbers, find out why they disagree. Most inventory disputes are not theft and not carelessness — they are two systems using different definitions. One counts a product as available as soon as it is received, the other only after it is put away. Both are right. They just do not agree.
So the first step is not a recount. It is agreeing on the definitions.
Decide who is allowed to move stock
This is the structural decision, and everything else depends on it. There are two workable models:
- The warehouse system is the single source of truth for physical stock. Everything else reads from it.
- The ERP holds the count and pushes it, with the warehouse executing movements but not holding its own number.
What does not work is both systems editing stock freely and reconciling afterwards. That arrangement produces a permanent difference and nobody can say which number is correct.
Youmanman Cloud Warehouse runs 极风 WMS with a free OMS and direct ERP connection. What matters for you is not the brand of the system, it is that you designate one authoritative side and document the direction of every update.
Define the states a product can be in
Stock that is received but not yet shelved, shelved but on hold, allocated to an order, picked but not yet packed, returned and awaiting inspection — each of these is a different state, and each needs to exist in your system or it will end up counted as available.
| State | Should it count as available? | Common mistake |
|---|---|---|
| Received, not yet put away | No | Counted as available, then oversold |
| Put away, good | Yes | — |
| On hold for quality check | No | Sold while held |
| Allocated to an order | No | Double-allocated across two orders |
| Picked, not yet packed | No | Oversold again |
| Returned, awaiting inspection | No | Restocked as if it passed |
The row that catches most new systems is the second-to-last. A returned item that goes straight back to available without inspection will eventually be picked and shipped, and the customer will be the one who finds out.
How to count: cycle counting beats a full stop
A full stocktake stops everything, costs a day, and produces one number that is stale by the time you finish. Cycle counting counts a slice of the warehouse on a rolling schedule, so nothing stops and the accuracy stays current.
Pick the slice by whatever causes the most damage when wrong: high-value items, fast-moving items, or items with a history of discrepancy. Then count them on a schedule rather than when you feel like it.
Blind counting is the technique worth insisting on. The counter is not shown the system quantity first, because a person told the expected number will tend to find it. Count what is there, then compare.
When the count differs, work through it in order
Do not adjust the system to match the count immediately. Find the cause first, because an unexplained variance will come back the same way next week.
| Likely cause | How to confirm | What to change |
|---|---|---|
| Counted in the wrong location | Check bin locations before adjusting | Fix the location data, not the quantity |
| Received but not shelved | Look in the receiving area | Fix the state definition |
| Allocated twice | Look for duplicate allocations on open orders | Fix the allocation logic |
| Damaged units written off without a record | Look for write-off entries | Add a write-off record requirement |
| Returned stock restocked unchecked | Look at recent returns | Add inspection before restock |
| Genuine shortage or loss | None of the above hold | Escalate as a loss claim |
Sync: push or pull, and how often
Once one side is authoritative, decide the direction and the frequency. Near-real-time is easy to ask for and expensive to deliver; a scheduled sync several times a day is usually enough for most sellers, provided that the fields you need for allocation are among them.
The fields that must be current for you to sell are available quantity, on-hand quantity, and inbound quantity. Everything else can lag.
Yunque Post pulls orders in through official platform APIs across 17+ platforms, so availability changes flow in from the platforms rather than from a keyboard. Combined with free storage for up to 180 days, a reorder sitting in production does not distort the picture.
FAQ
Q: How often should stock be counted? A: On a rolling schedule rather than in one big stop. Count the slices that cause the most damage when they are wrong — high value, fast moving, or historically inaccurate.
Q: Should the ERP or the warehouse hold the stock number? A: Pick one as authoritative and document the direction of every update. Two systems both editing stock freely will never reconcile.
Q: Should returned goods go straight back into available stock? A: No. Returned items should be inspected first. Restocking them unchecked is how a customer ends up receiving something that failed.
Q: What is blind counting? A: Counting without seeing the system quantity first, then comparing. A person who is told the expected number tends to find it; a person who is not told it counts what is actually there.
Q: The variance is real. Should I just adjust the system? A: Not first. Work through the likely causes in order, because an unexplained variance returns the following week. Adjusting the number without a cause hides the problem rather than fixing it.
Having your stock figures audited? Send us your SKU list and warehouse layout on our contact page and we will tell you where the gaps are.