Let me start with the conclusion: when the WMS and the ERP disagree, the problem almost never sits on reconciliation day. It sits in the thirty days before — in every receiving, putaway, pick, and count, one small action at a time. Monthly WMS–ERP reconciliation is not a month-end ceremony. It is a health check on those thirty days. Done well, variances shrink month after month. Done as a formality, they only pile up.
I have seen this failure three times. The first was a home-goods client whose month-end reconciliation turned up over a thousand variance lines. Finance and the warehouse blamed each other for a week before the cause surfaced: the supplier had over-delivered, the WMS recorded the actual receipt, the ERP booked the purchase order quantity, and nobody processed the variance between the two. The second was a 3PL whose variances clustered on a few big clients' goods. It turned out pickers, to save time, shipped whole pallets by scanning only the pallet code; the system back-calculated the quantity from the pallet, and nobody noticed a layer missing from one pallet. The third is the most typical: a count adjustment was made in the WMS only and never synced to the ERP, so the two systems went their separate ways from that day on, drifting further apart every month.
Three failures, one lesson: WMS–ERP reconciliation checks numbers, but it fixes processes. Before you adjust a number, find which step leaked.
The Monthly WMS–ERP Reconciliation Check, in Five Steps
What the monthly check fears most is not many variances — it is having no fixed process. The five-step process I give clients runs the same day every month, without exception.
Step one: freeze the cutoff. The end of the last working day of the month is the cutoff. At that moment, both systems export an inventory snapshot at the same time. Before exporting, confirm one thing: are there unfinished documents? Unfinished receiving, unapproved outbound orders, transfers stuck in transit — either finish them or explicitly put them on hold. Without a frozen cutoff, the two snapshots are photographs taken at different moments, and they will never reconcile.
Step two: pull the variance list. Line up both systems' quantities by item code and location, and list every line that differs. Sort the list into three buckets: quantity variances, value variances, and "orphan lines" that exist on only one side. Orphans deserve the most suspicion: stock in the WMS but not in the ERP, or the reverse, means a document never made it to the other system at all.
Step three: classify the variances. Never lump them together as "shrinkage" or "surplus." They have to land in specific categories. The five I use are: receiving variance, picking variance, count adjustment not synced, interface dropped a document, and master data mismatch. This step takes the most time and is the most valuable. The finer the classification, the easier the root cause is to find.
Step four: chase the root cause, category by category. For each category, take a few lines and trace the full document chain. A receiving variance means checking the receiving slip, the inspection record, and the putaway task. A picking variance means checking the pick task, the verification record, and the outbound order. Once you find the root cause, fix the process — not the number. Level the numbers without fixing the process and the same variance will come back next month from the same place.
Step five: adjust and re-verify. Genuine variances go through a formal adjustment process: variance report, approval, adjustment, re-verification. Adjustments need two-person verification — one person makes the adjustment, another checks it. After the adjustments, pull the variance list again and confirm the lines are cleared, leaving only known in-transit items or explicitly held lines. Only then is the month's exercise closed.
The five steps usually take two to three days. That is not short, but it beats a week of month-end arguments.
Five Reconciliation Variance Categories, Each With a Fixed Repair
Each category hides a familiar set of failure points. Here are the most common ones, ready to check against during the monthly check.
Receiving variance is the most common. The supplier delivers a different quantity than the purchase order — over or under — the receiver books the actual quantity in the WMS, and the ERP books the order quantity. The repair is a mandatory variance action at receiving: when actual receipt differs from the order, generate a variance document on the spot, get it approved, and follow it up. Variances cannot be left to drift in the system. Another pattern: goods fail inspection and go back, the WMS records the return, and the ERP never hears about it. The document chain from receiving through inspection to return has to run all the way to the ERP. It cannot break halfway.
Picking variance comes second. Pickers grab the wrong item, the wrong quantity, skip a scan, or scan twice. The repair sits in the verification step: before outbound, verify by scanning the item barcode — never by quantity alone. Shipping a whole pallet by scanning only the pallet code is a shortcut I have seen three times, and it went wrong every time. A pallet code proves only that the pallet left. It cannot prove the pallet was full. Every minute saved by skipping verification comes back doubled at the month-end check.
Count adjustment not synced ranks third. A count finds a variance, the WMS adjusts, and nobody syncs the ERP. The repair is simple: count adjustment documents must sync through the interface automatically, or the rule must say the ERP gets a manual sync the same day, with "this month's count adjustments" flagged as a separate item on the reconciliation list. Syncing cannot depend on memory. Anything that depends on memory will eventually be forgotten.
Interface dropped a document ranks fourth. Documents travel between the two systems; a network hiccup, an interface timeout, or a restart on the other side, and a document is lost — booked on one side, missing on the other. The repair is a daily interface check: compare both sides' document counts for the day, and re-send or raise an alert when something is missing. What interface drops fear most is silent loss: gone, and nobody knows until month-end. A daily comparison catches a lost document the same day.
Master data mismatch ranks fifth. Item codes, units of measure, barcodes — the master data in the two systems does not fully match, so the same item has two names. The repair is a single source of truth for master data, usually the ERP as master with the WMS syncing read-only. Creation, modification, and deactivation of master data all go through one entry point. The two systems must never maintain their own separate versions.
The Usual Failure Points: The Go-Live-Night Checklist
Once the reconciliation process is running, a few failure points remain that I check off on every go-live night.
The first failure point: the cutoff was never frozen. Someone finds freezing the cutoff too much trouble, exports the WMS numbers first, and pulls the ERP numbers two days later, mixing two days of business into the comparison. That reconciliation is wrong from the start, and every variance it finds is a false variance. Freezing the cutoff is non-negotiable.
The second failure point: adjustments without approval. The warehouse adjusts its own books, nobody re-verifies, the numbers balance, and the real problem gets buried. Adjustments must go through approval, and large variances escalate to the finance lead. Approval is not there to block people.
The third failure point: adjusting numbers without fixing the process. I have seen this three times; it is the same every time: the variances get leveled, everyone relaxes, and next month the same place produces variances again. After every adjustment, ask: what was the root cause, was the process changed, who owns the change, and when will it be done? Without that follow-up, reconciliation is a monthly numbers game.
The fourth failure point: orphan lines nobody owns. Stock in the WMS but not in the ERP, or the reverse — people see the quantities are small and leave them alone. Leave an orphan alone for six months and it becomes an account nobody can explain. Every orphan line needs a handling decision: backfill the document, return the goods, write them off, or record an explicit hold with a reason and a deadline.
One last yardstick: how well the monthly check works is measured by a single number — the count of variance lines. The first month may show over a hundred lines. With processes fixed, that should fall to single digits within three months. If it does not fall, there is still a leaked step in the process. Go back to the classification. The categories will tell you which one is leaking. That number is the report card for your reconciliation work.
By Lin Xiao, WMS Implementation Consultant







