Here is the conclusion up front: when the two systems meet in a manufacturing warehouse, nine times out of ten the interface is not broken. The master data is. Item codes, work order numbers, batch rules — if any one of those means something different in each system, receiving, putaway, issuing, and reporting will fall apart one after another in the first week after go-live. An interface that connects is not the same as data that agrees. I put that sentence on page one of every project kickoff document.

I have seen this exact failure three times. The third time was at an auto parts plant in the fall of 2024. The production line was rushing an order, and the warehouse spent every day hunting for material. The system said the raw material was in stock. The shelf said otherwise. After a full round of checks, the root cause was neither the warehouse nor the line. It was the master data mapping table between WMS and MES. The same item carried a code that differed by one suffix character between the two systems, so the warehouse system could not recognize the material on the MES issue order. Staff corrected orders by hand. Hand corrections scrambled the batches, and scrambled batches broke traceability. For that whole week, the planner's first job every morning was to reconcile the numbers before daring to schedule production.

Where WMS-MES Integration Breaks First: Receiving

Receiving in a manufacturing warehouse is not like receiving in a pure e-commerce warehouse. An e-commerce warehouse receives sealed cartons from suppliers. A manufacturing warehouse receives semi-finished goods, parts back from subcontractors, and leftover material returned from the line. MES cares whether a batch can go into a work order. WMS cares which location it sits in and which batch it belongs to. The first place the two disagree is the receiving dock.

The most typical scene I have watched goes like this (figures shown for illustration): the supplier delivery note says 500 pieces, the MES purchase order says 500 pieces, and WMS scanning recognizes only 480. The batch numbers on the other 20 pieces do not exist in MES at all. The receiver takes in 480, parks the remaining 20 in the inspection area, and waits for the planner to call the supplier. Those 20 pieces are invisible in the system. When the line urgently needs them, the warehouse says there is no stock and purchasing says the goods already arrived. Both sides have documents. Both sides are right. What is wrong is that nobody decided, during integration, who generates batch numbers and in what format.

The fix is not complicated, but it has to be written into the go-live night checklist and treated as non-negotiable: only one system generates a batch number, and the other system may only reference it, never invent its own. Receiving validation follows the generating system's batch. Goods that cannot be scanned are never manually released into a storage location. They go to the inspection area first, with an inspection record. The rule looks rigid. It saves about two hours of reconciliation every day after go-live.

Issuing and Reporting: Where MES and WMS Break Second

Once receiving is in trouble, issuing follows. MES issues material by work order. WMS picks material by location. A work order calls for 100 kilograms of a raw material (figures for illustration), and WMS holds that material across three locations and two batches. First-in-first-out, or issue by a specified batch? If the integration document does not spell it out, the issuer picks by experience. Pick right and nobody notices. Pick wrong and two batches get mixed, and quality traceability breaks on the spot.

There is a quieter trap in reporting. MES records the quantities the line reports. WMS deducts stock when material is issued. The line reports 100 finished pieces, but the material issued was only enough for 96 (figures for illustration). Where did the gap of four go? Many projects only discover at this point that nobody designed a variance process. Reported quantities and issued quantities do not match, and finance only finds the cost does not reconcile at month-end close. By then, tracing back means digging through hundreds of orders.

The table below is one I walk through line by line with every client doing this kind of integration. Look at the last column: every row needs exactly one owner. Writing two department names there means nobody owns it.

Step What MES owns What WMS owns Single owner
Item master data Item code, routing, unit conversion Location, batch attributes, storage conditions Master data administrator
Work order release Work order number, quantity, start time Staging by work order, location reservation Production planner
Receiving Arrival notice, inspection standard Actual received quantity, batch putaway Warehouse receiving team
Issuing Issue order, quantity to pick Picking by batch, first-in-first-out Warehouse issuing team
Reporting and deduction Reported quantity, completion confirmation Stock deduction, leftover return Production statistician
Variance handling Variance reason, replenishment approval Variance documents, ledger adjustment Both sides sign off

The table looks simple. The hard part is getting both sides to accept it. My approach is to print it before go-live, leave the owner column blank, and ask the people in the room to fill in names themselves. Any cell nobody fills in is a place that will break after go-live.

A Counterintuitive Call: Fewer Interfaces Break Less Often

Many clients want to integrate everything at once: work orders, items, batches, inspection, reporting, equipment status. My call is the opposite. Phase one connects only three things: item master data, work orders, and the stock deduction tied to issuing and reporting. Inspection results and equipment status wait until phase one has run steadily for a month. Fewer interfaces mean faster diagnosis when something breaks. With many interfaces, the first three days after a failure are spent arguing about which system is at fault.

I have seen this pit three times, and the lesson is the same every time: integration is not a technical job, it is a job of assigning responsibility. What data moves between systems, who generates it, who references it, and who fixes it when it is wrong — if those four questions are not written down clearly, the prettiest interface is useless. A home appliance plant did the opposite of the all-at-once approach. It connected only the three items in phase one, stabilized in two weeks, and added inspection and traceability in phase two. It went live a full month earlier than a neighboring plant that tried to do everything in one step. Slow is fast, and in this scenario that is literally true.

My go-live night checklist has only the five items people most often skip: has the master data mapping table been spot-checked line by line, at least 50 items; is there exactly one batch number generator; has anyone actually walked the inspection-area process once end to end; is the variance approver in the room; and is the fallback plan written down, so the warehouse can still issue material manually if the interface goes down. If any one of those five is unchecked, I will not sign off on go-live.

If you are working on WMS-MES integration, do not start by asking which protocol the interface uses. Start by asking your plant one question: who actually decides what an item code is. If nobody can answer that, everything after it is rework.

By Lin Xiao, WMS Implementation Consultant