Here is the conclusion up front: when you replace a WMS, never set the wms data freeze at a tidy clock time like "Friday at close of business." Set it at the seams between order statuses — in-flight orders finish in the old system, new orders start in the new one, and a reconciliation sheet covers the gap. That way the warehouse never stops shipping, and cutover night does not require an all-nighter from the whole team.
I have seen this pitfall three times, and the script is always the same. The project team picks a convenient moment — Saturday, 2 a.m. — freezes everything, shuts down the old system, and fires up the new one. Monday morning, hundreds of orders do not match. Pickers cannot find yesterday's waves in the new system. Customer service phones light up. The new system is not the problem. The freeze point was cut in the wrong place.
Why Cutting the Data Freeze by Clock Time Fails
Warehouse orders are not a tap you can turn off. An outbound order travels from order entry to waving, picking, checking, packing, and shipping — a full day or more. An inbound order crosses shifts from appointment to arrival, receiving, and putaway. If you freeze at 2 a.m. on Saturday, what happens to the orders stuck halfway? Either you force-close them in the old system, which makes the data fake, or you carry them into the new system as-is, where the status mapping almost never lines up.
The right approach is to cut by order status. Before cutover, sort every order into three buckets: completed, in progress, and not started. Completed orders get archived and never enter the new system. Not-started orders wait and get created in the new system after go-live. Only the in-progress orders need to go through the status mapping table, one by one. This is the most time-consuming step in the whole switch, and the one you can least afford to skip.
The Data Freeze Cutover: A Go-Live Night Checklist
Step one: freeze master data two weeks early. Product masters, location masters, supplier and customer records — data that changes rarely — get frozen first. Changes made in the old system after that point do not sync to the new one; they go on a manual log. On one project, master data was not frozen, and someone changed 200-plus product barcodes in the old system three days before go-live. The new system held the old codes, receiving scans failed on the spot, and the floor ground to a halt.
Step two: three days before cutover, take a status inventory. List every in-progress order by status: ordered but not waved, waved but not picked, picked but not checked, checked but not shipped. Count the orders and lines in each bucket, print the list, and put it on the wall. Those numbers are the reconciliation baseline for cutover night. Not one order short.
Step three: set the wms data freeze so in-progress orders finish their current step in the old system before handing over. An order that is picked but not checked gets checked in the old system and hands over before packing. An order that is waved but not picked gets its pick task cancelled in the old system and rebuilt in the new one. The cut must fall between process steps, never in the middle of one — I have seen this pitfall three times, and every time someone cut mid-step, a half-picked task ended up recognized fully by neither system.
Step four: on cutover night, do only three things — migrate, reconcile, and sample-test. Run the migration against the mapping table, verify every in-progress order's status one by one, and sign the reconciliation sheet on both sides. Then sample-test at least 20 orders end to end, from waving through shipping, in the new system before releasing it. If the sample test fails, roll back before dawn. Do not push through.
Step five: run dual-track reconciliation for three days after go-live. Every evening, compare the old system's and the new system's inbound and outbound volumes. If the difference exceeds 0.5%, find the cause before the next morning. Dual tracking is not distrust of the new system — it is keeping an escape route open. If something big breaks, at least you know which day it broke on.
The Failure Points People Underestimate
First, the timing gap on the inventory snapshot. The snapshot must be taken at the exact freeze moment, and the books locked immediately after. Nobody touches adjustments during cutover night. On one project, someone kept making cycle-count adjustments in the old system after the snapshot, and the new system went live hundreds of units off. It took two days to trace the gap back to that adjustment.
Second, serial and lot number mapping. Ordinary products map cleanly. Products with serial numbers, lot numbers, and expiry dates are where data gets lost. The mapping table needs a dedicated column showing exactly which old-system serial field lands in which new-system field, and you should verify 50 serial numbers individually before go-live.
Third, do not forget in-transit stock. Open purchase orders, transfer orders, and return orders are routinely left out of the status mapping table because they are not part of the warehouse's daily paperwork. The consequence: goods arrive and nobody claims them, they sit on the dock for three days, and only a customer complaint reveals the shipment does not exist in the system at all.
Here is the counterintuitive judgment to carry into the project: a slow cutover is not a failure. An all-nighter on cutover night is. A well-designed freeze point should let most people go home on time — because the arguments were already had during the two weeks of status inventory, and the numbers were already reconciled on the sheet. If the team is still inventing rules on the floor at midnight, the preparation failed, no matter how good the software is.
One last reminder: once the wms data freeze plan is set, get it signed by the warehouse manager, the customer service lead, and finance before implementation starts. The signatures are not bureaucracy. They mean all three sides accept the same reconciliation sheet on cutover night. A plan nobody signed is a plan anyone can disown on go-live day, with the words "those are not my numbers."
By Lin Xiao, WMS Implementation Consultant







