A WMS cutover moves on-hand inventory in a fixed order: freeze changes, post the last receipts and shipments, count or validate every location, load balances by location, license plate (LPN), lot, status and owner, reconcile the load against the legacy system and the ERP, and get finance to sign off before the first new transaction. Write the rollback criteria before the cutover weekend, not during it. The count is only as good as the freeze around it, and the load is only as good as the reconciliation after it.
This field note sets out that sequence step by step, with the open transactions that cause most of the trouble, a load-file specification, a fictional reconciliation, the go or no-go decision and the exit test for the weeks after go-live. It covers inventory balances only. Item, unit-of-measure and location master data are a separate readiness task, covered in our field note on the item master data that must be right before go-live.
The WMS cutover sequence at a glance
The inventory cutover has seven steps, and each one has an owner and an exit check. Do not start a step until the previous one has passed its check. Starting a step early because the clock says so is how a count gets lost.
| Step | What happens | Owner | Exit check |
|---|---|---|---|
| 1. Freeze | Stop master data edits, adjustments, transfers and new receipts and shipments in the legacy system | Implementation lead | Freeze confirmed in writing; last document numbers recorded |
| 2. Final transactions | Post every receipt and shipment that physically happened before the freeze | Receiving and shipping leads | No paperwork left unposted; staged stock listed |
| 3. Count or validate | Full count, or validated cycle counts with a sample recount | Inventory control | Every location counted or validated; recounts done |
| 4. Load | Load balances by location, LPN, lot, status and owner | IT or data lead | Load completed with zero rejected rows, or each rejection resolved |
| 5. Reconcile | Compare count, legacy book, WMS load and ERP | Inventory control and finance | Quantity and value differences within agreed tolerance, all explained |
| 6. Sign-off | Finance accepts opening balances; sponsor makes the go or no-go call | Controller and sponsor | Signed sign-off sheet |
| 7. First transactions | First receipts, picks and shipments in the new WMS, checked against the ERP | Area leads | First interface messages reconciled |
Full shutdown or zone by zone
There are two basic shapes. You can stop the whole warehouse, move everything and restart, or you can move one area at a time while the rest keeps running on the legacy system. A published migration playbook from one cloud WMS vendor names both: migrating everything “would require shutting down operations completely”, while the alternative runs the legacy system and the new WMS “simultaneously” with part of the warehouse shut down and being migrated (vendor playbook).
A full shutdown is simpler to reconcile and needs one cutover. Zone by zone keeps shipping but needs two systems, two sets of rules on the floor and a reconciliation for every zone. Choose zone by zone only if you cannot close the building for the time a full count and load take, and plan an extra reconciliation for every zone you add.
Parallel running is a different thing
Parallel conversion, in the ASCM Supply Chain Dictionary’s words, is where “the operation of the new system overlaps with the operation of the system being replaced” and the old system is discontinued “only when the new system is shown to be working properly” (ASCM Supply Chain Dictionary, 18th edition). For inventory on a warehouse floor, true parallel running means every scan happens twice, in two systems. That doubles the scanning work for as long as it lasts, which is why this field note assumes balances move once and reconciliation proves them.
What has to be ready before the cutover weekend
The cutover weekend should only execute a plan that has already worked in rehearsal. Everything that can be prepared, loaded and checked beforehand should be, so the weekend carries only the balances that change by the hour.
Load and validate everything that does not move
- Item master: item numbers, units of measure and conversions, lot and serial control flags, shelf-life rules, dimensions and weights.
- Locations: every location in the building, with its type, zone, capacity, status and whether it tracks license plates.
- Owners: for a 3PL, every client as an owner, with its items and any client-specific rules.
- Statuses and reason codes: available, on hold, damaged, quarantine, in inspection, and the adjustment reasons the floor will use.
- Users and devices: sign-ons, roles, handhelds and printers configured and tested at each station.
Each of these has its own load and its own check. Locations need physical labels that match the system before anyone counts, because a count posted to the wrong location is worse than no count.
Rehearse the load at least twice
Run the full inventory load into a test copy of the new WMS, using a real extract from the legacy system. Time it, count the rejected rows, fix the causes and run it again. Microsoft’s documentation for upgrading warehouse management from its older AX 2012 product to Dynamics 365 Supply Chain Management warns that you “might have to adjust data on the source implementation to fix issues” and that “several additional rounds of data migration might be required” (Microsoft Learn).
A rehearsal should end with three numbers: how long the load took, how many rows it rejected and why, and how long the reconciliation took. Those times go into the weekend plan with a margin.
Decide what you will not migrate
Not everything in the legacy system belongs in the new WMS. Decide early, in writing, what stays behind:
- Transaction history. Plan to start the new WMS with balances, not years of movements, unless you have a specific reason to load history. Keep the history reachable in the legacy system or an archive.
- Closed orders and receipts. Report them from the archive; do not load them.
- Zero balances and empty LPNs. Load locations, not empty records.
- Obsolete items. Write them off or move them to a disposal location before the count, not during the load.
Keep read-only access to the legacy system, or a full set of exported reports, for as long as finance and your auditor need it. Agree that period with them before the legacy system is switched off.
Prepare the people and the paper
- Count teams named, trained and assigned to zones, with a team lead per zone.
- Count tags or count sheets printed and numbered, if you count on paper.
- LPN labels printed for every pallet and tote that will be counted into the new system.
- The reconciliation report built and tested on the rehearsal data.
- A sign-off sheet drafted with the tolerances already filled in.
- A contact list with the decision makers’ numbers for the whole weekend.
The go-live readiness checklist covers the wider gate list for go-live; this field note covers the inventory part in depth.
Choosing the window: peak season, closes and the auditor
Pick the cutover window by working backward from the operating calendar. The best window is quiet on the floor, away from a financial close, far enough before peak to allow several weeks of stabilization, and agreed with the people who audit your inventory.
Stay away from peak
A new WMS slows the floor for a while, whatever the training. Going live just before peak season puts that learning curve into your highest-volume weeks. Plan the go-live so that the stabilization exit criteria described later can be met before peak starts, then work back to the cutover date.
Avoid the financial close
A cutover posts large inventory movements and adjustments. If it lands on a month-end, quarter-end or year-end close, finance has to separate cutover adjustments from ordinary period activity under time pressure. Choose a weekend in the middle of a period where possible, and agree with the controller how cutover adjustments will be posted and labeled.
Talk to the auditor before you fix the date
If an external auditor examines your inventory, tell them the cutover date and count plan before you commit to them. Ask three questions: whether they want to observe the cutover count, whether it changes the year-end procedures they had planned, and which legacy reports they will need to keep. Record the answers.
Whether cycle counts can replace a year-end count is a different question, with its own evidence requirements. Our field note on what has to be true before auditors accept cycle counts instead of a wall-to-wall count covers it.
Example (fictional). A 3PL’s fiscal year ends June 30. One client peaks in July and August with back-to-school orders, and two clients peak from October to mid-December. Working back, the team rules out June (year-end count and close), July and August (back-to-school), October to mid-December (holiday peak) and early January (returns). September is too short a gap between two peaks, and the last week of any month is out because of the close. That leaves a mid-period weekend between late January and May. It picks the third weekend in April, which leaves about ten weeks to meet the stabilization exit criteria before the July peak starts.
Plan the people as well as the date
The weekend needs counters, recount teams, IT, the data lead, inventory control, finance and the sponsor, with overlapping shifts. A holiday weekend gives an extra day but may thin the team. Check carrier pickups and customer delivery promises for the Friday before and the Monday after.
Freezing scope and handling open transactions
The freeze is the moment the legacy record stops changing so that the count can be compared with it. Every transaction that crosses the freeze line, and every document that is half finished, needs a rule written down in advance.
What the freeze stops
- Item master changes, including new items and unit-of-measure edits.
- Location changes: new locations, renames and status changes.
- Inventory adjustments, except those that come out of the cutover count itself.
- Bin-to-bin moves and replenishment.
- Receipts and shipments after their cutoff times.
ASCM defines cut-off control as “A procedure for synchronizing cycle counting and transaction processing.” At cutover, the same idea applies to the whole building: record the last receipt and shipment document numbers posted before the freeze, so anything later can be identified at once.
A rule for every open transaction
Open documents are where opening balances go wrong. Write a treatment for each type, test it in rehearsal, and assign an owner.
| Open transaction at freeze | Treatment | Watch for |
|---|---|---|
| Goods received but not put away | Post the receipt in legacy; count in the staging location; load to staging | Stock on the dock floor with no location |
| Purchase order with ASN, truck not yet arrived | Leave open in the ERP; the new WMS receives it after go-live | ASNs that must be resent or reloaded into the new WMS |
| Order picked and staged, not shipped | Choose before the weekend: ship before freeze, or unpick and return to stock | Staged pallets counted twice, once in staging and once on the order |
| Order allocated, not picked | Release allocations in legacy; reallocate in the new WMS after load | Orders that lose their place in the queue |
| Transfer between sites in transit | Count at the receiving site after arrival; keep in transit in the ERP | Stock that is in neither building’s count |
| Customer returns awaiting inspection | Count to a returns or hold location with status “inspection” | Returns loaded as available |
| Quality hold or quarantine | Load with the hold status and its reason | Holds released by accident during load |
| Consigned or client-owned stock | Count and load with the correct owner | Owner stock mixed with company stock |
Tell the people outside the building
The freeze affects suppliers, customers, carriers and, for a 3PL, clients. Tell them in writing at least a couple of weeks before the weekend, and again two days before:
- Suppliers: the last delivery appointment before the freeze and the first one after go-live. Ask them not to ship trucks that would arrive in between.
- Customers: the last order cutoff that ships before the freeze, and whether orders placed during the weekend ship Monday or later.
- Carriers: changed pickup times for Friday and any extra pickups on Monday.
- 3PL clients: the freeze window, when they will receive their opening balance, and who to call during the weekend.
A truck that arrives Saturday morning with no appointment is a receiving decision nobody should have to make during a count.
Make the freeze visible on the floor
A freeze that lives in an email is broken by the first busy supervisor. Tape off staging lanes, put signs on the dock doors, lock the legacy transaction menus if the system allows it, and post the cutoff times at every desk. Ask each area lead to confirm the freeze in writing.
Counting: a full count or validated cycle counts
Count every location at cutover unless you can show the legacy record is accurate at the level of detail the new WMS needs. That condition is harder to meet than it sounds, because a new WMS often tracks detail the old system never held: LPNs, lots by location, or statuses.
When a full count is the right call
- The legacy system does not hold balances by location, LPN or lot, and the new WMS will.
- Location accuracy from recent cycle counts is below your target, or unknown.
- Units of measure change in the new WMS, for example from cases to eaches.
- The building is being relabeled or reslotted at the same time.
ASCM’s dictionary says inventory accuracy targets “usually are 95 percent to 99 percent, depending on the value of the item”. It also notes that for location management it is “sometimes measured as the number of storage locations with errors divided by the total number of storage locations”. For a cutover, measure location accuracy, because the new WMS will direct people to locations.
When validated cycle counts can work
Validated cycle counts mean loading the legacy balances for locations that recent counts have proven accurate, and recounting a random sample at cutover to confirm it. This works only when the legacy record holds everything the new WMS needs, and when the cycle-count history covers every location recently.
Example (fictional). A DC has 6,400 active locations. Its cycle-count program counted every location at least once in the last 10 weeks, and location accuracy over that time was 98.9% against a 98% target. The legacy WMS already tracks LPN and lot by location.
The team loads legacy balances and recounts a random sample of 400 locations during the freeze. The acceptance rule, written in advance: if 98% or more of the sample matches within tolerance, the load stands; if not, the zone with the most errors gets a full count. The sample finds 394 matching locations, 98.5%, so the load stands. The six mismatches are corrected from the recount and investigated after go-live.
Set count tolerances before anyone counts
Decide in advance how close a count must be to the legacy record to be accepted without a recount. Base the tolerance on value, not on habit, and write it on the count instructions.
| Item group (example) | Recount when the difference is | Accept after recount when |
|---|---|---|
| High value or serialized | Any difference | Two independent counts agree |
| Fast movers, standard value | More than 2% or more than 5 units | Two counts agree within 1 unit |
| Slow movers, low value | More than 5% | Two counts agree within 2% |
| Bulk or weighed items | More than the scale tolerance | Reweigh agrees within the scale tolerance |
The values are examples. The point is that a counter at 3 a.m. should never have to decide whether a difference matters.
How to run the count
- Count in teams of two: one counts, one records, and they swap for recounts.
- Count blind: counters do not see the system quantity.
- Count by location, then by LPN within the location, then by lot: in that order, so nothing is counted without a place.
- Recount every variance above tolerance with a different team before anything is accepted.
- Count the hiding places: dock floor, staging lanes, returns cage, QC hold, maintenance cribs, trailers at the door.
- Control the paper: if you use count tags, account for every tag number, used or void.
Building the load file: location, LPN, lot, status and owner
The load file is the opening balance. It has one row for each unique combination of location, item, LPN, lot, status and owner, and every row must pass the same validation the new WMS will apply to a live transaction.
Fields and validation rules
| Field | Rule to validate before loading |
|---|---|
| Location | Exists in the new WMS, is active, and matches the physical label |
| Item | Exists, is active, and is allowed in that location type |
| Quantity and unit of measure | Positive, in a unit the item uses, converted with the new WMS’s conversions |
| LPN | Unique across the building; present only where the location tracks LPNs |
| Lot | Present for every lot-controlled item; matches the lot on the product |
| Expiry date | Present where the item is expiry-controlled; not in the past unless status is hold |
| Serial numbers | One row or child record per serial for serialized items; count matches quantity |
| Status | Available, hold, damaged, quarantine or inspection, with a reason for any non-available status |
| Owner | Present for every row in a multi-client building; the item is set up for that owner |
The LPN rule has a direct parallel in Microsoft’s migration validation, which checks that “On-hand inventory with pallets must exist only on license-plate-tracked locations” and that “On-hand inventory without pallets must not exist on license-plate-tracked locations”. Whatever your WMS, write the equivalent rule into the load validation, not into a fix-up after go-live.
Owner for 3PLs
In a multi-client warehouse, “owner” is the client who owns the stock, and every balance must carry it. Reconcile per owner as well as in total. A building can balance in total while two clients’ stock is swapped.
After sign-off, send each client its opening balance by item and lot, and ask for confirmation. Where clients take inventory by EDI, that may be an 846, which X12 names “Inventory Inquiry/Advice”; any cutover adjustments may go as 947s, “Warehouse Inventory Adjustment Advice” (X12). Agree the format with each client before the weekend.
Serial numbers, kits and catch weight
Serialized items need one record per serial, and the count must capture the serials, not only the quantity. Kits and assemblies need a decision: load them as finished items, or as components, but never both. Items sold by weight need the weight recorded at count time if the new WMS tracks it per LPN or per case.
What to do with rejected rows
Expect the load to reject some rows, and plan for it: rejected rows go to a named person with a time limit, not to a shared folder. Sort them by cause, because one cause often explains many rows: a missing location, an inactive item, a unit of measure without a conversion, a lot missing on a lot-controlled item.
Fix the cause in the reference data, not by editing the rejected rows one by one, then reload only the rejects. Keep the rejected file, the fix and the reload as one record; the auditor or controller may want to see how many rows were touched and why.
Example (fictional). Three rows from a load file for a multi-client building:
| Location | Item | Qty | UoM | LPN | Lot | Expiry | Status | Owner |
|---|---|---|---|---|---|---|---|---|
| A-14-03-B | EX-20118 | 40 | CS | LP0000812 | L2611 | 2027-03-31 | Available | CLIENT-07 |
| QC-HOLD-01 | EX-30544 | 12 | CS | LP0000977 | AX7731 | 2026-04-15 | Hold, short dated | CLIENT-07 |
| R-02-01-A | EX-41207 | 230 | EA | (none) | (none) | (none) | Available | CLIENT-11 |
The third row has no LPN because R-02-01-A is a carton-flow pick location that does not track license plates. The second row carries its hold status and reason into the new WMS, so it cannot be allocated on Monday.
Load in an order the WMS can accept
Load reference data first, then empty LPNs if your WMS needs them created before contents, then contents by location. Load holds with their statuses in the same pass; releasing everything to available and re-holding it later invites shipping held stock on Monday morning.
Reconciling against the legacy system and the ERP
Reconciliation compares four numbers for every item: what was counted, what the legacy system said, what the new WMS loaded, and what the ERP holds. The cutover passes when the differences are within tolerance and every difference above a threshold has an explanation.
The four-way comparison
- Count against legacy book. The book inventory, in ASCM’s definition, is “obtained from perpetual inventory records rather than by actual count”. Differences here are the true variances the count found.
- Load against count. Differences here are load errors: rejected rows, wrong conversions, typing errors. The target is zero.
- WMS total against ERP, by item. Differences here are adjustments the ERP still needs, or timing differences from transactions that crossed the freeze.
- Value against ERP value. Finance compares extended value at the ERP’s cost, because quantity can balance while value does not when units of measure are wrong.
Which system should hold the record for each number after go-live is a design decision, covered in our field note on which system should own each inventory number.
Where the variances come from
A variance line is not the same as missing stock. Before treating one as loss, check whether it is stock in the wrong place in the record or a transaction on the wrong side of the freeze; those can be explained and corrected before sign-off.
Example (fictional). After the count and load, the reconciliation shows 412 item-location lines with a quantity variance out of 11,850 lines loaded. Investigating them by cause, the team finds 96 lines from receipts and shipments posted in legacy after the freeze, which are reversed and reposted on the right side of the line. Another 88 lines are stock counted in staging lanes and the returns cage, which legacy held in a single “dock” location. There are 71 lines where legacy held cases and the count recorded eaches, and 54 lot or LPN keying errors corrected from the count sheets.
That leaves 103 lines of genuine variance, worth a net loss of $6,350 at standard cost against an ERP inventory value of $2,481,300, or 0.26%. The agreed tolerance was 0.5% in value with every line over $500 explained. Finance signs off, and the 103 lines are posted as cutover adjustments with their own reason code.
Agree tolerances before the weekend
A tolerance set on Sunday afternoon is easy to fit to whatever the reconciliation shows. Agree them with the controller weeks before, in three parts:
- A line tolerance for quantity, below which a variance is accepted without investigation.
- A value threshold above which every line must have a written cause.
- A total tolerance for the net value of all cutover adjustments, as a percentage of inventory value.
Put all three on the draft sign-off sheet, so the reconciliation is read against them rather than argued over.
Post cutover adjustments where finance can see them
Agree in advance where cutover adjustments post: in the legacy system before the load, or in the new WMS and ERP after it. Either works if it is consistent and labeled. Give them a dedicated reason code, so that cutover losses never mix with the first month’s ordinary adjustments.
Finance sign-off and the go or no-go decision
The go or no-go decision should be a reading of criteria agreed weeks earlier, not a debate at two in the afternoon on Sunday. Two documents settle it: the finance sign-off on opening balances, and the go-live criteria for the system itself.
What finance signs
- The reconciliation report: count, legacy book, WMS load and ERP, by item and in total, in quantity and value.
- The list of variance lines above threshold, each with its cause.
- The cutover adjustment total and how it will be posted.
- For a 3PL, the per-owner reconciliation and the plan to send each client its opening balance.
- The last legacy document numbers before the freeze.
Run the weekend from one room
Run the cutover from a single command point, physical or virtual, with one person in charge of the timeline. Hold a short checkpoint at the end of every step in the sequence table: done or not, open issues, and whether the next step can start. Keep two logs: an issue log with owners and times, and a decision log that records every decision and who made it. The decision log is what you will read on Tuesday when someone asks why a pallet was loaded to hold.
Rollback criteria, written in advance
Rollback criteria are the conditions under which you abandon the cutover and restart on the legacy system. Write them as testable statements, with numbers, before the weekend. For example:
- Opening balances not reconciled within tolerance by the sign-off time.
- The order release interface from the ERP to the new WMS failing its test orders.
- Label printing failing at more than one pack station after fixes.
- Handheld connectivity failing in any zone that ships on Monday.
If you roll back
A rollback is a planned procedure, not a retreat. The sponsor announces it at a set time. The legacy system reopens with its pre-freeze balances, adjusted only for count corrections finance chooses to post there. The count results, load files and reconciliation are kept for the next attempt, along with a short list of what failed and why. Suppliers, customers and carriers are told the same day.
The point of no return
Every cutover has a point after which rolling back costs more than fixing forward. A practical marker is the first shipment confirmed in the new WMS and sent to the ERP. From then on the two systems have diverged, and a rollback means unwinding real customer transactions. Put the point of no return on the timeline, with the rollback deadline before it.
A rollback plan names who decides, how the legacy system is reopened, how any transactions entered in the new WMS are reversed or re-keyed, and how the count results are kept for the next attempt. Our project planning work includes this plan as a named deliverable, alongside the cutover date and the risk register.
The first weeks after go-live: stabilization exit criteria
Going live is not the end of the cutover. The first weeks show whether the opening balances were right, because errors surface as short picks, negative balances and interface mismatches. Define in writing when stabilization ends, so the extra support does not stop too early.
Why the exit needs criteria
Research on ERP adopters by Markus, Axline, Petrie and Tanis found that “the success of ERP systems depends on when it is measured”, and that “many of the problems experienced in later phases originated earlier but remained unnoticed or uncorrected” (Journal of Information Technology, 2000). A cutover that looks clean on Monday can still carry a bad conversion or an unexplained variance that shows up in week three.
Example exit criteria
The values below are examples for a fictional single-site operation; set your own before go-live.
| Measure | Example exit criterion |
|---|---|
| ERP to WMS on-hand reconciliation | Within tolerance on 10 consecutive business days |
| Negative balances | None open at the end of each day |
| Location accuracy from cycle counts | At or above target for 4 consecutive weeks |
| Interface error queue | Cleared every day, with no message older than one business day |
| Short picks caused by record errors | Trending down week over week |
| Severity 1 support issues | None open for 2 consecutive weeks |
| Shipping performance | Back to the pre-go-live baseline |
Check the first transactions on Monday
The first hours show whether the load works in practice. Before the first shift ends, check each of these on real documents:
- The first receipt posts to the right location, and the ERP receives the receipt message.
- The first pick goes to a location that holds the stock, with the right lot.
- The first shipment confirms in the WMS, prints its labels and reaches the ERP.
- A held lot cannot be allocated or picked.
- One cycle count in a high-variance area matches the loaded balance.
- The interface error queue has been read, and every message in it has an owner.
Count early and on purpose
Schedule cycle counts in the first weeks for the locations most likely to be wrong: those with variances at cutover, staging and returns areas, and items whose units of measure changed. ASCM’s dictionary states that the key purpose of cycle counting “is to identify items in error, thus triggering research, identification, and elimination of the cause of the errors.” The first weeks after a cutover are when that purpose matters most.
Keep the people close
Stabilization depends on people on the floor who know the new screens, on every shift. Super users, extra supervisors and a short daily review of exceptions belong in the plan until the exit criteria are met, not until a date on the calendar.
What to do next
Start with a rehearsal, not a date. A timed rehearsal load and reconciliation will tell you how long the weekend really needs, and which data needs fixing first.
- Choose full shutdown or zone by zone, and write down why.
- List every open transaction type and its treatment at freeze.
- Decide full count or validated cycle counts, with the acceptance rule for the sample.
- Write the load-file specification and validation rules, including LPN, lot, status and owner.
- Rehearse the load and reconciliation at least twice on real extracts, and time both.
- Agree tolerances, the sign-off sheet and the cutover adjustment reason code with finance.
- Write rollback criteria and the point of no return, and put both on the timeline.
- Tell your auditor the date and count plan, and record their answers.
- Write the stabilization exit criteria before go-live.
The fictional example project shows where a cutover sits in a full engagement, from the first process walk to trained users. The go-live readiness checklist covers the rest of the go-live gate, beyond inventory.
