Systems and records
Inventory count correction data: testing allocation after a cycle count
By SourceX Editorial · Reviewed by Noah Loul ·
10 sections
Short answer
To test stock allocation after a cycle count changes available inventory, capture the state before the count, post the correction, then re-run every open commitment against the corrected available quantity. Pass only if each affected order, transfer and purchase line gets a decision from someone with authority, and the outcome is confirmed in a shipment, receipt or customer notice rather than a status field.
Checklist
Inventory count correction
Key takeaways
- Open counting work usually does not lock stock, so orders keep consuming a quantity that is about to change.
- A count correction is a chain event: the difference lands on reservations, released picks, transfers and the purchase orders sized on the old figure.
- The cheapest fix is often invalid for a reason held in another record, such as a supplier approval, a branch minimum or a no-substitution clause.
- Grade the outcome from shipments, receipts and notices, not from a Closed or Released status.
- Keep reason codes, reviewers and before and after figures, or past shortfalls disappear from the history.
How do you test stock allocation after a cycle count changes available inventory?
Capture the state before the count, post the correction, then re-run every open commitment on that item and location against the corrected available quantity. The test passes only when each affected order, transfer and purchase line ends in a decision a person with authority made, and the outcome shows up in a shipment, receipt or customer notice rather than in a status field.
That makes this a chain test, not a single check. A count correction changes one number, on-hand stock at a location, but the damage sits in everything that was already promised from that number: reservations on sales orders, picking work already released to the floor, a transfer to a branch and the purchase order that was sized on the old figure.
This page builds one worked case and a reusable task packet for it. The case is illustrative: the item, quantities, durations and order numbers are invented to show the structure, and none of them describe a real company. The point is the shape of the evidence, which a procurement or warehouse team can copy onto its own records.
Why an open count does not stop the system from promising the stock
In most warehouse systems, counting work does not lock the stock being counted, so commitments keep consuming it while the count is open. Microsoft's Dynamics 365 Supply Chain Management documentation says this directly: the on-hand inventory at a location with open counting work is not blocked and stays available for reservation and outbound processing.
The correction itself also arrives late and in steps. In the same documentation, a count that falls outside a worker's deviation limits goes to a Pending review status, and the counted value is rejected unless a reviewer approves the difference. The workers never see the expected quantity, a design meant to prevent intentional miscounts. Between the count and the approval, the system keeps allocating against the old number.
Odoo's inventory adjustment screen shows the same structure from another angle. Each line carries an On Hand quantity recorded in the database, a Counted quantity entered after the physical count, and a Difference calculated between them, along with the last count date and the user responsible. Nothing changes on hand until someone applies the count.
Which bucket absorbs the loss matters as much as the size of the loss. Shopify's developer documentation defines on_hand as the sum of available, committed, reserved and other states, where committed covers placed orders that are not yet fulfilled. When a recount lowers on hand below what is already committed, available cannot absorb the gap, and some existing promise has to give.
The starting state of the illustrative case
The case uses one bearing item stored in one pick location, three open sales orders, a branch transfer and an open purchase order. The count happens while a picking wave for the largest order is already released, which is the mid-workflow timing the test is built around.
| Record (illustrative) | State before the count | Linked to | Why it matters |
|---|---|---|---|
| Item BRG-62 at location A-12 | 120 on hand, 90 committed, 30 available | Item master, specification revision C | The number every commitment was made against |
| Sales order SO-1 | 40 units, ships tomorrow, picking work released | Customer contract with a no-substitution clause | Already on the floor when the count starts |
| Sales order SO-2 | 30 units, ships in 5 days, reserved | Standard terms, partial shipment allowed | Can take a split shipment |
| Sales order SO-3 | 20 units, ships in 12 days, reserved | Standard terms | Has time to wait for a receipt |
| Transfer order TO-7 | Not reserved, 10 units planned to branch B | Branch replenishment rule | Branch would drop below its own minimum |
| Purchase order PO-55 | 100 units from supplier S1, receipt in 9 days | Approved supplier list, revision C | The only confirmed inbound supply |
| Cycle count CC-14 | Counted 70 at A-12, difference of 50 fewer units | Worker deviation limit of 10 units | Goes to Pending review, not straight to on hand |
Which commitments does the shortfall reach, and in what order?
Rank the affected commitments by how little room each one has, not by order number or creation date. After approval, A-12 holds 70 units against 90 committed, so 20 units of existing promises cannot be met from this location, and the 30 units the system showed as available a moment ago are gone.
In the illustrative case, SO-1 comes first because its picking work is already released and it ships tomorrow. A picker sent to A-12 for 40 units will either find them or report a short pick, and the record of what happened next is the evidence that matters. SO-3 comes last because a receipt is due before it ships.
Reservation rules decide which orders the system thinks are protected. Microsoft's documentation explains that reserved inventory cannot be withdrawn for other orders unless the reservation is canceled, and that reservations can follow a first-ordered, first-delivered rule or give priority to certain customers and order types. Odoo sets reservation per operation type, at confirmation, manually or a set number of days before the scheduled date. The same physical shortfall therefore lands on different orders depending on configuration, and the packet has to record which rule was in force.
| Commitment | Slack before it fails | Candidate actions | Who can decide |
|---|---|---|---|
| SO-1, 40 units, released picking | Less than one day | Pick 40 now and short another order, or split and notify | Warehouse lead with sales approval |
| SO-2, 30 units, reserved | 5 days | Partial shipment of 10 now, 20 from PO-55 | Customer service with customer consent |
| SO-3, 20 units, reserved | 12 days | Hold and fill from PO-55 after receipt | Planner |
| TO-7, 10 units to branch B | Depends on branch minimum | Cancel or defer the transfer | Inventory planner for branch B |
| PO-55, 100 units inbound | 9 days to receipt | Expedite, increase quantity or add a second supplier | Buyer within approval limits |
Which constraint rules out the cheapest or fastest fix?
The obvious fixes often break a constraint that lives in a different record from the inventory figure. A test case is only useful if it includes at least one such constraint, so that a system or agent that grabs the first available supply fails visibly.
In the illustrative case there are three. Supplier S2 has stock and could ship in 2 days, but S2 is approved only for specification revision B of the bearing, and the item is now on revision C. The branch holds spare units, but moving them puts branch B under its own replenishment minimum. A substitute part is in stock, but SO-1's contract forbids substitution without written approval.
Each of these is cheaper or faster than the right answer, and each is invalid for a reason that only a linked record can show: the approved supplier list, the branch rule or the contract clause. The grader should check that the chosen action cites the record that permits it, not only that the order shipped.
- Supplier route: valid only if the supplier is approved for the current specification revision on the date of the order.
- Branch transfer: valid only if the sending location stays at or above its own minimum after the move.
- Substitution: valid only if the customer agreement allows it or a written approval is attached.
- Expediting the open purchase order: valid only within the buyer's approval limit, or with the approval that raises it.
The count correction and affected commitment task packet
The packet is one record per count correction that links the count to every commitment it touched and to the outcome of each. A team can build it from the count journal, the reservation history, the work exception log and the purchasing records, and use it either to audit a past incident or to grade a system that handles the next one.
At a glance
The count correction and affected commitment task packet
- Count identifier, item, location, count method (planned, threshold or spot) and the worker or device that recorded it.
- Expected on hand, counted quantity and difference, with the time the count was entered and the time the difference was approved or rejected.
- Reviewer identity, the deviation limit that sent the count to review and the reason code chosen for the adjustment.
- A snapshot of on hand, committed, reserved and available for the item and location just before the correction posted.
- Every open commitment on the item and location at that moment: sales orders, transfers, production picks and released picking work, each with quantity and due date.
- The reservation rule in force (for example first ordered, first delivered, or reservation a set number of days before the scheduled date).
- Inbound supply that could cover the gap: open purchase order lines, expected receipt dates and the supplier approval status for the current specification revision.
- Constraints attached to each commitment: substitution rules, partial-shipment terms, branch minimums and buyer approval limits, each with the record it comes from.
- The decision for each affected commitment (reallocate, split, hold, cancel, expedite, substitute) and the person or rule that made it.
- Customer or internal notices sent because of the decision, with time sent and any reply.
- Outcome evidence: shipment confirmation, goods receipt, or cancellation record for each commitment, with dates.
- A recount or follow-up count, if one was done, and whether it confirmed or reversed the correction.
- A label stating whether the case is real or illustrative, and which records were redacted or summarized.
Which linked records prove the decision and the outcome?
The proof is the downstream record that shows the decision took effect: the shipment that left with the quantity promised, the receipt that refilled the location, or the notice the customer received. A status of Closed on the count or Released on the order says only that a step finished, not that the promise was kept.
For the illustrative case, the evidence chain runs as follows. The approved count CC-14 links to a reason code and the reviewer. The SO-1 short pick links to a work exception and a reallocation to another location, or to an adjustment that removed the reservation. SO-2 links to a split shipment and a customer consent. SO-3 links to the PO-55 receipt and the later shipment. TO-7 links to a cancellation that cites the branch minimum.
Customer-facing promises can carry their own rules. For U.S. mail, internet and telephone orders, the Federal Trade Commission's business guide to its Mail, Internet, or Telephone Order Merchandise Rule says sellers need a reasonable basis for any shipping time they state, or for shipping within 30 days when none is stated, and must offer buyers the choice to accept a delay or cancel when they cannot ship on time. That is general information, not legal advice; whether it applies to a given order, especially a business-to-business one, is a question for counsel.
| Commitment | Decision record | Outcome record | What a status field alone would miss |
|---|---|---|---|
| SO-1 | Short-pick work exception and reallocation | Shipment confirmation with picked quantity | Released work can close with fewer units than ordered |
| SO-2 | Split-shipment approval and customer consent | Two shipment confirmations | A partial shipment can read as complete |
| SO-3 | Hold note linked to PO-55 | Goods receipt, then shipment | The order looks on time until the receipt slips |
| TO-7 | Cancellation citing the branch minimum | Branch on-hand history | A canceled transfer looks like a planning error |
| PO-55 | Expedite request within the buyer's limit | Goods receipt with quantity and date | A confirmed PO is not a received one |
Counterexamples that stop a grader rewarding the wrong move
Every failure case needs sibling cases where the right answer is different, or a grader learns to reward caution instead of judgment. Four counterexamples cover the common traps.
- The count was wrong. A recount finds the missing units in an adjacent bay, so the correction should be reversed and no commitment touched. A system that cancels orders on the first count fails.
- The stock is elsewhere. Unreserved units sit in another location in the same warehouse. Microsoft's documentation describes short-pick work exceptions that can reallocate picking to locations with unreserved quantities, automatically or with the worker choosing; that path is correct here, and a purchase is not.
- The correction came after the shipment. SO-1 already left with 40 units before the count was approved, so the shortfall falls only on later orders. Grading has to use event times, not the order of rows in an export.
- The difference is inside the deviation limit. A 6-unit gap is accepted without review under the worker's limit, and the right test is whether available stock was recalculated before the next reservation, not whether anyone was notified.
Reason codes and adjustment history decide whether the record can be reused
A count correction is only useful later if the history keeps the reason, the reviewer and the before and after figures, rather than just the new on-hand balance. Many exports carry only the current quantity, which makes every past shortfall look like it never happened.
Systems give teams the fields to keep that history. Microsoft's documentation describes counting reason code policies that can be mandatory or optional for cycle counts, spot counts, threshold counts and counting journals, set by warehouse or by item, and recommends deciding up front whether reason codes should drive the financial offset account. Its short-pick setup lets a work exception adjust inventory to zero at the short-picked location and remove the order line's reservations through the adjustment type.
Before using these records as a dataset or test suite, check three things: that reason codes were actually required during the period, that adjustment history was retained rather than overwritten, and that the commitment records can be joined to the count by item, location and time. If any of the three fails, label the affected period rather than filling the gaps by inference.
Where real count-correction histories come from
Real cases come from companies whose warehouse, order and purchasing systems have kept count journals, reservation changes and exceptions over several years. Those histories are hard to recreate synthetically because the hard part is the constraints, such as supplier approvals and contract terms, that sit in other systems.
For a distributor or manufacturer that keeps count journals, reservation changes and purchasing exceptions, SourceX can turn a selected slice of that history into a licensed dataset, running the rights review, contracting, preparation and delivery. Buyers generally want consistent history of two or more years and tens of thousands of records in one system, although rarer workflows can qualify with less. Fit is checked through a short company assessment with no uploads, and no warehouse record is delivered before the company has approved the scope and terms and signed.
Frequently asked questions
What is inventory count correction data?
It is the record of a physical count that disagreed with the system quantity: the expected and counted figures, the difference, who reviewed it, the reason code and the adjustment posted. Useful versions also keep the commitments open at the time and what happened to each. Without those links, the data shows that stock changed but not what the change broke.
Does a cycle count lock inventory while it is being counted?
Usually not. Microsoft's Dynamics 365 documentation says on-hand stock at a location with open counting work stays available for reservation and outbound processing. That is why a count can change the available quantity in the middle of a picking or ordering workflow.
Which orders should absorb a count shortfall first?
The ones with the most room to recover, such as orders that ship later or allow partial shipment, unless a customer priority rule says otherwise. Orders with released picking work and near ship dates usually need the first decision. Record the rule used, because a different reservation policy would have chosen different orders.
How is a count correction different from a short pick?
A count correction comes from a deliberate audit of a location and changes on hand after review. A short pick is reported by a picker who cannot find the quantity the work asked for, often in the middle of fulfilling an order. Both should link to the same commitment records, and a short pick often triggers a count.
Can these records be used to test an AI agent or planning system?
Yes, if each case keeps the state before the count, the open commitments, the constraints that apply and the verified outcome. Pair every failure case with counterexamples, such as a wrong count or stock in another location, so the test rewards correct judgment rather than caution.
Sources
- Microsoft's Dynamics 365 documentation says on-hand stock at a location with open counting work is not blocked and stays available for reservation and outbound processing, and that counts outside deviation limits go to Pending review. Source
- Dynamics 365 counting reason code policies can be mandatory or optional across cycle, spot and threshold counts and counting journals, by warehouse or item. Source
- Dynamics 365 short-pick work exceptions can reallocate picking to locations with unreserved quantities, adjust inventory at the short-picked location and remove reservations. Source
- Dynamics 365 reserved inventory cannot be withdrawn for other orders unless the reservation is canceled, and reservations can follow first-ordered, first-delivered or customer priority. Source
- Odoo's Physical Inventory page records On Hand, Counted and Difference quantities with the last count date and responsible user. Source
- Odoo offers three reservation methods, At Confirmation, Manually and Before scheduled date, configured per operation type. Source
- Shopify defines on_hand as the sum of available, committed, reserved and other inventory states, with committed covering unfulfilled orders. Source
- The FTC's business guide says sellers need a reasonable basis for stated shipping times and must offer buyers a delay-or-cancel option when they cannot ship on time. Source
See if your company qualifies
A short company assessment. No data uploads are needed.