Inventory Allocation Systems for Manufacturers: From Promise to Reallocation

When a manufacturer considers an inventory allocation system, the first question is not whether a screen can display stock on hand. The real question is whether the business can promise a specific quantity to an order once, preserve that promise through picking and shipment, and release it correctly when the order changes. A factory may have enough book stock but still fail a delivery promise. It may also hold stock for a customer who no longer needs it. This guide for factories in Thailand and ASEAN explains the difference between a preassigned allocation quota and an order reservation, the state transitions behind a delivery promise, the respective authority of ERP, WMS and MES, and what to test in an RFP, FAT and SAT.

Define what the inventory allocation system must control

Order reservation is more than on-hand quantity minus open orders. It decides which demand may use which supply, at a specified item, site, inventory status and date. It tracks that right through change, cancellation, warehouse work and shipment. Our guide to manufacturing order backlogs deals with backlog ageing and delivery-review routines. This article focuses on the evidence behind the promise to each order line.

Before comparing software, draw one process map and answer five questions. At which event is inventory held: order entry, approval or release to the warehouse? Does the rule consider only physical stock, or also purchase orders and planned production? How specific must the commitment be: site, warehouse, status, batch, expiry date or even a customer-designated serial number? Who may release it after cancellation, quantity reduction, quality hold or reprioritisation? Which system is authoritative for the reservation and which event reconciles it with physical movement? A feature checklist labelled “real-time stock” cannot answer these questions.

Microsoft’s documentation describes automatic or manual reservation against a sales order. Reserved inventory cannot be withdrawn for another order unless the reservation is cancelled in whole or in part; physical stock and ordered but not yet received stock are separate options. These are product-specific capabilities, useful as questions for your RFP rather than a promise about every system. Microsoft: Reserve inventory quantities.

Allocation quotas and order reservations are different objects

A planner might earmark a portion of a scarce component for contract customers before their individual orders arrive. That is an allocation quota for a channel, region or customer group. Once an actual order line exists, a quantity assigned to that line is an order reservation. Microsoft Inventory Visibility explicitly describes allocation as a virtual pool established before a sales transaction, while soft reservations are usually linked to an actual sales line. When both are used, its documentation recommends allocating first and then reserving against the allocation. Microsoft: Inventory Visibility inventory allocation.

Confusing the two creates opposite failures. If sales treats a quota as a customer-specific promise, several orders may compete inside the same quota. If the factory reserves only against incoming orders, the first order may consume supply that contract or service obligations required it to protect. Keep a quota ID and an order-reservation ID distinct, and record how consumption of the quota relates to each order. Do not simply subtract both from the same on-hand figure: that double counts the same commitment.

For illustration only, imagine 100 saleable units, a 40-unit contract-customer quota and a 60-unit general quota. Order A consumes 15 units of the contract quota. The system should show 25 units of quota remaining and 15 units reserved to order A. The example numbers are hypothetical. In real operations, first exclude quality holds, prior reservations, incompatible batches and unit-of-measure errors from saleable supply. The calculation and event owner must be explicit in the design document.

Do not compress availability into one number

“100 units on hand” does not mean that 100 units can be promised today. Some units may be under inspection, on quality hold, already reserved, in another warehouse, or unsuitable for a customer’s batch requirement. Stock ready to ship today also differs from supply expected from a supplier or production next week. Show physical on-hand, saleable physical stock, reserved stock, unreserved saleable stock, planned receipts and candidate promise dates separately. If sales, planning and warehouse staff attach different meanings to the same “available” field, a new system merely speeds up disagreement.

Available-to-promise (ATP) assesses supply and demand for a delivery commitment. Oracle says its ATP rule can specify the supply to search, such as on-hand, purchase-order supply or factory work orders, and the demand the rule can fulfil. Oracle: Available-to-Promise Rules. An ATP candidate is therefore not automatically a physically picked item. If the promise relies on future supply, decide who re-evaluates affected orders and tells customers when a supplier or work order slips.

As a conceptual starting point, saleable physical stock equals physical stock less quantities unavailable for quality or safety reasons; unreserved saleable stock equals that result less effective physical reservations. A future-date promise adds supply expected by that date and subtracts demand ahead of the new order. Actual equations depend on product configuration, dimensions and timing. Do not treat every open purchase order as certain supply. Determine what confirmation and production-capacity evidence is needed before giving a firm date.

Specify every availability-changing event

Receipt confirmation, inspection release, transfer, stock adjustment, new order, reservation, release, pick, shipment, return, production completion and scrap can all affect availability. For each event record its origin, event ID, item/site/batch dimensions, quantity, timestamp, duplicate-message handling and compensating action. Use an idempotency key so that replaying the same message does not count it twice. The accounting stock update and promise availability may have different timing; state the acceptable delay per process.

In Microsoft’s soft-reservation example, multiple order channels query a common available-for-reservation value. A successful request raises the soft-reserved amount and lowers what remains available; later integration can offset the soft reservation as an ERP hard reservation takes over. The transferable design lesson is to give concurrent channels one contention point and avoid double counting during the transition. Microsoft: Inventory Visibility reservations.

Model exact states from order to shipment

State names alone are not sufficient. Define entry and exit conditions, permission, quantity changes and audit evidence. The following names are business concepts, not standard labels promised by a particular product.

Partial reservations and split promise dates are normal

Suppose an order line requests 100 units: 50 can ship today, 30 depend on completion next week and 20 have no reliable date. One “reserved” flag on the whole line wrongly suggests all 100 are secured. Keep supply commitments as separate line segments with quantity, source and promise date. This example is illustrative. If partial shipment, substitute items, interplant transfer or customer revisions are common, keep a revision history rather than overwriting the first answer.

When order quantity falls, specify which commitment is released: the last future segment or the exact date the customer cancelled. That choice changes the date on which another order may use the released supply. When quantity rises, evaluate only the increment while preserving the original firm commitment. If the increment cannot be promised, show “original quantity firm; increment unconfirmed.” A release-all/reacquire-all implementation may let competing orders take the original supply between steps unless the whole operation is atomic.

For quotations and web orders, soft reservations need explicit expiry. Test the case where an approval or payment arrives after expiry and release. Record cancellation as a reasoned reversing action with timestamp and actor rather than deleting the prior event. A cancellation before pick, one during warehouse work and a return after shipment are different processes; do not force all three through one generic release API.

Choose when to fix batch and location

An order can promise ten units from a warehouse without immediately assigning a shelf or pallet. When warehouse waves and first-in-first-out picking should choose the physical goods, the order might fix site, warehouse and stock status while the WMS assigns location and handling unit at pick. If the customer requires a batch, serial number, expiry condition or certificate, those dimensions may need commitment earlier.

Microsoft’s warehouse-level reservation documentation distinguishes a hierarchy in which demand carries site, warehouse and status while warehouse logic selects location, from a hierarchy in which batch or serial is specified by the customer demand. Microsoft: Flexible warehouse-level dimension reservation policy. Fixing dimensions too early can force release and re-reservation for every bin move. Fixing too few may let the WMS pick stock that fails the customer requirement. Include substitute-batch approval and notification in the transition model.

Our handheld inventory integration guide explains scan timing and stock movements. For allocation, the acceptance question is whether the scanned physical unit can be traced to the reservation and order line, and whether a wrong pick returns a meaningful exception. Link reservation ID, pick task and shipment line; a successful scan alone is not enough.

Assign authority across ERP, WMS and MES

It is normal for ERP and WMS both to show an inventory figure. It is unsafe when they update the same promised quantity under different rules without an agreed system of record. ERP or order management commonly owns customer orders and commercial promises; WMS owns storage locations, picking and actual warehouse movement; MES or production management owns work orders and completion evidence. Existing architecture can differ, so assign an authority to each event instead of assuming product names dictate responsibility.

When two or more channels promise stock immediately, route them through one atomic decision point. Independent local stock copies can each sell the last unit. Decide whether a network outage means “take the order without a confirmed date” or “stop acceptance.” An undefined “sync later” policy cannot protect the customer promise.

Prevent double subtraction across soft and firm reservations

An order soft reserved in a sales portal may later enter ERP. If ERP adds a firm reservation while the soft one remains, one order consumes availability twice. If soft reservation is released before firm reservation succeeds, another order may seize it. Design the transfer as one correlated state change. Keep source order ID, line ID, reservation ID and supply segment ID. FAT must show that retries, timeouts and duplicate messages neither consume nor release inventory twice.

The same issue arises when MES changes planned output into completed stock. Do not add the planned amount and the physical completion as two separate supplies. Specify when the plan is replaced, whether inspection must pass before stock is saleable, and how the lot is linked to customer orders. A delayed WMS shipment message also needs a protected in-progress state so ERP does not release goods already being picked.

Design release and reallocation as daily work

Cancellation and delivery-date changes are routine. If only an administrator can release stock from a hidden screen, sales cannot use genuine availability and the warehouse accumulates obsolete holds. Record a reason such as customer cancellation, quantity reduction, expiry, quality hold, priority change, discrepancy or accepted substitute. Preserve the original promise date, approver and notification. Decide by item class whether freed stock flows automatically to the next eligible order or a planner reviews the candidates.

Priority can depend on contract commitments, line-stop risk, export cut-off, batch requirements and other business rules. Separate inviolable constraints from priorities that an authorised person may override with a reason. Show the affected orders, quantities and promise-date changes before a reassignment is committed. Removing a previously communicated firm promise should require sales approval and a customer-notification record, not just a new stock calculation.

A quality hold should identify every order expecting the affected batch. The business may release those reservations or pause them in an exception state while alternatives are reviewed. When the original batch later becomes available, avoid leaving both original and replacement stock promised to the same order. A stocktake shortage needs a similarly explicit rule for which commitments remain and which customers need a revised date. Recording an order against unavailable stock may be legitimate, but label its date unconfirmed rather than making negative stock appear available for physical allocation.

Compare implementation options by behaviour and responsibility

An existing ERP module, package, WMS integration, shared reservation service or custom application may all be reasonable. “Supports reservation” can mean physical stock only, ordered supply as well, a fixed batch, a partial quantity or a multi-site commitment. Group items into standard production, customer-specific, expiry-controlled, service and prototype classes. For each, specify the minimum reservation dimension, release rights and promise horizon before evaluating products.

Test the existing ERP’s standard capability first if it is already the order and inventory authority. If several channels take concurrent orders and ERP receives them with delay, evaluate whether a common soft-reservation layer is required. If WMS owns real batches and bins, avoid forcing ERP to fix every location prematurely; send the customer’s constraints and let WMS choose an eligible unit at pick. When MES supplies future-output estimates, distinguish plans from confirmed completions.

Compare cost on a common scope: licences or development, master-data cleanup, interface changes, handheld process changes, training, exception monitoring, support and future site expansion. “ERP integration included” says little until event ownership and acceptance evidence are specified. The RFP should make the boundary visible so vendor estimates are comparable.

Write an RFP with testable scenarios

Ask each vendor to run the same anonymised sample data: items, units, warehouses, batches, expiry dates, stock statuses, order lines, requested dates and expected supply. Include these scenarios:

1. Two channels simultaneously request the last available unit; only one receives a firm promise.
2. A customer-group quota is set, then consumed by two partially reserved orders.
3. A shortage is promised from future production, which later slips.
4. Part of an order is cancelled and the released supply appears as a reallocation candidate.
5. A customer-designated batch goes on quality hold and an alternative requires approval.
6. WMS picks a batch different from the requirement and shipment is stopped.
7. An ERP interface message is replayed without double reservation or release.
8. After an outage, an expired soft reservation meets a late approval.

For each scenario, specify input, expected state transition, visible quantities, emitted events, logs, permissions and recovery action. Require vendors to distinguish standard capability, configuration, custom development, workaround and unsupported behaviour, and to demonstrate the screen and data/API evidence. Reservation hierarchy, future supply and quota-to-order consumption deserve particularly precise answers.

FAT and SAT prove different boundaries

FAT checks the logic and interfaces in a controlled environment: concurrent requests, partial commitments, release, retry idempotency, permissions and audit logs. Simulate a timeout where the other system completed the action despite the missing response. A correct-looking screen is not enough if the underlying quantity was subtracted twice. Reconcile before and after by item, site and batch.

SAT runs the real process: order entry, Thai shop-floor users, handheld picking, shipment, quality hold, MES completion, approval and customer notification. Check local languages and date/time formats, work crossing shifts, network outages and external warehouses. An API that passed FAT can still fail when real labels and batches enter the warehouse. Accept with a traceable reservation ID, quantity and event history rather than a subjective “looks usable” judgement.

Migrate open promises, not just stock balances

Before cutover, inventory open order lines, already communicated promise dates, manual holds, quality blocks, ordered supply, production plans and customer-specific batches. An Excel entry labelled “reserved” may represent a firm customer promise, a group quota or a sales note; classify it before migration. Moving every ambiguous entry into firm reservation distorts availability on day one.

The reconciliation sheet should separate physical stock, hold stock, quota, soft reservation, firm reservation and unconfirmed demand by item, site, batch and status. Name the approver and deadline for discrepancies and the point when the old system becomes read-only. New orders working on launch day do not compensate for lost commitments from the old system.

Measure promise quality after launch

Alongside stock accuracy, measure promise-date changes after confirmation, double promises, expired soft reservations left active, time from release to reassignment, unresolved exception holds, and reservation-to-pick mismatches. Define the denominator consistently: all order lines, only firm commitments, or a selected item group. Set targets after observing the factory baseline; do not apply a generic external improvement percentage to your operation.

Start with products where a broken commitment hurts most: customer-specific parts, unstable supply, constrained batches or expiry-sensitive goods. A narrow initial scope is useful, but code systems, event IDs, audit and permissions should allow later expansion. Shared meanings for each state and named exception owners matter more than another dashboard.

FAQ

How does order reservation differ from inventory management?

Inventory management records what physical goods exist, where they are and how they move. Reservation records which demand has a right to use some of those goods and when that right may be released. A single application may provide both, but the authority and event boundary still need definition.

Can an allocation quota alone confirm a customer’s delivery date?

Usually it cannot prove a promise to a particular order. A quota protects supply for a group. The order’s quantity, date and constraints must be evaluated, and the quota-to-reservation consumption must be recorded. Product terminology varies, so test the actual transitions.

Is released stock immediately available to another order?

That depends on its physical stage. Unpicked stock may be reusable after an authorised release. Picked goods, quality-held stock or shipped goods need WMS cancellation, inspection or a return before they are saleable. A screen-level release is not evidence of physical availability.

Should ERP or WMS own the reservation?

Separate the commercial promise to an order from selecting a physical bin and recording shipment. ERP or a shared service may own the first; WMS may own the latter. Decide from your existing standard functions, number of channels, batch constraints and latency tolerance.

What should we prepare before requesting proposals?

Prepare representative open orders, saleable and held stock, manual reservations, future supply and communicated dates. Map states from entry to shipment, change/cancellation approvers and ERP/WMS/MES events. Use the same map for the RFP and acceptance tests.

Conclusion: define promise states before buying inventory features

A manufacturing inventory allocation system becomes useful when it connects quota, ATP candidate, soft and firm order reservations, warehouse work, shipment consumption, release and reallocation. Start with the item classes, promise rules, authoritative quantities and exception owners. Give vendors identical scenarios and verify quantities and history in FAT and SAT. That makes a delivery commitment testable rather than merely visible on a screen.

TOMAS TECH can help factories in Thailand map order, inventory, warehouse and production responsibilities and turn reservation and release cases into an RFP and acceptance plan. You can contact us while the implementation scope is still being defined.

Sources

• Microsoft Learn: Reserve inventory quantities
• Microsoft Learn: Inventory Visibility reservations
• Microsoft Learn: Inventory Visibility inventory allocation
• Microsoft Learn: Flexible warehouse-level dimension reservation policy
• Oracle Supply Chain Planning: Available-to-Promise Rules