How do you assess the storage software?
In the warehouse software presentation, the reader sweeps out a series of labels and the inventory on the screen appears to have changed, and it seems to be automatically managed. To judge whether it can ensure that the accounts are equal, the material and batch is correct. The same consignment goes through the same channel repeatedly, will it repeat the deduction? There's a transfer during the count, whether the difference is a system error or an inconsistency in the reconciliation time point? The assessment should start with these questions that can be reproduced, rather than simply accepting an "accuracy rate."
The total amount is equal. Why is it still possible that the warehouse could be misdelivered?
A hypothetical test example: 100 items of the same material are managed as "packages" and 90 items on the A account; 100 items on the B account and 110 items on the B account. 200 items in the general ledger, and 200 items in physical form, are not valid for either batch, but the system continues to sort out the goods by batch, enforces the EOS rules or traces quality problems. The summary results cannot be a substitute for the correct details.
This means that the inventory reconciliation dimension is determined before acceptance. Usually, a combination of material, owner, warehouse, storage level, batch, serial number, stock status and unit of measure should be determined according to business needs. The frozen inventory cannot be commingled into available quantities, other cargo owners ' goods cannot be used to cover current cargo shortages, and the whole box and individual items cannot be added directly to each other.
The accuracy of the inventory is not a uniform eligibility line for all warehouses. Enterprises should rely on the consequences of the error, the retroactive requirement, the value of goods and the existing baseline agreed indicators. More crucially: whether the same indicator is used in the reports of different manufacturers, with the same denominator, dimensions and time points.
Disassembly the “accuracy rate” into a recalculated indicator
It is recommended that at least separate checks be made on identity, quantity, location, status and processing time limits. Do not compress them into a percentage, otherwise the amount may be offset, the misdirected or the frozen situation error may be concealed.
| Indicators | Recommended statistical approach | The border must be separate |
|---|---|---|
| Line differences | Compare book and physical, record inconsistent lines and total reconciliation lines by agreed dimensions | All accounts and all accounts should be reconciled, and not only the system records are available |
| Absolute volume difference | The absolute difference is calculated by line at the same calibration and compared to the agreed base | Separate units of measure; separate reporting on zero-time basis |
| Position and Status Difference | Check the actual location of the object, its pending or frozen status | Quantity of objects that are at wrong repository, wrong owner or wrong state |
| Inventory submission delay | From the time of the scheduled confirmation, to WMS inventory submission or ERP endorsement | Time collected, time displayed on interface and time submitted for business cannot be mixed |
| Unusual closure | Unusual number of records, backlogs, results of reviews and closure of evidence | The event that was hidden or abandoned by the system is not properly completed |
Using the two batches above, the agreed method, 'sum of absolute quantity differences per line ÷ total book quantity,' gives (10 + 10) ÷ 200 = 10%. Using 'discrepancy lines ÷ lines checked' gives 2 ÷ 2 = 100%. These metrics answer different questions, and neither is a universal industry standard. Keep the calculation in the acceptance report rather than entering only a good-looking result.
It is also suggested that business constraints not offset by averages be set, such as the fact that wrong owner's inventory cannot be used for current orders and that the objects to which release is prohibited cannot be passed in full quantity.
What are the checks between RFID recognition and accounting?
The original RFID readings mainly provide label identification and context. To be a credible inventory input, you should also be bound by the label in turn, the object level, the read area, task matching, business status and inventory submission verification. An error at any point may occur "Equipment read all, stock is still wrong."
- Tie:The label is the single item, box or tray, and whether the binding record is valid? When the same label is erroneously tied to another material, reading can only spread errors more quickly.
- Level:The tray and the box is read at the same time, will it be double-counted? Do you still have old box-to-trail relationships after detaching?
- Border:Are the labels of the adjacent areas entering the current operation? Stay, retreat or return within the corridor, generating several candidate events?
- Operations:Are the objects current documents, the batch and the cargo owner in compliance with requirements, and are the accounts allowed to be checked or frozen?
- Submitted:Is there a single change in inventory for an effective event, is the failure clearly identifiable and is compensation records maintained?
Reads that do not find business documents should not be discarded, nor should they directly generate a change in inventory. The more appropriate treatment is to enter a state of readiness for matching or review, to retain events and reasons for failure. For physical continued passage, controls should be applied to the rules of the field operation, rather than leaving an unprocessed alarm only behind.
Key to dynamic inventory: comparison of book and physical items at the same time
The round of inventory often takes time. Between the beginning and the end, there may be a receipt, selection, transfer library and mission cancellation. If you take the initial booking inventory, it is directly compared to the physical amount at the end, even if each operation is correct.
A regional freeze can be used, first to determine the scope of an inventory and limit associated operations; or a certified dynamic inventory program can be used to keep a snapshot of the inventory and valid changes during the inventory. The former will describe the impact of the freeze, the latter will indicate the time point for physical observation, change attribution and late event processing. Only one full-portrait snapshot will be saved, and this does not automatically solve the problem of different times of the book-to-desk count.
The inventory process also validates the amount of occupied and physical quantities, is not installed, is moving over to the warehouse, does not complete the exit and zero-unit conversion. Microsoft warehousing documents provide a cycle count and inventory discrepancy review process, respectively, and indicate that the result may need to be audited; such treatment cannot be replaced by an interface for automatic stock-up updates.[1]
The following questions may be asked during the evaluation:The inventory operator completes a warehouse disk at 10:00, and the other person removes a box from the warehouse at 10:02, and the count results are uploaded at 10:05. The system compares on what time point? If you do not re-enter the terminal until the next day, will it cover current stocks? Please answer the commercial log and inventory, rather than just show the final quantity.
The event design should distinguish between the time of the business and the time of the system's recording. GS1 EPCIS also clearly distinguishes between eventstime and recordtime; this standard provides a model for events, but does not mean that all WMS must implement EPCIS.[2]
Let the same thing get repeated. Check if the system will double-deduct.
A differentiated acceptance question is that inventory submission has been successful, but the interface is lost and the collection program retransmits the event. A qualified design should be able to query or return the results of the original processing, avoiding double-deductiveness. Only a duplicate reading is hidden in the interface, or only a label is recorded in the temporary memory, which does not prove that the inventory submission is equal.
What needs to be verified is that “the same business transaction duplicates the submission still produces the same treatment”, not that “this label will never be processed again.” The next legitimate release from the same tray should generate new operational identities; and the same re-transmitting network should follow the original event identity.
| Active settings | Check results | Evidence required to be retained |
|---|---|---|
| Repeat the same library event and restart the service before retrying | Inventory changed only once, and the original disposal remains searchable | Incident identity, status of disposal, inventory services and receipt |
| Break the net, cancel the original mission and finally re-transmit the incident. | Re-read the old instructions as mission status is correct. | Time of occurrence, time of receipt, mandate version and reasons for refusal or review |
| Blending into non-owner, non-volume or unbound labels | Unusual, not subject to review by an equal number | List of targets, read-in list and breakdown of final documents |
| Read original tray and sublabel after untraceing the tray | Use of current valid relationships to avoid duplication or omissions | Change in box-to-carry relationship, breakdown of events and basis for counting |
| Two people submit the same subject's conflicting operations at the same time. | Dealing with conflicts by operational constraints, not through final writing | Service results, conflict information and manual processing portals |
WMS and ERP are right. Why is it still wrong on both sides?
RFID addresses the source of data on site, WMS is responsible for warehouse operations, ERP may take over orders, master data and financial accounting. Even if the WMS internal quantities are correct, the parties have different understandings of the owner, batch, unit, document status and synchronization time points, and there are still system differences.
The evaluation should draw from several complete operations: Receipt Bills, WMS Treasury Services, RFID incidents, ERP returns and subsequent write-off records. Check how many boxes are, when part of the goods is received, whether the failed retest is duplicated, and whether return or revocation can be reversed in reverse along with the original document. It cannot only look at the interface returning "successful" but also check the outcome of the recipient's actual business.
If data are late, the statement should indicate whether it is “sync” or “already inconsistent”; if manual adjustments are required, the pre- and post-adjustment values, approving officers and reasons should be retained. Batch replenishments to allow acceptance figures to meet the standards will mask real process problems.
Eventually, the manufacturer will deliver a chain of evidence that can be traced back.
Prepare an expected list of subjects before testing, inventory baseline and operational rules; retain collection events in the test, changes in tasks and inventory services; and then list differences, root causes, manual actions and results. High-risk scenarios should be individually concluded, not diluted by a large number of simple cases.
The assessment can be that: “What tests are completed in designated storage areas, and the object, and the speed and operating conditions, which problems have been closed, which still need to be handled manually and which have not yet been covered.” Such reports can guide procurement and go online more than “support RFID, with a high level of accuracy in inventory”, and also provide a basis for subsequent re-packaging, adding equipment or adjusting processes.
Technical reference
The following documents are used to check the time of inventory process and event concepts, which does not represent a sample of the present document that is measured by the manufacturer, nor does they constitute a uniform acceptance indicator.
- Microsoft Learn:Cycle countingSee the periodic inventory process and review of variances.
- GS1: EPCIS data model and event fieldSee eventstime, recordTime.

Ms. Lei
Mr. Pang