Night-shift errors cluster—when wave, bin, and check are misaligned, training cannot patch structure
The 'high accuracy' on the report is comforting, but when looking at absolute error lines, customer complaints and production line stoppages often cost more than a scanner. What needs to change is not 'being a little more careful,' but reducing the structures that are prone to errors.
What you'll take away
- First, categorize the errors: taking the wrong item, picking too many or too few, mixing up orders, batch errors, and inventory location mistakes. The countermeasures for each are different.
- Wave size is constrained by the number of rows, path duration, and item similarity; bigger doesn't necessarily mean faster
- Similar SKUs being adjacent is a hotbed; it must be countered with isolation, reinforced labeling, and forced scanning.
- The review is conducted by sampling or full inspection based on customers, goods value, and historical errors, with a fixed annual proportion likely turning into a signing ceremony.
- The denominator for accuracy should be based on outbound lines or pieces, and should be considered together with the FIFO enforcement; it is prohibited to check only the quantity without checking the batch.
A stack of 'looks correct' turnover boxes on the review counter
Forty minutes before the night shift ended, the reviewer opened a box: the item number was correct, the quantity was correct, but the batch was not the one assigned by the system. The picker said, 'It's easier to take from there'; the team leader said, 'The customer is waiting, send it out first'; the next day a customer complaint came in, but the report still counted this order as 'accurate'—because only the quantity was checked at the time.
If accuracy is a structural design issue, emphasizing responsibility will only result in a seasonal curve of 'dropping a bit when busy.' Large wave sizes, similar items placed next to each other, containers not bound, and checks that are only for signing—if these four structural gaps are not fixed, training is just a placebo.
First, give the error a name
No classification, so the countermeasure can only rely on guesswork. It is recommended to unify the reason codes at the four points of discovery: picking, checking, customer rejection, and production line feeding.
| Type | Common root cause | Priority action |
|---|---|---|
| Picked the wrong SKU | Similar products are adjacent, labels are weak, codes are close | Location Isolation Check Code/Scan Code Enforcement |
| Pick more / Pick less | Unclear packaging specifications and confusing units | Packaging Master Data Count/Weight Recheck |
| linked orders | Excessive wave size, mixed use of containers, incorrect sowing grid | Container bound to orders/waves; grid check |
| Batch / SN Error | FIFO can be bypassed, mixed batches | Batch/SN Mandatory Verification |
| Wrong inventory location entry | Placed on the wrong shelf, warehouse transfer not posted | Shelf Verification Cycle Counting |
| Wrong document | The wave includes canceled lines and versions that have not been updated | Locked Order Effective Version Before Generation |
If the Top type and corresponding storage area cannot be sorted out recently, it means it is still in the 'retraining' stage and has not entered governance. When errors pile up during the night shift, first look at the wave row numbers and lighting labels, then discuss whether people are tired.
Wave is a design; bigger is not necessarily better.
Excessively large waves lengthen the path, increase container mixing and interruption recovery costs, and accuracy and labor efficiency may both drop. Segmentation is at least constrained by these factors:
| Constraint | Approach | Remarks |
|---|---|---|
| Row / Item Limit | Frozen after actual measurement by warehouse type and tool | Breaking the wave beyond the limit, without relying on people to tough it out |
| Path Duration | Control within continuous operation section | Overtime waves should be able to interrupt continuous picking and have a clear status |
| Item similarity | High-confusion SKUs and the same task require strict verification | Avoid putting it into the same task when there is no verification |
| Order Priority | Isolation between urgent orders channel and regular waves | Prevent last-minute orders from disrupting the entire batch |
| Batch strategy | Allocation has been locked according to FIFO/FEFO | Wave optimization cannot reverse open batch allocation |
Sowing or picking: Sowing is suitable for structurally stable and concentrated items; for multiple varieties in small batches, picking or mixing is more common. A two-week pilot to compare errors and route durations is more effective than slogans. When receiving congestion causes simultaneous receiving, shelving, and picking, drastic changes in storage locations can distort wave paths—accuracy projects need to share the same management calendar as scheduled receiving.
Easily confused shelves are cheaper than 'even more careful'
Errors with similar SKUs often concentrate on a few shelves. First make a 'commonly confused pairs' list, rather than relabeling the entire warehouse:
- Physical isolation: If they look very similar and only differ in specifications, do not consider them separated just because there is a single column between them.
- Label reinforcement: Amplify key differences such as specifications, polarity, left and right parts, etc., and align with the system display fields.
- Scan code mandatory: Tracing and high-mixed materials are prohibited from “confirming no sweep”.
- Packaging Master DataClearly indicate the number of inner packages and the hierarchical units, specially for handling picking too much or too little.
- Lighting / Electronic Tags(High-density area): Still needs to be linked with containers and orders; otherwise, orders will still get mixed up.
When the equipment is not easy to use, first address usability: a high scan failure rate will force on-site skipping; once skipping becomes a habit, accuracy projects are basically meaningless.
Review is a risk valve, not a second labor.
The fixed annual sampling ratio, once the pattern is known on-site, causes re-inspections to degrade into a signing formality. Strategies should change according to customer level, value of goods, expiration dates, and historical errors: for example, full inspections for new customers initially, full inspections for high-value items, increased sampling for areas or personnel with high error rates, and gradually reducing sampling in stable low-risk areas. Strategies should be written into the system to avoid team leaders verbally saying, 'Let's be stricter with sampling today.'
Errors found during outbound review need to be written back to the wave and picking tasks, forming a drill-down of "who made a mistake in which storage area and under which wave structure," rather than just recording an overall accuracy rate. Without this layer, next month we will still rely on guesswork for cutting-edge technology.
First stabilize the definition, then discuss the target.
| definition item | Suggestion | Watered-down writing |
|---|---|---|
| denominator | Outbound line or item | Only use the outbound order header, multiple lines are diluted |
| Molecule | Including re-inspection findings, customer returns, and production line material errors | Only count customer complaints |
| Due Date | By delivery date | Postponed until next month to 'disappear' |
| Wrong batch | Accounted as an error | Ignore quality and storage |
| Goal | Decide after the definition stabilizes | First set extremely high goals, then adjust the standards. |
Do not use a single KPI to pit efficiency against accuracy. A more stable approach is a dual-track system: for efficiency, look at average lines per person and path duration, with the safeguard being that the scan skip rate does not increase; for accuracy, look at line-level accuracy and type improvement, with the safeguard being that the average lines per batch do not go out of control; for compliance, look at FIFO/batch compliance rate, with the safeguard being a limit on exceptions. When a safeguard is breached, first reduce batch size and increase checks, rather than increasing penalties—penalties in the face of structural deficiencies will only lead to more skips and subsequent corrections.
System Handles: Task Structure Scan Interception Review Strategy
JEDP-WMS inbound and outbound operations cover wave picking and outbound verification, and support SN/batch/SKU/container code tracking. The key factors affecting accuracy are how tasks are divided, whether incorrect codes are blocked, and whether verification is risk-based—not having an extra paper checklist. Wave generation should follow rules for segmentation, isolate urgent orders, and exclude canceled lines from waves; picking execution should scan location/items/container and block incorrect codes; batch management should be consistent with FIFO enforcement, and batches should not be changed arbitrarily by default; containers should be linked to orders/waves to reduce order mix-ups.
It is not necessary to implement everything in twelve weeks at once: first, freeze reason codes and row-level definitions → then prioritize Top1 type changes for storage location/tag/packaging → next, tighten wave limits and container binding → finally, implement dynamic review and coordinate with batch verification. When storage locations and master data are still disorganized, first enforce shelving standards and cycle counting; when outbound volume is very low or scanning devices are unavailable, first supplement hardware and perform simple checks, without forcibly implementing a complex wave engine. Pure bulk loose material filling is not suitable for a row-level picking model and should switch to measurement verification and tank/batch control. When the effective version of orders is frequently voided, first stabilize order collaboration, then stabilize picking.
Interruptions, Queue Jumping, and 'Universal Skip'
Wave picking design is often ruined by three common on-site habits. First, when a task is interrupted to handle an urgent issue and returns to find that containers and documents don't match, the risk of order mixing rises sharply — interrupted picking must retain the task status and already picked lines, and resuming cannot rely on memory. Second, urgent orders are directly inserted into an ongoing wave, and after the path is disrupted, pickers start 'picking what is convenient first' — urgent orders should go through isolated lanes or independent short waves, not contaminate the entire wave. Third, skip codes prepared for scan failures are set too generously: a single network glitch can become a permanent workaround, rendering accuracy metrics effectively meaningless.
If the rules for the line-side warehouse, consignment warehouse, and finished goods warehouse are different, the wave strategy and verification intensity need to be configured separately. Using a single "warehouse-wide wave limit" table indiscriminately will either slow down the slower warehouses or make high-mix areas too loose. The cost of errors is also different for production material picking versus sales shipment: picking the wrong materials can stop the production line, while shipping the wrong goods can lead to customer complaints and return logistics—the verification strategy should be tiered according to the type of outbound activity, rather than using the same sampling rate as a shortcut.
The accuracy rate of inventory on reports, wave efficiency, and error causes should share the same source. If the on-site PDA has one set of numbers and the office Excel has another, improvement efforts will always get stuck in disputes over the figures. When compared with cycle counts: if the top cause of discrepancies in the same storage area is wrong shipment or receipt, and the top picking error is also taking the wrong item, it indicates that location and label management share a common root cause. Projects should be merged, rather than having separate mobilization meetings for warehouse management and picking.
First, draw an error pie chart
If errors in the past thirty days cannot produce a classified pie chart, any procurement request using the 'upwave algorithm' should be suspended first — the algorithm amplifies the existing structure, including the existing chaos. Once the pie chart is ready, focusing on only the Top1 type for two weeks is usually more effective than launching five picking technologies at the same time. If Top1 is about picking the wrong item, prioritize isolation and scanning; if it’s about order mix-ups, prioritize container binding and wave reduction; if it’s a batch error, don’t talk about efficiency yet, talk about enforcement first. Two weeks later, if Top1 remains unchanged, first check the skip rate and exception permissions, then look into the personnel.
- WMS Warehousing Solution: Wave Picking and Outbound Verification
- First-in, first-out enforcement
- Arrival Appointment and Receiving Rhythm
- Cycle Counting and Account Reconciliation
Want to apply this in your plant?Resources has matching self-assessments and templates—or book a live demo.
FAQ
Quick answers on this topic—still stuck?Contact
Does every order need to be reviewed?
High-risk clients, expiration dates, high value, and frequent historical errors are recommended for full inspection or dual review; low-risk items can be spot-checked, but the proportion and sampling method must be auditable and not predictable on site.
How to choose between seeding and picking methods?
Stable order mix and concentrated SKUs favor put-to-wall (batch/sow); high-mix low-volume often favors pick-to-order (or a hybrid). Decide from pilot error and labor metrics, not slogans.
Can paper-based picking achieve high accuracy?
It can reach a certain level, but it is difficult to trace back to the person and the time, and it is also hard to intercept immediately in case of misclassification/mistakes. The value of the system lies in interception and record-keeping, not just printing tasks.
How to set the accuracy target?
First stabilize the statistical criteria (row level, including in-warehouse findings and customer rejections), then backtrack the target based on claims and line stoppage costs; announcing the target first and then changing the criteria is equivalent to inflating the numbers.
Will accuracy and FIFO conflict with each other?
No. Using the wrong batch is itself a type of error; use verification to ensure compliance, use waves and storage locations to improve efficiency, and do not turn off verification to speed things up.