Pick face empty while reserve is full? Replenishment must join the wave—not the keeper’s eyeballing

Total stock is enough—the pick face still runs dry

The accuracy rate of the morning inventory is okay, and the night shift scans the empty picking positions. The warehouse manager said it was replenished during the day but been emptied by a large order; where is the replenishment order? Voice message on WeChat. The storage location stock keeper said it was replenished, but the system didn't transfer it—both sides are exhausted, and the picking positions on the books are still empty.

Out of stock is divided into two types:Really lacking(Insufficient funds available in the entire account) andFake leave(The goods are in the storage location, but the picking location rhythm fails). False shortages are treated as 'not enough staff,' and adding people only raises direct picking exceptions, while FIFO and storage capacity both fail. Treating false shortages as real shortages and increasing procurement will also raise inventory and capital occupation—the problem, however, is not in procurement.

Replenishment must be a task object

Without tasks, it is impossible to count 'who is late and which SKU they are late on.' If it cannot be counted, parameters cannot be adjusted. If the completion is written back but still reads the old snapshot, 'the person has completed it, but the system still blocks it' will occur — the site immediately resumes direct picking, and trust is broken through once.

Receiving must be mutually exclusive: if two people receive the same task at the same time, it may result in double supplementation or no supplementation at all. Reassignment should leave a trace; verbally saying 'you go handle that one' does not count as receiving. Failure reason codes must be distinguishable: batch locked, storage location master data error, channel busy, device occupied—using the vague 'incomplete' will only force warehouse staff to continue coordinating via WeChat.

Sizing: avoid both extremes

Too small: tasks flood the screen, forklifts run empty. Too large: picking locations are overcrowded, overflow returns require secondary labor.

Start with this working formula (illustrative, not the only algorithm):

Replenish qty ≈ max(min hold − on-hand, wave projected gap), then trimmed according to the maximum capacity of the picking location. When there are packaging specifications, round up according to full boxes/full pallets to avoid half-box tasks flooding the screen.

Grading by ABC: A is more diligent and faster, C is larger batch and lower frequency. Using the same safety stock for the entire warehouse is a classic reason for replenishment project failures. In the first week, first ensure 'there is a trigger,' and in the second week, adjust parameters based on empty scans and task density — getting it right in one go is almost impossible, and continuously adjusting two types of parameters makes it impossible to attribute the cause.

Inventory pending inspection must not be replenished into pick locations. Status filtering is more important than 'scanning for quality check an extra time during replenishment': if items pending inspection enter the picking surface, the wave will amplify the risk to every customer order.

Wave engine vs replenishment—who goes first

Recommended order:

1. Generate/Pre-allocate wave demand;
2. Calculate the picking location gap and issue replenishment;
3. Release the picking task after replenishment is completed (or in case of exception release).

If picking is done before replenishment, empty scans will become the norm. Warehouses that allow 'picking while replenishing' must mark dependencies on tasks: picking tasks for SKUs that have not been replenished remain frozen or have lowered priority. Exceptions must be recorded with the reason and authorized person—otherwise, 'picking while replenishing' will degenerate into permanent direct picking.

Multiple consignor/consignment picking locations: The replenishment source batch must include the owner; it is forbidden to use owner A's stock to replenish the picking surface for owner B. If the owner is wrong, it is harder to correct afterward than if the quantity is wrong.

Dynamic pick faces work the same: trigger on “wave demand − current allocation < 0”; the object is still a task, not a verbal ask.

Direct pick: OK as rescue, not as default

Allow storage location direct picking for emergencies, but at the same time meet:

• Authorized roles are limited;
• The system records the reason for direct picking;
• Try to automatically generate post-replenishment write-backs to avoid long-term gaps in picking location accounts.
• Direct selection rate weekly report — consistently high indicates that the parameters or personnel allocation have failed, not heroic behavior.

When direct picking is praised as 'flexible', the master data of the picking locations will deteriorate rapidly: the system thinks there is stock, but the site only trusts familiar people. If direct picking is not written back, the triggers for inventory counting and replenishment will read incorrect picking location levels for a long time.

Three master-data rot modes—fix before blaming people

1. Pick location capacity fill 9999 → Never trigger;
2. Minimum holding amount greater than capacity → task always fails;
3. Storage → Picking transfer relationship not maintained → Cannot dispatch.

In the first week after going live, separately list and clear these three types of errors. Channels and devices are also constraints: concurrent limits for the same lane, off-peak scheduling for large items; failure reason codes should include 'channel busy' and 'device occupied.' When the proportion of master data errors is high, stop parameter adjustments and first clear capacity and relocation relationships — parameter adjustments cannot fix wrong master data.

Draw clear boundaries with other “shortage” talks

Mixing the two sets of 'replenishment', the warehouse management and the workshop will wait for each other. If the inventory window locks the picking surface source for a long time, night shift empty scans will skyrocket—either stagger the windows, or allow restricted replenishment and leave exceptions.

Top SKUs: three metrics in one week

Select the batch of materials with the highest warehouse picking frequency, set minimum holding, and enable automatic replenishment. Then look at: the number of empty scans at the picking location, the direct picking exception rate, and the on-time completion rate of replenishment tasks. If the empty scans can't be reduced, first check the capacity and batch locking; if the direct picking rate is praised, it indicates that the task hasn't become the default route yet — so don't expand the whole warehouse or add staff for now.

Record the top 3 reasons for failure for two consecutive weeks, and only adjust one category of parameters (for example, only adjust the minimum holding of category A). The three metrics of empty sweeps, direct picking, and on-time rate remain unchanged, and expanding the coverage will only copy the WeChat scheduling to more aisles.

Lot locks, wave splits, and rush inserts

After a replenishment task occupies a batch, that batch should not be silently taken by other waves. If splitting of occupancy is allowed, the task should display 'Partially Locked' along with the remaining quantity; if failure occurs due to being fully taken, the reason code should be 'Batch Occupied' and trigger a source recalculation—do not just show a red 'Failed' and let the warehouse staff guess.

If an urgent order is inserted before the replenishment is completed, the completeness check should be rerun: either increase the replenishment priority of that SKU or split the wave and leave exceptions. Directly releasing the pick for the inserted order and bypassing replenishment dependencies is a common reason for a rebound in the empty scan rate.

Hot and cold zones and large items: When the replenishment path is too long, the minimum holding can be increased according to the 'replenishment lead time,' rather than 'just adding a bit more based on feeling.' If the path time is not included as a parameter, adjusting the parameters will forever swing between running empty and overstocking.

Shifts and replenishment windows

If replenishment tasks are concentrated in the same aisle during peak picking times, empty scans and congestion will occur together. Schedule replenishment windows according to shifts: after wave pre-allocation, prioritize replenishing Class A; Class C can be replenished off-peak. Tasks generated outside the window enter the queue and are not forced to grab equipment immediately.

When there is limited manpower during the night shift, the minimum holding can be increased according to the shift, rather than copying the day shift parameters. If the shift parameters are not separated, the night shift empty sweeps will be mistakenly diagnosed as 'not enough people,' and adding more staff won’t meet the points.

Window conflicts with cycle count

When a picking location is being inventoried, replenishment tasks will fail or mess up the accounts. Choose one of the two rules: either pause replenishment to that location during inventory lock and reroute to nearby locations; or allow 'restricted replenishment during inventory' and force a re-inventory of that location. Without a rule, the night shift will pass the blame back and forth between 'empty scans' and 'inaccurate inventory.'

Before closing the inventory difference, the minimum hold trigger for this SKU can be downgraded to manual confirmation to avoid automatically generating a string of invalid replenishments due to an incorrect stock level.

Definition of replenishment complete

Replenishment 'completion' must include: deduction from the source location, increase to the target picking location, matching batch/owner, and no capacity overflow. A 'verbal completion' missing any step will cause the set completeness check to read a false stock level. PDA confirmation sequence is recommended to scan the source first and then the target; scanning only the target without scanning the source will cause the storage location account to drift.

• WMS warehouse solution
• Wave Picking Accuracy
• FIFO hard gate
• Line-side logistics docking WMS