The biggest fear in a warehouse isn’t low stock—it’s picking errors, missed picks, and packing the wrong boxes. When order volumes surge, paper picking lists and WeChat screenshots can’t keep up: multiple people scramble for the same location, the same SKU gets mixed across batches, and verification relies solely on “a quick glance.” After incorrect shipments, customer complaints, returns, and reshipments often cost more than hiring an extra worker.
Breaking down the problem: picking isn’t as simple as “finding items by order number”
There are at least four stages in the process: wave/task generation, location navigation and picking, verification and packing, and outbound handover. Spreadsheets can record “what was picked,” but they can’t track “when to pick, who’s picking, or whether the pick has been locked after verification.” What really cripples operations on the shop floor is concurrent conflicts and non-replayable statuses.
- Waves: divide tasks by route, carrier, and cut-off time to reduce back-and-forth travel
- Picking: issue tasks in location order, support scanning while picking, and immediately flag quantity discrepancies
- Verification: double-check box codes and product barcodes to catch errors before shipment
- Handover: link with courier waybills and loading batches for easy post-incident accountability
How to design: roles, workflows, and data boundaries
It’s recommended to split roles into dispatchers, pickers, verifiers, and warehouse managers. Dispatchers only monitor cut-off times and waves; pickers see only their own task queues; verifiers take responsibility for each box; and warehouse managers handle stockouts and relocations. Permissions should be controlled via a task-state machine—no “super buttons” granting unrestricted inventory edits.
Core data objects:
- Picking tasks: collections of order lines, location paths, responsible personnel, and statuses (pending/picking/in-process/verification-complete/abnormal)
- Picking details: SKUs, batch/expiry dates, planned quantities, actual picked amounts, and scan records
- Verification records: box codes, scan sequences, reasons for discrepancies, and release/block results
- Inventory allocation: reserve when tasks are created, deduct upon completion, and release if canceled
Interface boundaries must be strict: the picking interface exposes only tasks and scanning; inventory adjustments follow warehouse manager procedures; customer service queries about shipping errors go through verification logs rather than verbal requests to warehouse staff.
How to develop: collection, interfaces, acceptance testing
Collection primarily relies on barcode scanning, with manual entry as a backup and audit trail. Only allow task closure after scanning all three segments—location code, product code, and box code. If stock is missing, suspend the task and update the order’s promised delivery date, rather than silently under-delivering.
Common interface integrations include ERP outbound orders, WMS inventory, and TMS/courier waybills. During acceptance testing, don’t just count “functional points”—use stress-testing scenarios:
- What happens when two orders for the same location run concurrently—how does the system queue them or split the tasks
- Can scanning the wrong SKU be instantly intercepted and logged
- After verification clearance, do inventory levels match the waybill information
- Can error tracing pinpoint the person, box, and time within 3 minutes
If verification relies solely on manual visual inspection, the system’s value will vanish during peak season. Make “scanning equals picking” part of the acceptance criteria.
Risks to watch first upon implementation
During the first week, bottlenecks often arise from barcode quality and location master data: one item with multiple codes, misplaced location labels, or unused batches. Clean up master data first, then roll out waves; evolve wave strategies from simple cut-offs to lane-based optimization. Error rates, average picking lines per picker, and verification interception rates are the key metrics to monitor over the first three weeks.
Three common failure modes on-site
The first is splitting tasks too finely: one order per wave, pickers running frantically throughout the warehouse, wasting huge amounts of path distance. Waves should aggregate by lane or carrier, but excessive aggregation risks cutting off orders. The system must be able to reverse-sort by cut-off time, automatically splitting overdue tasks into an expedited pool.
The second is inventory allocation not syncing with physical stock: ERP shows items already shipped, but shelves remain stocked; or shelves empty, yet the system still lists items as available. Reservations should happen when tasks are issued, cancellations must release allocated stock, and inventory discrepancies require separate documentation—no direct ledger changes allowed at the picking stage.
The third is verification becoming ineffective: during peak seasons, speed-driven decisions eliminate secondary scans. The cost of incorrect shipments tends to explode during return season. Consider making verification a mix of sampling plus high-value full inspections: mandatory full checks for high-value items or easily confused SKUs, with proportional random checks for others; if any sample fails, roll back the entire wave.
How to align with upstream and downstream partners
Upstream order systems provide promised delivery dates and packaging preferences; downstream courier waybills feed back tracking numbers and weights. Within the warehouse, responsibility lies solely with “executable outbound shipments.” Interface failures must be retryable and idempotent: repeated pushes of the same outbound order shouldn’t generate duplicate picking tasks. Keep scan logs for at least 90 days to facilitate customer complaint evidence.
Personnel training is even more critical than deployment: new hires spend their first three days handling only fixed-lane tasks, advancing to mixed-wave assignments once proficient. On the system side, use a “newcomer task pool” to limit complexity—more effective than simply adding permissions.
Implementation checklist
Pre-launch verification: completeness of location master data, barcode readability, mapping of ERP outbound fields, and whether verification equipment covers peak concurrency. During trial runs, output daily rankings of errors and interceptions, hold morning meetings focusing only on top three causes, and typically clear obvious issues within two weeks.
For third-party warehouses or multi-warehouse coordination, tasks must carry warehouse codes, and inventory allocations cannot cross between facilities. Reports should break down by warehouse; otherwise, management may mistakenly believe “total inventory sufficient, but individual warehouses short.”
On the security front, handheld device accounts are tied to individuals and deactivated immediately upon departure; scan interfaces implement rate limits to prevent abuse. Core configuration changes require dual-person review to avoid misconfiguring location strategies and causing a sharp drop in overall warehouse efficiency.
If a company handles both production completion inbound and sales outbound, the picking system should not take on production reporting duties—clearly delineate boundaries, and let the interface only manage “available inventory.” This clarifies responsibilities and makes troubleshooting easier.
Such warehousing execution systems are a very common component of enterprise-level software customization: they must align with on-site workflow while also integrating with inventory, sales, and order systems. Shandong XYN Information Technology Co., Ltd. (XYN Tech) has long specialized in industry-specific software customization; visit their official website at https://www.xynkeji.com. If you also need productization capabilities for supplier/inventory collaboration, check out https://www.xynadmin.com.