Rework raised but progress still shows complete? Scrap and rework must enter the WO state machine—not only yield

Yield looks fine—shipments still slip

The weekly quality report shows an increase in the pass rate for each inspection, but the plan completion rate has declined. Breaking down the work orders: many 'completed' ones are actually 'pending rework' that were manually marked as finished, with the rework recorded in another log. The warehouse prepares boxes according to the system, but the physical items are still in the rework area.

If rework/scrap is only a statistical field for quality purposes, the execution chain will pretend that the process is still progressing. Planning, warehousing, and costing are all reading the same lie. When first-pass yield is inflated by 'verbally fixing it before reporting completion,' the direction of improvement will also be skewed—the problem is pushed out of the system rather than being solved.

Disposition is an event—it must change status

Only changing the 'nonconforming quantity' without changing the status is equivalent to telling the plan 'it can still be scheduled further.' Making concessions permanent and without a time limit is equivalent to writing the nonconformance strategy into the master data. Quick rework codes below the threshold can exist, but the status still needs to be changed for a short cycle—relying entirely on word of mouth, progress will definitely drift. 'Return to main process' is an exit condition, not a sixth type of long-term hanging business state.

Rework route: short, hard, gated

Common breakpoints:

• Still follow the original process, omitting necessary steps such as disassembly/cleaning/repair welding;
• The work reporting point is still at the original collection point, and the progress instantly jumps to full;
• Rework is automatically marked as completed without quality confirmation.

Rework routing should be short and strict: the gate should be clear, and after completion, it should return to the designated main process node. Directly jumping to the endpoint is prohibited. Anti-jump stations are also effective in the rework state—otherwise, rework becomes a legitimate jump station path. After key feature changes or jumping stations, the first piece must be forced to go through production again before mass production.

The reporting screen directly shows the next step: which station to go to, what to scan, and who will check it. Only issue quality notifications without changing the UI; rework status will be returned via verbal instructions. Rework must pass re-inspection or quality confirmation before returning to the main routing; 'complete rework and pass the station at once' is a hidden bypass.

Scrap and hold: execution must bite the quality chain

MES standalone scrap: At least freeze execution and inventory disposition. Combined with QMS: NCR isolation conclusions should be able to block station and outbound. MES scrapped, WMS can still ship — is one of the most expensive cracks across systems.

Serialized items: The SN status follows the work order. It is prohibited for the same SN to be in "Completed and Ready to Ship" and "Under Rework" at the same time. Partial rework of batch items: split the quantum batch or mark the work-in-progress batch to avoid the entire batch being considered ready to ship. In case of conflict with the isolation strategy, the stricter freeze shall prevail.

Downgrading and reuse must follow the concession/downgrade process. Inventory attributes should be separated from genuine products, and scanning as genuine products is prohibited. Concessions must be structured with the customer scope, quantity, and term specified, and reviewed upon expiration—concessions without a term will become 'why is this batch still being shipped?' after six months.

Interfaces to planning kit and ship

Materials and capacity occupied by rework orders should be visible in the completeness and commitment calculations—otherwise, the plan will stock and package materials based on 'completion,' and missing parts will be discovered again once rework ends. Pre-shipment task verification: SN/batch must not be in rework or scrap isolation.

When a batch fails the random inspection and triggers the isolation of the entire batch: the in-process items that have not yet been inspected must also go into isolation or be temporarily held, and it is forbidden for some quantities to continue being reported as completed to affect the yield rate. If the promised delivery date is still externally based on the 'system completion,' sales will take the blame for the site — the state machine does not write back to the plan, which is the upstream of a reputation incident.

Location and cost: leave no orphan WIP

Rework items must have a system location (rework warehouse/line-side rework position); scrap items should go to the quarantine warehouse or scrap warehouse. The execution ledger and the storage ledger can be in separate systems, but the location semantics must be aligned—quality may say it is in the quarantine area, while the warehouse may say it is still at the line side, which will inevitably result in ownerless WIP during inventory.

Rework hours by default go into the original work order cost, with responsibility codes divided into Operation/Material/Design. Hours not included in cost, the team tends to 'complete the repair before reporting completion.' If the first-pass yield assessment does not take hidden rework into account, the data will push rework out of the system. At the same time, observe: rework occurrence rate, closure duration, and differences in verbal rework audits.

Customer site rework: still write back to the original work order or create an after-sales rework order, and freeze the corresponding SN/batch before resending. If the site repair is not written back, traceability and quality statistics will be systematically underestimated. Include the boundary between after-sales and MES in the scope description to avoid both saying 'closed'.

Repeated defect codes should trigger a quality escalation into CAPA: the execution chain handles it immediately, and CAPA management handles recurrence. If both sides are missing, the production line will only rework repeatedly, and the first-pass yield looks very good on the report.

Anti-pattern checklist (shop-floor check)

• Verbal rework, system still completed;
• Return to the original process collection point for rework, progress jumps to full;
• Scrapping only changes the quality statistics, WMS can still issue;
• Concessions have no time limit and no customer scope;
• Rework hours are not included in the cost, and the first-pass yield looks very good;
• Rework status is not integrated with shipment verification, and the plan is to prepare boxes according to completed work.

One defect code—close the loop

Select the most frequent defects: judgment → rework status → rework process → reinspection → return to main process. Check the number of false reports after completion, overdue rework work orders, and interceptions caused by scrap being erroneously issued. If the false reports cannot be reduced, do not roll out the 'intelligent rework recommendation' across the entire plant yet.

Re-inspecting and sampling one batch for scrapping: both execution through the station and outbound were intercepted; sampling one batch for concession: the deadline and customer scope are readable, and there is a pending review when it expires. If both samples fail, don’t expand the line yet—expanding the line will only copy verbal rework to more work orders.

Multi-op lines and outsourced rework

If rework involves crossing workshops or outsourcing: the status must still be visible on the main work order, and the outsourcing section is handled using related tasks. Rework outsourced has no back-to-factory re-inspection checkpoint, which is equivalent to giving the 'completion right' of non-conforming products to external parties. Upon returning to the factory, a mandatory re-inspection point is implemented, and only if it passes will it return to the main routing.

In a scenario of process chaining, if the upstream process has reported completion and the downstream process finds defects: the upstream completion should be rolled back or transferred to 'Pending Rework Association,' and the downstream process is prohibited from continuing to pull for completion as if the items are good. If both sides pretend to be complete, WIP and commitments will expand simultaneously.

Cost: outsourced rework fees default to the original WO or a dedicated charge, but must roll up to defect and responsibility codes. Parking fees in “misc” with no defect key leaves quality improvement and supplier claims with nothing to grab.

How to audit hidden rework

Weekly抽certain proportion of 'one-time station' work orders: check whether the physical items have ever entered the rework area and whether the quality records show any signs of verbal rework. Audit differences should be reviewed with the team, rather than just being reported in the quality notice.

If the pass rate rises while the rework rate and audit discrepancies also increase, it indicates that the assessment is rushing data into the system. At this point, first stop the 'pass rate ranking pressure' and first get the state machine and working hours into the system running properly—otherwise, applying pressure will only train people to falsify data.

Concession expiry and traceability

Concessions that expire without review: automatically return to quarantine or block new shipments, and generate a quality to-do task. Upon expiration, only an email reminder is sent without preventing execution, turning the concession into a perpetual pass. The trace query must be able to list the issued SN/batch by concession order number—when a customer complaint occurs and 'who approved this batch and which customer is limited' cannot be found, the frontline will bear the responsibility.

Minimum status-machine set

There is no need to implement complex mesh processes all at once. A minimal set of states is sufficient:To be reworked, Reworking, Pending re-inspection, Scrapped/Quarantined, Concession release; After qualifying, exit the rework state and return to the main process node. These five states cannot run properly; adding more recommendation algorithms just gives a verbal facelift to the rework. Weld the smallest set first, then discuss branching processes — if it cannot be drawn, do not expand the defect code range for now.

• MES manufacturing solution
• Process Anti-Skip Station and SOP
• Non-compliant isolation cold storage stopped at the station
• Scan code to report work collection point