A reliable on-demand service needs controlled order states

A digital queue cannot describe the physical restart on its own. After a stoppage, the service must reconcile the approved job with printed length, panel position, blank identity, cut-set completeness and cure state. The aim is to resume only the verified remaining scope. Restarting the original job from its first instruction may duplicate panels or garments; skipping ahead without locating the last good article may omit them.

The restart record should identify the interruption, affected job, last verified physical state, quarantined material and authorised next action. Equipment-specific documentation and vendor testing must establish the actual stop and restart procedure. This chapter makes no claim that a particular printer can resume at an arbitrary panel or that an interrupted cure can be continued safely. Where the process state cannot be verified, the affected work remains held until an approved test, reprocessing route or rejection decision resolves it.

Thermal and maintenance dependencies belong in that decision. Garment-printing guidance includes heat curing, and sector guidance identifies the heat demands of drying and curing.[S15][S17] Model-specific maintenance instructions can also require routine ink handling even when production is intermittent.[S31] Those examples support recording process and maintenance state, not transferring one machine's settings or schedule to another.

Exhibit 42 converts common failure cases into evidence and recovery requirements. Each row ends with a controlled release decision. None permits an operator to hide uncertainty by changing the job identifier or starting the whole order again.

The table shows why recovery cannot be only a software function. The release-intent record may prove that a job was authorised, but physical inspection establishes what now exists. The quality record determines whether an affected article can proceed. Operations should retain the original identifiers through quarantine, rework and remake so that the later performance review can count the event and its cost accurately.

Held material should carry its order, job, specification, physical state and reason for hold. Its location and status must prevent accidental return to production. The disposition options should be explicit: release after evidence, rework under an approved instruction, remake through a new recovery intent, return to usable stock where valid, or reject under the applicable waste procedure. Scrapping an item does not close the recovery record. The event still consumed material, labour and capacity, and its cause still needs review.

A restart test should include the awkward cases before the service makes dependable promises. The pilot can interrupt a queued job, delay an acknowledgement, stop after partial printing and remove a panel from a controlled cut set, all under an agreed safe protocol. Success means that the team reconciles the order and physical scope without duplication or omission. It does not mean that the software remained available throughout.

The order remains open after printing because the buyer has not yet received and accepted the promised article. GFT proposes separate events for packed, dispatched, delivered, accepted, disputed and refunded. Each event records quantity as well as status. A parcel scan concerns logistics. Buyer acceptance concerns the agreed product and service. Neither should overwrite the other, and neither follows automatically from a completed print job.

Packing requires reconciliation of the order, size set, quantities, garment identities where used, care information and consignment. If the order ships in parts, each consignment should identify the quantities included and the balance remaining. Dispatch records what left the service and under which carrier reference. Delivery records what the carrier reports as received. Acceptance records the buyer's decision under the agreed terms and window. These distinctions allow a missing parcel, a wrong size and a rejected print to enter different recovery routes.

The programme's performance-management design calls for delivery measurement and accepted-output discipline.[^G02] WP20 applies that discipline to the complete order. A fast dispatch followed by a quality dispute is not a fulfilled success. Equally, a buyer delay in confirming receipt should not be relabelled as a production defect without evidence. The record should preserve the event source and timestamp so that operational lateness, carrier delay and acceptance delay can be analysed separately.

Commercial closure occurs under the service's defined terms. The retained-order state should follow the agreed cancellation, refund and return window, with any open dispute still visible. Net retained sales must reconcile to credits, refunds and remakes rather than remaining equal to the original checkout value. Where an order closes partially, the ledger should show accepted and unaccepted quantities and their commercial treatment. A simple closed flag is not enough.

Proof of closure should be proportionate to the service. A small operation might use a signed delivery record and a dated buyer confirmation. A more integrated service might consume carrier events and an online acceptance response. In both cases, the event needs a source and an owner. Manual entry is not automatically unreliable, but an unattributed status change cannot provide a dependable audit trail.

A return record should help the service correct a specification, process or promise. GFT proposes cause categories for sizing, print appearance, durability, wrong item, late delivery and buyer cancellation, with secondary detail where needed. The category should describe the reported and verified condition, not assign blame. Buyer error is not an adequate technical finding. Approved size chart did not match buyer expectation identifies a different corrective route from wrong size packed.

Cause capture begins with the buyer's report in its own terms. The review then records the evidence used to confirm or revise that initial classification. A sizing complaint may require the garment measurements, approved pattern, size label and intended wearer record. Print appearance may require comparison with the approved artwork and physical reference. Durability may require the care history and relevant test evidence. Fit research cautions that simulation does not remove every physical fitting issue, so a digital approval cannot by itself dismiss a later fit complaint.[S07]

The cause code should point to the record that can change. Sizing links to the size definition, avatar coverage, pattern revision and packing check. Print appearance links to artwork, material lot, print recipe and inspection. Durability links to material qualification, cure and care instructions. Where rubbing performance is relevant, the public ISO scope distinguishes dry and wet rubbing tests, which supports retaining the test condition in the evidence rather than recording a generic colour complaint.[S11]

Some cases have several causes. Late delivery may coincide with a wrong item, and a buyer cancellation may follow an earlier delayed promise. The record can hold a primary remedy and secondary contributing causes without forcing false simplicity. The reviewer should also separate reason for return from disposition. An item returned for lateness may be unused and resaleable under the service's rules, while an item with a durability failure needs technical containment. Commercial treatment and physical disposition are different decisions.

Recurrence review is where cause codes earn their keep. A repeated sizing issue should reopen the relevant technical pack or size communication. Repeated wrong artwork should reopen release readback and job identity controls. Repeated late delivery should test the original capacity gate and promise policy. Each corrective action needs an owner, a due state and a later effectiveness check. Counting returns without connecting them to those records produces reporting, not learning.