During Production Inspection in China: When DUPRO Is Worth the Cost

A DUPRO is worth paying for when it can change the unfinished order

A during-production inspection is worth considering when a failure can spread across the unfinished order, the relevant operation and representative output will be visible, and a competent inspection can produce evidence early enough for an authorised decision owner to contain affected work, correct the cause and verify new output.

That is a stricter test than “the factory is 30% complete.” A percentage says little about what the inspector can see. Thirty per cent may be useful if every mould cavity is running and early finished units exist. It may be useless if the critical coating has not started, one colour is absent, or the inspected goods were prepared separately from the production line.

A during-production inspection is often called DUPRO, DPI, inline inspection or in-process inspection. Provider definitions vary, but the useful idea is consistent: examine relevant work in progress and early output before the order is complete. ISO's current inspection-body standard describes inspection broadly as examining materials, products, plants, processes or work procedures and determining conformity against stated requirements. It also emphasises competence, impartiality and consistent operation for inspection bodies (ISO/IEC 17020:2026).

The result is still bounded. A DUPRO records what was observed at a named site, on a named date, against a defined scope. It does not guarantee the unobserved balance, the completed lot, the delivery date or Australian compliance. Final inspection, product-specific testing and the buyer's release decision remain separate gates.

If there is no useful production event to observe, no controlled method, no competent resource for that method, no authority path for an adverse result or no recovery window, do not book the same visit and hope it becomes useful. Move the checkpoint, change its scope or select a control that can answer the actual decision.

Decide with exposure, not a standard percentage

Current inspection services do not even use one completion window. SGS describes its DUPRO service at 30–50% complete, while QIMA describes during-production checks across a 20–80% range. Those ranges describe provider practices. They do not identify the last useful control point for a particular material, process or defect.

Start with the failure you need to prevent from multiplying. Then ask when it becomes observable, when correction becomes difficult and how much recovery time remains.

This is OPL's operational decision method, not an industry formula. The fee should be compared with the credible exposure the checkpoint can still change: affected units, consumed material, rework complexity, schedule loss, downstream testing, freight and destination handling. Do not invent a standard order-value threshold. A low-value order with a safety-critical failure may justify stronger controls; a high-value order may still be poorly served by a visit that cannot observe its main risk.

Make that comparison explicit before booking. Put the inspection fee and plausible recheck cost beside low, base and high downside ranges if the failure continues to the next gate. Build those ranges from the suspect quantity, recoverable unit exposure, material already consumed, likely rework or scrap, schedule recovery, required retesting and downstream costs that would arise only after shipment. Do not present this as expected-value mathematics unless the probabilities are defensible. The visit is not cost-effective merely because the order value exceeds the fee; it needs a credible path to change one of those consequences.

Choose the production event the inspector must witness

The booking instruction should name an event, not only a date or percentage. The useful event is where the risk becomes visible while the decision remains reversible.

First output from a new or changed process

Use this checkpoint when a variable has changed: a new site, line, tool, cavity, material batch, subcontracted step, specification revision, packaging source or production method. Ask the factory to confirm that the changed input or process is actually running and that normal output—not a prepared display sample—will be available.

For a moulded part, the event might be first stable output from every cavity after a tooling repair. For a sewn product, it might be completed output from the first normal cutting and assembly batches across every size in scope. These are hypothetical examples. The right event depends on the product, process and failure mode.

Representative output across variants

One good unit cannot show variation across two lines, four cavities, three shifts or several component lots. Define the sources of variation before the visit and require the report to identify which were observed.

Representative does not mean “statistically representative” by default. It means the planned coverage includes the distinct production sources relevant to the risk. If the buyer also needs a formal lot-acceptance decision, the lot, random-selection method, sample plan and acceptance rule must be separately defined.

Verification after a known corrective action

A mid-production recheck can test whether a correction is controlling new output. It should identify the original nonconformity, the affected population, the action taken, when the corrected process began and which new units were checked.

Do not assume that repairing selected units removed the cause. The material, tool setting, instruction, fixture or process control may also need to change. The report should not declare a systemic issue closed merely because a small prepared sample now looks acceptable.

Confirm the inspection is ready before anyone travels

A visit can fail before the inspector measures anything. The wrong site is operating. The named SKU has not started. Only one colour is present. The critical component is unavailable. The test fixture is missing. Production is almost finished, so the factory has no realistic correction window.

Require a dated readiness confirmation covering:

• the legal seller and actual production address;
• the line, operation or subcontracted site that must be observed;
• current quantities at raw-material, work-in-progress, finished and packed stages;
• the SKUs, colours, sizes, cavities, lines, shifts or lots available;
• the critical operation's planned start and stop times;
• the controlling documents and approved references at the site;
• required instruments, fixtures, power, consumables and competent operators; and
• the unfinished quantity and schedule remaining after the visit.

Production photographs or a video call may help confirm logistics, but they are evidence with limits. They may not establish material identity, random access, measurement accuracy or what happens outside the frame.

If the readiness conditions are false, the result should be blocked or scope incomplete. It should not become a pass simply because the inspector found a few finished units to photograph. Record what was unavailable, which conclusion cannot be made and whether the event should be rescheduled or replaced by a different control.

The readiness check should also sit inside the evidence-based production schedule. A technically useful checkpoint booked after the recovery window has closed is operationally late.

Send a decision brief, not “check quality”

An inspector cannot turn an undefined expectation into an objective result. Send one indexed pack that tells the inspector what controls, what to observe and what decision the buyer needs to make.

Start with the current purchase order, legal seller, actual factory, SKU and quantity. Attach the current factory-ready specification, drawing, bill of materials, artwork and packaging revision. If a golden sample is a comparison reference, identify it and state which measurable requirements still control when the sample and document differ.

Then extract the relevant rows from the order's quality-control plan. For each characteristic in scope, define:

• the requirement and tolerance;
• the process stage where it can be observed;
• the inspection or test method;
• the instrument or fixture and any required competence;
• the units, lots, lines, cavities or shifts to cover;
• the defect or nonconformity definition;
• the photographs, readings, records and sample identity to preserve; and
• the action authority if the result is adverse.

This reflects two sound principles without turning either into a universal DUPRO standard. ISO's quality-management principles favour a process approach and evidence-based decisions (ISO: Quality management principles). ISO/IEC 17020:2026 also makes competence and impartiality relevant to inspection-body selection.

If accreditation matters for the product, method or relying party, use the ILAC signatory directory to reach the appropriate accreditation-body register. Then verify the provider's current scope. A logo alone is not a scope check.

For Australian consumer products, the ACCC's sourcing and testing guidance says suppliers are responsible for product safety and should examine factory controls, tested raw materials, followed assembly processes and quality checks. It also warns that testing a pre-production sample is generally not enough without further quality assurance. That supports stronger evidence through production; it does not make a factory inspection a compliance certificate.

Read the report as a bounded observation

A useful report separates what was observed, what population may be affected and what remains unknown. Without those boundaries, a green status can create more confidence than the evidence earns.

What was observed

The report should identify the date and time, actual site, line or operation, production state, controlled revisions, units and records checked, methods used, readings obtained, photographs taken and people who acknowledged the findings.

QIMA's current sample-report description includes quantities of semi-finished, finished and packed products alongside production and product checks. That is useful context, but the buyer's product-specific brief must decide what matters.

What population may be affected

A material lot, tool cavity, machine, shift, subcontracted process or revision start time can help bound suspect output. The report should state the evidence for that boundary. If the factory cannot trace output to the relevant source, say so. Do not silently assume the inspected pieces describe every unit produced before or after them.

What remains unknown

List absent variants, unopened areas, unavailable records, unperformed tests, incomplete packaging and unfinished production. State which later evidence must close each gap.

Keep AQL in its proper role. ISO 2859-1:2026 defines AQL-indexed acceptance-sampling schemes for lot-by-lot inspection. A process observation or a handful of early units is not automatically an acceptance decision for the completed order.

If the report applies a sampling plan to a defined early lot, record the plan and outcome precisely. Use the separate AQL inspection guide for the lot, random-selection, sample-size and Ac/Re questions.

Turn the finding into a controlled next step

The inspector records evidence. The buyer's authorised decision owner decides what production may continue, subject to the agreement and any specialist advice.

Use one explicit status:

• Continue: the defined checkpoint evidence supports continuing the named work. Later gates remain required.
• Continue with a condition: unaffected work may continue, but a named step, SKU, material or output remains blocked until specified evidence closes the condition.
• Contain and correct: identify and segregate the suspect population, stop the failure from spreading, implement the correction and preserve proof.
• Recheck: inspect new output or correction evidence at a defined event before the affected work continues.
• Hold affected work: do not release the named material, operation, population or shipment while a safety, compliance, baseline or systemic uncertainty remains unresolved.

Every action needs an owner, due date, affected scope and closure evidence. “Factory will improve” is not an action record. Neither is a photograph of one repaired unit if the affected population and cause remain unknown.

Do not assume the report means the supplier must rework, refund, absorb inspection costs or accept a deduction. Those consequences depend on the agreement, facts and applicable law. If the issue persists into the shipment-release stage, use the failed-inspection decision guide to assess containment, correction, reinspection and disposition without treating a report label as an automatic remedy.

Know when another control is better

DUPRO is not the default answer to every quality risk.

• Use a first-off or initial-production check when the key question is whether a new input, tool or setup can produce an acceptable first output before normal production proceeds.
• Use pre-shipment inspection when the decision concerns the completed lot, final assortment, quantities, workmanship, labels or export packaging.
• Use a container-loading check when the main risk is shipment identity, count, container condition, seal or handling during loading.
• Use a competent laboratory or specialist when the conclusion requires chemical, electrical, mechanical, food-contact, safety or other product-specific testing that an onsite visual or functional check cannot establish.
• Use regulatory or legal advice when the decision turns on the meaning of an Australian rule, mandatory standard, ban, licence, certification or contractual remedy.

Where an Australian mandatory standard applies, products must meet it before supply, and importers are among the suppliers covered by that obligation (ACCC Product Safety: Product safety standards and how to comply). The exact standard, test method, model and evidence matter. A DUPRO may confirm that a controlled component or process is being used; it cannot substitute for evidence the applicable requirement actually demands.

If the risk is still unclear, first choose the inspection stage by asking where the failure becomes visible and when it becomes expensive to reverse.

During-production inspection booking checklist

Before confirming the visit, make sure the instruction answers all of these questions:

• What exact failure could spread if production continues?
• What production event will make that failure observable?
• Which seller, site, line, process, SKU, revision and order are in scope?
• Which materials, component lots, cavities, lines, shifts or variants must be present?
• What requirement, tolerance, sample or approved reference controls?
• What method, instrument, fixture and competence are required?
• What coverage or sampling instruction applies, and what does it not prove?
• What photographs, readings, records and sample identities must be preserved?
• How will suspect work be identified and contained?
• Who may issue continue, condition, recheck or hold instructions?
• How much unfinished production and recovery time will remain?
• What testing, final inspection and shipment-release gates still follow?

If several answers are blank, the order is not ready for a meaningful inspection booking. Complete the brief or choose a control that can answer the actual question.

Inspect while the decision is still reversible

Do not buy a DUPRO because a template says the order has reached a standard percentage. Buy it when a named failure can still be seen and there is a defined authority path to contain, stop or condition affected work before the failure multiplies.

Identify the risk-producing event. Confirm that normal, representative production will be visible. Give the inspector controlled requirements and an evidence brief. Define who can continue, condition, contain, recheck or hold the affected work. Then keep product-specific testing and final release in place.

That turns a factory visit from a progress photo exercise into a decision checkpoint—without asking it to prove more than it observed.

Sources

• ACCC Product Safety — How to source and test products
• ACCC Product Safety — Product safety standards and how to comply
• ISO 2859-1:2026 — Sampling procedures for inspection by attributes
• ISO/IEC 17020:2026 — Requirements for bodies performing inspection
• ISO — Quality management principles
• ILAC — MRA Signatory Search
• SGS — During Production Check
• QIMA — During Production Inspection Sample Report
• QIMA — The Importance of During Production Inspections