On this page
- What actually differs
- Where each stage sits in the sequence
- The cost curve: what a defect costs at each stage
- The check types, and what the industry names mean
- Corrective action when an in-line finding happens
- How this differs from AQL — two different questions
- Which mix for which order
- Who inspects what
- A worked example: the same defect, two inspection plans
- How to write the quality plan into the order
- The quality plan
Twenty-four hours before a shipment is due to leave, a carton is opened at final inspection and the problem is obvious: the front placket on a run of dresses has been sewn a few millimetres off-centre. It is a small error, it is consistent across the batch, and it was entirely avoidable — it was visible at the sewing stage on the first garment, and again on the tenth.
At that point the available responses are all expensive. Sort the batch, rework the plackets, re-inspect, move the shipment. The defect was never a final-inspection problem. It was an in-line problem that reached the end because nothing at that stage was looking.
That is the practical difference between inline and final inspection. One acts while a defect is still cheap to correct; the other decides what to do about a defect that already exists. Both are necessary, and neither substitutes for the other.
What actually differs
The distinction is not thoroughness. It is what the check is capable of doing.
| In-line control | Final inspection | |
|---|---|---|
| When | During cutting, sewing and finishing | At or near production completion, before shipment |
| What it can do | Correct the process: adjust, re-set, re-train, re-cut | Detect, contain, and trigger sorting or rework before the goods ship |
| What it costs when it finds something | Minutes of adjustment | Sorting, rework, re-inspection, schedule impact |
| What it produces | A process correction, and a record of it | A lot decision, and a report |
Two phrasings are worth correcting, because both are common in sourcing documents and both understate what final inspection is for.
"Final inspection happens before packing." Not reliably. Final or pre-shipment inspection is often conducted when production is substantially or fully complete and goods are partly or mostly packed — which is precisely the point of a pre-shipment check, since packing and cartoning are themselves things that can go wrong. If you are relying on "before packing" to mean the goods are still loose and easy to correct, confirm that with the inspection protocol rather than assuming it. Where repacking is on the critical path, an inspection that happens after packing is a different event from one that happens before it.
"Final inspection can only detect." More precisely, it cannot prevent the defect from being created — but it can detect, contain and trigger a remedy before the goods leave. Containment is real value: a defect found at the port is a claim; the same defect found in the factory is a sort. What final inspection cannot do is stop the fault from being made, which is why it cannot be the whole programme.
The two are not alternatives. In-line control is how a factory stays inside the standard; final inspection is how the buyer confirms it was met, and how unshipped goods are stopped from becoming a dispute.
Where each stage sits in the sequence
In-line control governs the process
In-line checks are placed where a decision is still reversible. Three natural points.
At cutting. The marker has just been laid and the first pieces are off the blade. A cutting check catches a marker error, an off-grain piece, or a fabric fault that would otherwise be cut into dozens of garments. This is the cheapest inspection point in the whole sequence, because what it corrects is fabric rather than a garment.
During sewing. Assembly is where most correctable defects originate: seam alignment, stitch density, tension, trim placement, measurement drift across a run. Checks here catch a setting rather than an individual garment — the point is not to inspect pieces, it is to inspect whether the process is producing correct pieces.
At finishing. After trimming and around pressing, checks cover what only becomes visible on a finished surface: press marks, surface damage, shade variation across assembled pieces, clean edges and closures. How the manufacturing process runs from design to finished garment shows the full sequence these checks sit inside.
Final inspection governs the lot
Final inspection answers a different question: should this lot ship? It is a sampled decision against an agreed standard — the mechanism is described in what AQL inspection actually governs — and it produces a record of what was inspected, what was found, and against which plan.
It is genuinely valuable, and it is structurally incapable of improving the lot it is inspecting. Everything it finds has already been made.
The cost curve: what a defect costs at each stage
Buyers often compare inline and final inspection as though they were two line items of the same kind. They are not: they sit at different points on a curve that rises steeply as a garment moves forward.
| Where the defect is caught | What the correction costs |
|---|---|
| Fabric on receipt | A piece of fabric, or a conversation with the mill |
| At cutting | Re-cut a piece; no garment lost |
| During sewing | Adjustment of a setting; minutes |
| At finishing | Unpick and re-sew a completed operation; hours |
| At final inspection, before packing | Sort, rework, re-inspect; possibly delay the shipment |
| After packing | Unpack, sort, rework, repack; the packing work is paid for twice |
| After shipment | Recalled goods and the customer's trust |
Three things about this curve are worth stating plainly.
Nothing about the defect changes along it. Only the point at which it was noticed. The same off-centre placket is a two-minute setting adjustment at the sewing stage and a batch-wide repair at the end.
The steepest step is packing, not inspection. Once goods are packed and cartoned, a remedy has to undo work that was already done correctly — and cartons are often already labelled and counted. This is the practical reason the timing of a final or pre-shipment inspection matters more than its sample size.
The late find also has no slack. A defect found during production competes with production time that still exists. A defect found at the end competes with the delivery date, which does not move.
There is a second, less obvious cost. When a lot fails at the final gate, the remedy is usually a full sort, because the defect must be located on every piece rather than corrected at source. A process check at the sewing stage finds the cause once; a final sort finds the symptoms individually, on every garment.
Why the first check in a run matters most
One check sits at the steepest part of that curve: the first-off check. It is not the most thorough check in the sequence, and it is not the one that finds the most defects. It is the one that catches a systematic error — a setting, a marker, a sequence — before it is repeated across a run.
The distinction matters because defects come in two kinds. A random defect, present on one piece, is found by inspecting more pieces and corrected by repairing the piece. A systematic defect, present because something in the process is wrong, exists on every piece made since the last correct setting — and inspecting more pieces only tells you how many.
A first-off check is the cheapest available test for the systematic kind, because it happens when there is exactly one garment to look at. The tenth garment is still cheap. By the hundredth, the cost has moved. At the final gate, the same finding is a batch.
Which is why the question worth asking a factory is not "how many pieces do you inspect at the end", but "what do you check on the first pieces, and who signs them off before the line continues". The answer to the second question tells you more about a quality programme than the sample size does.
The check types, and what the industry names mean
"In-line inspection" is not one activity, and the vocabulary in sourcing documents is not consistent between factories and third-party inspection bodies. Knowing which check is being promised is what makes the promise worth anything.
| Check | When it runs | What it is actually for |
|---|---|---|
| First-off / first-piece check | On the first pieces after a line starts | Establishing that the settings, the marker and the sequence are right before the run proceeds |
| Patrol / in-line check | Periodically through the run | Catching drift — a setting that has moved, a machine that has changed, a new operator on the station |
| In-line audit at a station | On the garment as it passes a defined operation | Verifying one operation or measurement at the point where it can still be corrected |
| Fabric check on receipt | Before cutting | Finding fabric faults while they are still fabric |
| Cutting check | After the first pieces are cut | Finding marker and grain errors before the full run is cut |
| Appearance and finishing check | Around trimming and pressing | Catching surface faults that only exist on a finished garment |
Alongside these sit the inspection-stage names used in third-party protocols, which describe when an external inspector attends rather than what the factory does:
| Third-party stage | Common meaning | Where it sits |
|---|---|---|
| DUPRO (during production inspection) | An inspection while production is running | Mid-run — closest in spirit to in-line control, but performed externally and at a point in time |
| FRI (final random inspection) | A sampled inspection when production is complete | At or near the end of production |
| PSI (pre-shipment inspection) | A check before shipment, commonly conducted when production is substantially complete and goods are partly or largely packed | Immediately before dispatch |
| Container loading supervision | Watching the goods go into the container | At dispatch |
Two cautions. First, FRI and PSI overlap in practice, and the terms are used differently by different bodies — what matters is the condition of the goods at the moment of inspection (loose, partly packed, fully cartoned), not the label on the report. Second, a DUPRO is not a substitute for in-line control: it is a snapshot taken by someone who is not on the line every day, and it cannot adjust a setting that has already been wrong for a week.
One more term belongs on this list but is a different kind of activity: 100% sorting. It is not an inspection stage; it is a remedy, and it is what a failed lot usually becomes.
Corrective action when an in-line finding happens
An in-line check that logs a finding without linking it to an action has the cost of inspection and none of the benefit. The loop is short, and it is the same loop every time.
| Step | What happens |
|---|---|
| 1. Stop and contain | Identify the pieces already affected since the last verified-good point, and hold them. Containment is the step most often skipped, and without it the same defect is found again at the final gate |
| 2. Establish the root cause | Is it a machine setting, an operator technique, a pattern or marker issue, or an incoming material problem? The answer decides who fixes it |
| 3. Correct the cause | Adjust the setting, re-train the operator, correct the operation sequence. Correcting the piece fixes one garment; correcting the cause fixes the run |
| 4. Verify the first corrected pieces | Check the next pieces deliberately, rather than assuming the correction worked |
| 5. Resume, and record | Note what was found, what was contained, what was changed and what was verified. This record is what makes a later final-inspection report interpretable |
Two notes on why this matters commercially.
Step 1 is where most of the money is. A defect contained at the moment of discovery is a repair. The same defect discovered three days later is a sort of everything produced in between, and nobody knows which pieces those are without a record.
Step 5 is what the buyer usually ends up asking for. When a lot fails at the final gate, the first useful question is not "how many defects" but "was this visible earlier, and what was done about it?" A factory with an inline record can answer. One without cannot, and the conversation becomes a matter of opinion — which is where arguments about responsibility usually start.
How this differs from AQL — two different questions
These two subjects are often discussed as though one contained the other. They answer different questions, and a complete quality plan needs both.
| AQL | In-line vs final | |
|---|---|---|
| The question it answers | Given a sample, do we accept this lot? | Where in the process do we look, and what can we still change? |
| Its type | A sampling decision framework, with risk split between producer and consumer | A process-control timing decision |
| What it produces | An accept/reject decision and a report | Corrections, containment records, and fewer defects reaching the sampled gate |
| What it cannot do | Improve the lot, or prevent a defect | Decide whether a lot is acceptable, or split risk reproducibly |
A programme built only on AQL inspects correctly and discovers late. A programme built only on in-line checks has no reproducible basis for the ship/don't-ship decision and no defensible record of it. This is why the two are normally described together: AQL governs the decision at the end; in-line control governs how much is left to decide about.
Which mix for which order
Neither extreme is right for every order. The useful question is not "inline or final" but "how much in-line control does this order justify?" Five factors decide.
| Factor | Pushes towards more in-line | Pushes towards final-focused |
|---|---|---|
| Order size | Larger runs — a process error multiplies | Very small runs, where a full sort is quick |
| Construction complexity | Structured, lined, embellished, multi-component | Simple, single-fabric styles |
| Novelty | A new style, new fabric, new factory, first run | A repeat of a proven style |
| Value per piece | High-value garments, where rework is costly | Low-value pieces, where sorting is cheaper than control |
| Time before dispatch | Tight delivery, with no room for a sort at the end | Generous timelines with slack for remedy |
The pattern is consistent: in-line control earns its place wherever an error would repeat. A defect that occurs once is an inspection problem. A defect that occurs fifty times because a setting was wrong is a process problem — and process problems are what in-line checks exist to catch.
A working default
Most well-run programmes do not choose. They run a sequence with checks at each stage and a sampled decision on top:
- Material check on receipt — before cutting.
- Cutting check — first pieces, before the run proceeds.
- First-off and patrol checks in sewing — at the start of the run and periodically through it, watching the process rather than the pieces.
- Finishing and appearance checks — after trimming and pressing.
- Final sampled inspection — the lot decision, at or near completion.
This is the six-step structure we run, and its logic is the curve above: put the checks where correction is still cheap, and keep the final gate as the decision rather than the discovery.
Who inspects what
Inspection is performed by different parties at different stages, and the most common source of dispute is that two parties both believed the other was looking.
| Stage | Normally performed by | What the buyer needs from it |
|---|---|---|
| Fabric on receipt | Factory | Confirmation that the approved fabric and reference were checked |
| Cutting | Factory | Assurance that the marker and grain were verified before the run |
| In-line sewing and finishing | Factory | The in-line record, including containment and corrections |
| Final random inspection | Factory, buyer, or a third party | The report, the plan it was conducted against, and the decision |
| Pre-shipment review | Buyer or third party | The document set, the packing condition, and any open items |
The split is not arbitrary. In-line checks sit inside the factory because they correct the factory's own process in real time; a third party arriving mid-run cannot adjust a setting that was wrong a week ago. Final inspection is where outside eyes are most useful, because it is the last point at which goods can be stopped before they ship — which is why third-party final inspection is a common arrangement, usually at the buyer's cost.
Where it goes wrong is when the split is left unstated. A buyer who assumes the factory inspects every piece, or a factory that assumes the buyer's inspector covers a defect class that was never agreed, both end up in the same conversation after the carton is sealed. Naming who does what, at which stage, against which reference, costs one paragraph in the specification.
A worked example: the same defect, two inspection plans
A run of lined dresses has a recurring fault: the lining hangs slightly longer than the shell at the hem, reading as a soft fold. It can be corrected by re-positioning the hem before the final close — but only if it is found before the hem is finished.
Plan A — final inspection only. The fault is found at the final gate. By then the hem is closed on every garment, so the correction is a re-open and re-sew across the batch, followed by re-inspection. The lot passes on the second attempt, several days late, and the schedule absorbs the cost.
Plan B — first-off and patrol checks at the lining stage. The first-off check catches it on the first dress, before the hem is closed. The correction is a setting: the lining is re-positioned, and the operator is shown the approved relationship between shell and lining. The affected pieces since the last verified-good point are contained and checked. A patrol check later confirms the correction held. Nothing is re-opened, and the hem never becomes a repair.
The defect is identical in both plans. What changed is where it was noticed — and therefore whether the remedy was a setting or a batch. That is the whole economics of in-line control, and the reason checks belong at the stage that creates the fault rather than at the stage that ships the order.
How to write the quality plan into the order
The plan does not need to be long. It needs to name seven things, and it should travel with the tech pack so it governs the sample, the bulk and any repeat.
| # | What to write | Why |
|---|---|---|
| 1 | The checks — which stages have an inspection point | Makes the sequence explicit rather than assumed |
| 2 | What each check covers — fabric, cutting, sewing, finishing, appearance | Prevents gaps between stages where nobody claims responsibility |
| 3 | Who performs each check | Removes the most common cause of dispute |
| 4 | The reference each check uses — measurement chart, approved sample, defect classification | A check without a reference is an opinion |
| 5 | The corrective action loop — who contains, who establishes the cause, who verifies the fix | Turns a finding into a correction rather than a note |
| 6 | The final inspection plan — standard, edition, level, values per class, and the condition of the goods | Makes the lot decision reproducible; see what AQL actually governs |
| 7 | The remedy if the lot fails — sort, rework, re-inspect | Agreeing it in advance turns a failure into a process step |
Two lines buyers forget
The condition of the goods at final inspection. Whether the assessment is done on loose garments, partly packed goods or sealed cartons changes what is practically correctable and what the report means. Naming it removes a whole class of argument.
What happens between the checks. The gap between an in-line check and the final gate is where defects accumulate unobserved. Where a style has a known weak point — a closure, a trim, a seam in a difficult fabric — it is worth naming a check specifically for it rather than relying on the general sequence. Common defects found in dress production sets out the ones that most often escape a general sequence.
The quality plan
Everything above reduces to one page. It states where the checks are, what they are capable of doing, and how a finding becomes a correction.
| # | Stage | Check | Can it correct? | Owner | Reference |
|---|---|---|---|---|---|
| 1 | Goods in | Fabric inspection on receipt | Yes — reject or re-source the roll | Factory | Approved fabric reference |
| 2 | Cutting | Cutting check on first pieces | Yes — re-cut before the run | Factory | Marker, grain, approved sample |
| 3 | Sewing | First-off check | Yes — settings, sequence | Factory | Approved sample, construction sheet |
| 4 | Sewing | Patrol checks through the run | Yes — drift correction | Factory | Same, plus the in-line record |
| 5 | Finishing | Appearance and finishing check | Partly — surface corrections | Factory | Defect classification |
| 6 | Pre-dispatch | Final random inspection | No — detect, contain, trigger remedy | Factory / third party | Sampling plan, classes and values |
| 7 | Any stage | Corrective action loop | Yes — contain, cause, verify, record | Factory, reported to buyer | The plan above |
The difference between this and most inspection documents is row 7. Without it, the plan describes where to look, and row 6 describes what to do when the looking found something too late.
If you want an inspection plan agreed before production rather than improvised at packing, send us the style and your quality requirements. Quality control covers how we run the sequence, the production process shows where each check sits, and what to send an OEM clothing manufacturer covers the inputs that make the plan answerable. Where a defect is best understood as a specific construction fault rather than a stage problem, what a garment quality programme should check is the companion piece.