Home › Field notes › Change Control in Bag Production: Material, Line, Facility and Process

Modular Bag Pro · 2026-10-11 · 2,520 words · 11 min read

Change control in bag production is the discipline of declaring every substitution — of material, component, production line, facility or process — before it reaches a bulk run, and proving the substituted output still answers the approved specification. Four classes cover it: material, line, facility and process. Verification has to be completed before the affected goods ship, and release stays on an attribute plan at AQL 2.5 taken from ISO 2859-1 at general level II. A run occupies 35-50 days at 500 pieces per reference, with confirmation pieces built in 6-10 working days and in 12-15 where the work is intricate. Figures are indicative, quoted FOB Xiamen, and cover civilian load-carriage goods only; this is general trade information rather than legal advice.

What Counts as a Change: Four Classes

A change is any alteration to what was approved, whether or not anyone thinks it matters. The test is not whether the alteration is visible in the finished product; the test is whether the input or the method differs from the one that was baselined. A different supplier of the same cloth at the same specification is a change. So is the same cloth run on a different line.

Material changes cover cloth, coating, webbing, thread, foam, hardware and any printed or transferred finish. They are the class buyers watch most closely, because they are the ones that show up in laboratory results — a coating substitution moves water resistance, a thread substitution moves seam behaviour, a plating revision moves corrosion performance.

Line changes move the build from one production line to another. Facility changes move it to a different building with different equipment. Both are invisible in a photograph and both can move dimensional consistency, because jigs, operator habits and machine settings differ even when the drawings are identical.

Process changes alter how a thing is done without altering what it is: a different stitch pattern, a different seam construction, a different bonding temperature, a different assembly sequence. These are the hardest to detect from a finished piece and the easiest to introduce without noticing, which is exactly why they belong on the same register.

Judgement: Record anything that differs from the baselined input or method — same-specification suppliers and line moves included — because the test for a change is a difference from the baseline, never a visible difference in the finished goods.

Change Classes Compared by Verification Depth

Verification depth should follow the consequence of the class, not the convenience of the moment. A swap that cannot move performance needs confirmation; a swap that can move performance needs measurement; a move that changes equipment needs both measurement and a fresh approval piece. Applying the shallowest convenient check to every class is how a silent substitution reaches a customer.

The last row is the one worth reading twice. Each class has a characteristic silent failure, and in each case the failure happens because somebody assumed continuity rather than testing for it. A finish supplied under the same commercial name is not guaranteed to be chemically identical year on year.

Spec rule: Match verification depth to consequence — laboratory re-test for material, a fresh approval piece for a line or facility move, and both for a process alteration — and hold release until the required evidence exists.

Triggers That Are Most Often Missed

Most unmanaged changes are not deliberate substitutions; they are consequences of something else that was treated as administrative. Five triggers account for the majority, and each one is detectable if somebody is looking for it.

A supplier swap at the same nominal specification is the first, and it is the most dangerous because it looks like non-events. A second is a new colourway: pigment load changes coating behaviour, so a dark shade and a light shade of the same finish are not automatically interchangeable in performance terms. A third is a tooling rework — a cavity re-cut or a die re-machined produces parts from a tool that no longer matches the one that was qualified.

The fourth is a component revision inside the same part family. A buckle that keeps its part number while its moulding tool is renewed is a new part for evidence purposes, and the same applies to a slider whose plating bath changes. The fifth is a seasonal repeat with no declared change at all, where finish chemistry or component sourcing drifts quietly between seasons.

Two habits catch nearly all of these. Keep a baseline file naming every input with its supplier and part number, and re-read that file at the start of every repeat rather than only when something appears to have changed.

Registering a trigger costs almost nothing and is the whole point of the exercise. A single line — date, what moved, which references are affected, what verification is planned — converts a substitution from a surprise into a managed event. Programmes that skip the line on the grounds that the change is minor are the ones that later cannot say when a particular input entered the build, and by then the question is being asked by somebody with a claim rather than by a colleague with a query.

Selection rule: Treat supplier swaps, new colourways, tooling rework, silent component revisions and every seasonal repeat as change triggers, and re-read the baseline input file at the start of each repeat rather than waiting for a visible difference.

Verification Required Before a Change Ships

Verification is not one activity; it is a set of activities matched to what the change can affect. The discipline is to decide the set before the change is made, because deciding afterwards means choosing tests that are quick rather than tests that are relevant.

Two conditions attach to every row. Specimens must come from the changed input or the changed equipment, not from stock built before the change — a report on the old cloth says nothing about the new one. And the report has to be in hand before the affected goods are released, not in transit with them.

Where a change affects a claim the brand makes publicly, the retest also protects the listing. A water-protection figure quoted in retail copy is only defensible while the finish that earned it is still the finish being used.

Verdict: Draw specimens from the changed input or equipment, never from pre-change stock, and hold the affected lot until the report exists, because a result that arrives after the goods have shipped is a record rather than a control.

Notification Windows and Who Approves

Notification is where most change-control systems fail, and the failure is usually about timing rather than content. A notification sent after the run is cut is not a notification; it is a report of something that has already happened. The window has to sit before the point at which the change becomes expensive to reverse.

A workable structure has three windows. Material and component changes are declared before the affected run is cut, because after cutting the material is spent. Line and facility moves are declared before the affected run starts, because output from the new equipment needs an approval piece before it can be released. Process alterations are declared before assembly begins, because after assembly the parts cannot be recovered.

Approval authority mirrors the consequence. A change with no performance consequence can be accepted by production engineering and reported. A change with a performance consequence needs the buyer's technical reviewer. A change that moves cost or date needs the commercial owner as well. Nobody should approve a class of change whose consequence they cannot carry.

Silence is not approval. A notification that receives no reply should escalate on a stated clock rather than being assumed accepted, and the escalation route should be named in the change register rather than improvised when it is needed.

Takeaway: Declare material changes before cutting, line moves before the run starts and process alterations before assembly, escalate unanswered notifications on a stated clock, and never treat silence as approval.

Retention Pieces, Traceability and the Golden Reference

Change control is only as strong as the ability to prove what was shipped. Three artefacts supply that ability, and programmes that keep all three can answer questions that programmes keeping none of them cannot answer at all.

A retention piece from each lot, stored with its lot number and dispatch date, is the first. It allows a claim arriving months later to be assessed against the goods that actually left rather than against a description of them. Storage conditions matter: dark, dry and sealed, because a faded or mouldy reference argues against whoever kept it.

Traceability is the second. Each lot should link back to the input batches it was built from — cloth lot, component batch, plating batch — so that a failure confined to one input can be bounded rather than assumed to affect everything. Without that link, every problem is potentially every shipment.

The golden reference is the third and the one most often lost. Sealed, dated and held by a named owner, it is the only artefact that lets a future season be compared with what was actually approved. Where a change has been accepted, the reference is updated deliberately and the update is recorded, never quietly swapped.

Bottom line: Keep a retention piece per lot with its lot number, link every lot to the input batches it was built from, and maintain a sealed golden reference updated deliberately on each accepted change.

What Happens When a Change Ships Without Notice

An unannounced change is discovered in one of three ways, and none of them is pleasant: the buyer spots it at goods-in, a customer returns the product, or a laboratory test run for another purpose fails unexpectedly. By then the conversation is no longer about the change; it is about who knew what and when.

The commercial consequence depends on whether performance moved. Where it did not, the issue is contractual — a term was broken and the remedy is negotiated. Where it did, the issue is a product problem with a recall-shaped cost, and the remedy is whatever the destination market requires rather than whatever the parties prefer.

Evidence decides the outcome in both cases. A change register with dated entries, a retained piece from the affected lot and the input batch record let the discussion be bounded to a specific quantity. Without them, the prudent assumption is that every shipment since the last known-good lot is affected, and the scale of the problem expands to match the uncertainty.

Our production team treats an undeclared substitution as a non-conformance regardless of whether the goods pass inspection, because inspection is drawn against the specification and not against the baseline. Passing an AQL 2.5 check on a substituted input is not evidence that the substitution was acceptable.

The remedy afterwards is procedural: re-baseline the affected reference, re-qualify the changed input, issue a revised approval piece, and add the missed trigger to the register so the same path is closed next time. Doing that once, properly, is usually enough to stop the same class of substitution recurring, because the register entry turns an informal habit into a named item that somebody reviews at the next repeat.

Where the goods have already reached the destination market, the first practical step is bounding the exposure rather than arguing about blame. Lot records, input batch links and retention pieces establish which shipments are actually affected, and a bounded quantity is a manageable problem while an unbounded one is not. Brands that can produce those three artefacts within a day consistently settle such matters faster and on better terms than brands that need a fortnight to work out what was shipped.

Programme Gates, the Register and the Change Record

On the production side, 4,950 m² of SGS-verified floor carries 149 machines across 7 production lines, staffed by 137 people turning out 200,000 units per month. Experience in the trade goes back to 2004 and the company was formed in 2014. Quotations are issued inside 24-48 hours. Confirmation pieces are built in 6-10 working days and in 12-15 where the work is intricate; the run itself needs 35-50 days at 500 pieces per reference. Goods are released on an attribute scheme drawn from ISO 2859-1 at AQL 2.5 — critical defects must be nil, major no more than 2.5 and minor no more than 4.0 — shipped FOB Xiamen under T/T 30/70 terms, and sea then takes 25-35 days, air 5-8 and courier 3-5.

The register itself needs five fields to be usable: the date the change was identified, the class, the affected references and lots, the verification performed with its result, and the approval with the name and date. Five fields, one line per change, held with the reference rather than with the order.

Review cadence should be event-driven with one calendar anchor. Review the baseline at the start of every repeat, and immediately on any of the five triggers: a supplier swap, a new colourway, a tooling rework, a component revision or a seasonal repeat with no declared change.

Change control connects to the rest of the programme: tooling rework and its one-off cost sit in the tooling and screen charge guide, the approval piece a line move requires is described in the pre-production approval process, and material swaps re-qualified against the MOLLE attachment grid need the interface dimensions re-measured. Verification is booked through the testing and inspection services desk and the register for a specific reference is best opened at the enquiry desk.

Five fields per change — date identified, class, affected lots, verification with result, and approval with name — plus a baseline review at every repeat and on every trigger, is what keeps a substitution from becoming a product problem.