How to Build a Luxury Bag QC Process

What a bag quality process is actually for

The purpose of a quality process is not to find faults. It is to find them early enough that the correction costs little, and consistently enough that they can be traced to a cause. A process that only inspects finished pieces will catch faults at the most expensive possible moment, after the leather is cut, the hardware is set and the bench hours are spent. A process with gates upstream catches the same fault while it is still a material decision.

The second purpose is consistency of judgement. Two artisans looking at the same topstitch will disagree. A written standard with a measurement and a reference removes the disagreement, which is worth more than any individual inspector's skill.

Decision point one: incoming material

The first gate sits at intake, before anything is cut. Leather is checked for shade against the signed reference, for substance against the specified thickness, for surface damage, and for hand. Hardware is checked for finish, plating shade and mechanical function. Lining and thread are checked against their references. If material fails here, the decision tree is short: it does not enter production.

• If shade measures inside the agreed Delta E limit under D65, release the lot to storage.
• If shade is outside the limit but bandable, sort into families and allocate one family per order.
• If substance or temper is outside specification, return the lot to the tannery; never cut and hope.
• If hardware plating shade drifts against the reference clasp, quarantine the batch and log the bath identity.
• If damage or scar density exceeds the cutting allowance, downgrade the lot to a style where it can be cut around.

Decision point two: first piece sign-off

Once cutting begins, the first complete piece off the line is built slowly, by the most experienced hands available, and then examined against every line of the specification. This is the single highest-value hour in the whole process. A construction error found here affects one piece; found at final inspection it affects five hundred. The first piece becomes the working reference that inspectors compare against for the rest of the run.

What the first piece must prove

It must prove that the pattern works, that the material behaves as expected under the machine settings chosen, that the hardware seats without distortion, and that the finished piece matches the approved sample in every measurable dimension. Where it does not, the pattern, the settings or the material is corrected before the line is released.

Decision point three: inline monitoring

Between the first piece and final inspection, the process needs eyes at intervals rather than at the end. We take a checkpoint every two working hours at each station: edge finishing, topstitch tension, hardware setting and assembly alignment. Each checkpoint is a small sample, but its purpose is trend detection rather than lot acceptance. A stitch tension drifting over a shift is visible in a two-hourly check and invisible in a final inspection that only sees finished pieces.

Decision point four: final inspection

Final inspection is the gate everyone knows, and it is the one that should be doing the least work. Its job is to confirm that everything upstream held. On a 500-piece lot we draw 50 pieces at general inspection level II of ISO 2859-1 and assess each one externally, functionally and dimensionally. Major defects are counted against AQL 2.5, which accepts three and refuses four. Critical defects carry zero tolerance.

Decision point five: disposition

A refused lot is not a failed programme; it is a decision to be made. Disposition has three routes. Sorting separates conforming pieces from non-conforming ones and is appropriate when the defect is sporadic rather than systemic. Rework corrects the fault on affected pieces and is appropriate when the correction is stable and verifiable. Refusal returns the lot to the bench for rebuilding, and is the right answer when the cause is not yet understood.

The gates in sequence

1. Confirm the specification in writing: dimensions, tolerances, materials, hardware, thread and the approved reference sample.
2. Gate one: inspect and release material against signed references before any hide is cut.
3. Build the first piece slowly, examine it against every line of the specification, and sign it off as the working reference.
4. Release the line and run inline checkpoints every two working hours at every station.
5. Record every checkpoint finding, including clean ones, so a trend is visible rather than remembered.
6. Gate four: draw the sample at general inspection level II, classify every finding, and apply the accept and reject pair.
7. Dispose of the lot: sort, rework or refuse, and write the root cause note before the next lot begins.

Records that make the process auditable

A process without records is a set of good intentions. We keep the material intake report, the first piece sign-off with its measurements, every inline checkpoint log, the final inspection report with photographic evidence, and a retain sample from each lot held for twenty-four months. The retain sample is the piece that settles any later question about what was actually shipped, and it is worth more than any written description when a dispute arises a season later.

A new style versus a repeat order

The process is not identical for every order. A first production run of a new style carries unknowns that a repeat does not: the pattern has never been cut in volume, the material has never been worked at speed, and no first piece reference exists yet. We run that order with a heavier inline presence, checkpoints every hour rather than every two, and a first piece built twice so the second confirms that the corrections to the first were stable.

A repeat order of a style with a clean history can run lighter. The pattern is proven, the first piece reference is in the file, and the stations know the construction. Where ten consecutive lots have passed, the programme qualifies for reduced inspection under the switching rules of ISO 2859-1, which lowers the sample size without lowering the specification. Reduced inspection is withdrawn the moment any lot is refused or any construction change is introduced.

Corrective action: closing the loop

Inspection without corrective action is an expensive way of finding the same fault repeatedly. Every refused lot and every recurring inline finding generates a written note that names the cause, the correction and the date it took effect. The note is attached to the specification, so the next order of that style is built against the corrected document rather than against the one that produced the fault.

Three causes account for most of what we write: a specification that was ambiguous, a material that behaved differently from its reference, and a station setting that drifted out of tolerance. Each has a different correction. Ambiguity is fixed by adding a number, material behaviour by adjusting the intake standard, and station drift by shortening the checkpoint interval until the station is stable again.

What the process costs and what it saves

Inspection is not free. On a 500-piece structured bag lot, a level II final inspection takes about two and a half working hours, and the inline checkpoints add a few minutes at each station across the run. Set against that, reworking one closure on a returned piece costs around USD 4.80 in parts and bench time before freight, and a refused re-order costs more than both. The gates upstream are where the return actually comes from: the first piece sign-off is an hour that protects five hundred pieces.

Building the process onto your own programme

Start by writing the specification down. Most programmes that struggle with quality do not have an inspection problem, they have a specification problem: nobody wrote the tolerances, so nobody can enforce them. Once the specification exists, the gates follow naturally, because each gate is simply a point at which the specification is checked against reality.

The second step is to put independence on the other side of the inspection. An inspector who reports to the same lead measured on output will eventually be persuaded to pass a borderline lot, and no written standard survives that pressure. Independence, even at the scale of a single atelier, is what turns a set of gates into a system that lasts through a busy season.

Standing AQL, major defects

2.5, ISO 2859-1, single sampling, normal inspection

Critical defects

Zero tolerance, one occurrence refuses the lot

Final inspection sample

General level II, 50 pieces on a 500-piece lot

Accept and reject at AQL 2.5

3 accepted, 4 refused

Inline checkpoint frequency

Every two working hours at each station

First piece sign-off

Full specification check before line release

Material intake check

Shade measured to Delta E 1.0 under D65, substance by dial gauge

Retain sample

One piece per lot held for 24 months

Final inspection bench time

About 2.5 working hours on a 500-piece lot

Rework cost reference

Around USD 4.80 in parts and bench time for one closure

Minimum order quantity

500 pieces per style per colourway

Sampling lead time

6 to 10 working days, 12 to 15 for complex constructions

Bulk lead time

35 to 50 days from approved sample

Programme history

Export work since 2014 across 7 production lines, about 200,000 pieces per month

Shipment terms

FOB Xiamen

Frequently Asked Questions

People Also Ask