Build a Handbag QC System That Scales as You Add Styles

How much does quality control cost once a handbag brand runs more than one style? Three programmes illustrate it cleanly. A three-style programme spends roughly USD 60-95 per style per lot: a shared inline visit, a shared final inspection, and lot files that reuse one template. At nine styles the number drops to USD 45-70 per style, because inspections consolidate into combined visits. At eighteen styles it drops again to USD 30-55 per style, but only if the specification architecture was built for it — programmes that copied documents per style typically see the number go the other way, because the cost is in maintaining eighteen versions of roughly the same page. This page prices the three tiers, sets out the four triggers that decide when a style earns more inspection, and gives the document structure that makes the eighteenth style cheaper than the third.

Quality cost per style at three programme sizes

Three Styles, Nine Styles, Eighteen Styles

The economics change twice on the way up, and both changes come from the same place: an inspection visit is priced per day, not per style. One man-day at USD 220-320 covers a lot of 500 pieces comfortably. Two lots running side by side in the same building can often be covered within one and a half days rather than two, because carton selection, opening and closing routines are shared. Where a programme has three lots finishing in the same week, the third lot's incremental inspection cost can fall to little more than the extra hours.

What does not consolidate is documentation. Every additional style carries its own drawing, defect annex and golden sample file, and if each of those is a separate hand-made document, the cost per style stays flat while volume grows. This is the point at which most programmes either plateau or get expensive, and it is entirely a structural question rather than a sourcing one. Eighteen slab copies of the same defect list is not eighteen specifications; it is one specification with eighteen opportunities to be out of date.

Where the Money Actually Goes at Scale

Split the spend and the reason becomes obvious. Roughly sixty per cent of a growing programme's quality cost is people inspecting, twenty-five per cent is chamber and laboratory testing on components, and fifteen per cent is administration — maintaining the files, reconciling versions, answering questions that the documents should have answered unaided. Only the first of those three falls with scale naturally.

That last pair is what separates the two kinds of eighteen-style programme. The one with a shared architecture answers a new style in an afternoon, because ninety per cent of the specification already exists and only the differences need writing. The one with per-style documents spends four days rediscovering decisions it already made eleven times, and usually reaches a slightly different conclusion each time — which is how two styles in the same collection end up with incompatible tolerance bands.

The Four Triggers That Upgrade a Style

Not every style deserves the same depth, and pretending otherwise wastes money on products that do not need it while leaving genuinely risky products under-covered. Four conditions are enough to decide.

1. New construction. A silhouette the line has never made gets an inline visit plus twelve-unit dimensional measurement rather than five, until it has passed two lots cleanly.
2. New material or new component source. A first leather or a first hardware partner gets full component testing: cycling, static pull and chamber work, regardless of how simple the style is.
3. Functional claim. Anything that must perform — a device sleeve, a fit-driven strap, a closure given a specific retention load — gets testing on quantities rather than visual checks.
4. Compliance exposure. Children's items, and anything sold into markets with specific substance restrictions, gets per-component restricted substance verification rather than reliance on a past pass.

Every other style runs the standard set. In most programmes that means roughly one style in four carries extra cost in its first lot, dropping to one in ten in subsequent seasons as constructions and components become proven. The result is a quality budget that grows far more slowly than the collection, which is what scaling actually means in practice. The nine QC practices list is the standard set referred to here, and worth keeping as the baseline layer.

One Specification Architecture, Not One per Style

The structure that keeps the eighteenth style cheap has three layers. A general specification covers everything true of every bag the brand makes: defect classes with examples, measurement conditions, thread and needle defaults, reference-edge conventions, packaging and marking. A style annex carries only what differs: dimensions with tolerances, the six load joins, materials, and the finish or colour notes. A lot file records what happened: the revision in force, golden sample photographs, the inline note and the final report.

• Nothing is repeated. If two annexes carry the same sentence, that sentence belongs in the general specification.
• Revisions are numbered together. Annex R3 against general specification v2.1 is unambiguous; two files each called 'latest' are not.
• The general layer changes rarely and deliberately. A change there touches every style, which is exactly why it should be slow and reasoned.
• The annex is the only layer that changes often. Adding a style therefore means adding a short document, not rebuilding a long one.

A useful test for whether the architecture is working: ask how long it takes to write the specification for a new colourway of an existing style. If the honest answer is under an hour, the structure is doing its job. If it involves opening a previous file and saving it under a new name, the structure is a filing system rather than a specification, and its cost will scale linearly forever.

Tiering Inspection Depth by Style Risk

Three tiers cover most ranges. Tier one is a proven style of proven construction with an unchanged component source: final inspection only, standard sample sizes, no chamber work. Tier two adds an inline visit at 15-30 per cent completion and twelve-unit dimensional measurement. Tier three adds component testing and, for the first lot of anything new, retained-unit destruct testing at six units rather than three.

The rule that makes this work is symmetrical: styles climb tiers automatically on a trigger, and step back down after two consecutive clean lots. Without the downgrade rule, every style drifts permanently to tier three and the savings never arrive. With it, a mature programme typically settles with twenty to thirty per cent of styles at tier three, forty per cent at tier two and the remainder at tier one.

What a Shared File Structure Looks Like

Naming conventions carry more weight than they appear to. One folder per lot, named from four fields — purchase order number, style code, colourway and ship date — means anyone a year later can locate everything about a shipment without opening three systems. Inside it, four items: the revision-in-force annex, the golden sample photographs, the inline note if one was taken, and the final report with one conclusion line at the top.

• One conclusion line per report. Whether the defect family was systemic or incidental is the sentence that changes the next specification.
• Photographs attached, not linked. Linked images break within about two years with tedious reliability.
• Rectification notes filed with the lot. What was fixed, how many units, and who paid, since that pattern repeats.
• Component test certificates filed once, referenced many times. Put them against the component code rather than against each lot that used it.

That last item is a quiet source of savings. A chamber test certificate for a metal component stays valid for as long as the component source and its plating process are unchanged, so it belongs to the component rather than to the order. Filing it once and referencing it from every subsequent lot removes a charge from every repeat purchase, which across eighteen styles is a genuinely meaningful number by the end of a year. Sedex and BSCI audit frameworks both expect exactly this traceability, so the same structure serves a compliance visit at no extra effort.

The Cost of Getting the Structure Wrong

The failure mode is specific and easy to recognise in hindsight. A programme reaches ten or twelve styles, quality cost per style has not fallen at all, and inspection reports start disagreeing with one another because two documents define the same defect differently. Nothing has broken visibly; the brand simply pays tier-three prices for tier-one products, forever, and every new style adds another document whose differences from the others nobody can recall.

There is a second cost that does not appear on an invoice. Undocumented judgements have to be re-made by whoever is in the room, so quality outcomes become dependent on individuals rather than on method. That is tolerable with one person checking three styles and unmanageable at eighteen, and it is the reason some brands find their quality inexplicably declining just as their volumes justify doing the work properly. The remedy is unglamorous — write the general layer once, properly — and it is considerably cheaper than discovering the gap in a container. How each defect gets classified in the first place is covered in the quality control reference.

A Two-Quarter Plan for Moving up a Tier

For a programme sitting at two or three styles preparing to grow, the sequence below takes roughly two quarters and does not require stopping anything. It assumes the stitching and hardware defect family definitions have already been settled, since those form most of the general layer.

1. Weeks 1-2: write the general specification — defect classes with examples, measurement conditions, thread defaults, reference-edge conventions.
2. Weeks 3-4: convert existing styles to annexes against it, noting every place they conflict, then resolve each conflict once.
3. Weeks 5-8: establish tiering and apply the four triggers to the existing range, writing each style's tier into its annex.
4. Quarter two, weeks 1-4: introduce the downgrade rule and run the first two lots under it, checking that nothing degrades.
5. Quarter two, weeks 5-8: consolidate inspection visits for styles shipping in the same week and measure the saving per man-day.

The expected result at the end of two quarters is a drop of roughly twenty-five to thirty-five per cent in quality cost per style, with inspection depth unchanged or higher on the styles that actually carry risk. Programmes that skip weeks 1-2 and go straight to tiering tend to reach the same place about six months later, having written the general specification three times in three slightly different forms.

Triggers for Going Back Down a Tier

The downgrade rule needs conditions written down, otherwise nobody ever applies it. Three are sufficient: two consecutive clean lots at the current tier with no critical defects and no systemic major family; no change to construction, material or component source since the last predicted pass; and no open rectification note older than sixty days. Meet all three and the style moves down one tier from the next lot.

One caution on the first condition. 'Clean' has to mean something countable, not merely that nobody complained. A lot that passed with five majors found against an accept number of seven is not clean in any useful sense — the operating characteristic curve under ISO 2859-1 means such a lot has roughly a genuine defect rate near the limit — and downgrading on that basis steadily erodes protection. Set the bar at a number: downgrade when two consecutive lots sit below half the accept number, not simply below fail.

Sample Size Under ISO 2859-1 at Each Lot Size

The standard inspectors work to scales the sample to the lot, and the relationship is not linear: a lot four times larger is not inspected four times as hard. Knowing the sample size and the accept number at each lot size is what makes a report readable, and it is what stops a buyer treating a pass on 50 units as evidence about 5,000.

• A 500-piece lot is 50 units with an accept of 3. Finding three majors passes the lot and says nothing comfortable about the other 450.
• The operating characteristic curve matters. A lot accepted at the accept number has a genuine defect rate near the limit, which is why the downgrade rule uses half the accept number.
• Switching rules change the sample. Tightened inspection after two rejected lots raises the sample size without raising the cost of the visit.

Defect Classes and the Numbers Attached to Each

Three classes and three numbers turn a report from an opinion into a decision. Critical defects are zero-acceptance at every lot size because they are safety or compliance failures. Majors and minors carry the accept numbers the standard sets, and the distinction between them is the single judgement most programmes never write down, which is why two inspectors can read the same bag differently.

• Write the examples into the general specification. A class without examples is a class that will be argued about at the inspection table.
• Critical is never negotiable. Zero accept means zero, and a single critical in 50 units stops the lot.
• Count separately. Majors and minors are drawn against different accept numbers; adding them together is the most common reporting error.

Four Numbers to Track Each Quarter

Cost per style is the number the tiering is judged on, but it is a lagging measure and it does not say whether inspection got better or simply cheaper. Four measures together do: quality cost per style, defects per hundred units, first-pass yield and re-sort hours. Track them per lot and the fourth one tells you when a tier change is working.

• Track per lot, not per year. A yearly average hides the one lot that cost three times the rest.
• First-pass yield is the cheapest measure. It is already on the inspection report and it predicts next quarter's cost better than anything else.
• Set the review at quarter end. Four numbers reviewed quarterly beats a spreadsheet reviewed never.

Frequently Asked Questions

People Also Ask