The fixings are for a different model

The situation

A European DIY retailer had ordered 640 units of flat-pack modular storage from a factory in Guangdong — six models, two door widths and three heights, packed one unit to a carton, filling one 40HQ and one 40GP. Production ran over five weeks. Assembly instructions were issued by the buyer’s product team at revision C, and the hardware schedule attached to the purchase order listed every part by number, description and quantity per model.

The factory reported production complete. Bagged hardware was weighed, checked against the schedule by quantity, and packed. Nothing in that sequence compares a part number. The count was right for all six models, and the bag label showed the correct model code and the correct revision — because the label is printed from the work order, and the work order was current.

Two models were wrong. M-04 (1,860 mm, two-door) and M-06 (1,500 mm, one-door) had been packed with the hardware kit from revision B — the fit before the board thickness changed. The bags were labelled revision C. The contents were revision B.

Why the count matched

Four of five line items were the wrong part, and the bag was still correct on the only number anybody had recorded.

How it was found

The bag audit found it. The assembly found it faster, and the assembly found things the bag audit could not.

The pre-shipment inspection opened five hardware bags per model, drawn from spread positions across both containers, and laid each out against the schedule line by line. Forty-four of the affected bags were identified by carton number before anything was opened for assembly — M-04 at cartons 61–240 and M-06 at cartons 366–485, 300 units in total.

One unit of each model was then assembled from production packaging, with the factory’s own instructions and the buyer’s tools. **M-04 would not complete step four.** The 6 mm cam bolts entered the 8 mm cams freely and rotated without pulling the panels together; the assembly was abandoned twice and completed once with substituted bolts purely to establish what the remainder of the build looked like.

This is the finding that changes the report. The bag audit establishes that the wrong parts were supplied. The assembly establishes **that the product cannot be built with the parts in the box** — and it does that in about eleven minutes, on a unit, in front of the factory. No amount of documentation establishes that, because nothing in the paperwork claims the unit is buildable.

• **Open bags spread across the run, not from the top of the stack.** A revision change applies from a point in the production sequence onwards, so affected cartons cluster. Bags from the first, middle and last production days give the boundary in three draws.
• **Check by part number, against the drawing, not by description.** Descriptions are written by the person specifying the part and are routinely shared across sizes. The part number is the only field that cannot be satisfied by a similar-looking part.
• **Build one unit per model from its own production carton.** Not a showroom unit, not a pre-assembly sample — a unit packed the way the rest were packed. The factory assembles units all day; an assembly line that completes every unit is the most persuasive evidence in the report.
• **Record the step where the build stops, in the factory’s words.** *Joint will not draw together at step 4* is a finding. *Hardware incorrect* is an opinion.

What assembly finds that counting cannot

The cost of running an assembly test on one unit per model is a few hours of an inspector’s time and a little factory labour. The cost of the same finding at the destination is a technician on a call-out per unit, a return consignment, and a retailer whose customers assembled flat-pack furniture at a weekend.

Where a superseded kit actually comes from

Nobody chooses the wrong revision deliberately. Four ordinary mechanisms produce this, and each of them is removed by a sentence rather than by inspection.

• **The revision that never got closed out.** The buyer changed the board from 16 mm to 18 mm and reissued the hardware schedule. The factory was told. The old kit stayed on the shelf beside the new one, in identical bags, with no marking — and the older shelf is closer to the packing bench.
• **A kit built once and repeated.** Hardware bags are built in bulk so the packing line does not stop to assemble them per unit. A kit built for the first hundred units of a production run is reused for the rest, which is correct practice for a stable kit and the mechanism behind every failure when the kit is not stable.
• **The work order carries the label, not the contents.** Label printing and bag filling are separate operations, so the label reflects the current revision by construction. In effect the documents say revision C while the parts say revision B, and the label is the more visible of the two.
• **Parts picked from a bin by description.** Where a dowel or a screw is bought in bulk and drawn from stores at the bench, the person drawing it matches what is written on the bin. A bin marked *dowel 10* will supply a 30 mm dowel to a bag that needs 45 mm, and the bag will still hold 46 pieces.

What the report recorded

The report on this order was short, because the remedy is cheap if it happens before stuffing. What makes it useful is that each statement is separable and each carries a number.

• **Which units, which cartons.** 300 units of 640 — all 180 units of M-04 and all 120 of M-06. Identified by carton number: 61/640 to 240/640 and 366/640 to 485/640. The four remaining models were audited at five bags each and returned clean.
• **Which parts, per line item.** Four of five lines in the schedule incorrect on the two affected models, with scheduled and found dimensions written side by side rather than summarised as *hardware non-conforming*.
• **Whether the goods build.** One unit of each affected model assembled from production packaging; M-04 will not complete step 4, M-06 assembles but the doors will not close. This is stated separately from the parts finding, because the remedy for the second is a specification gap.
• **What was not verified.** The remaining 300 hardware bags were recorded as *not audited*, not as *correct*. The affected boundary was established by production date rather than by opening every bag, which is a sampling result and is labelled as one.

What the specification should have said

Written this way, the superseded kit is not a finding. It is an impossibility — because the bag states its revision, only one revision is permitted at the bench, and one unit per model has to build before the doors close. **The point of the specification is not to catch this at inspection. It is to make the inspection have nothing left to catch.**

• **Schedule the hardware by part number, per model, and print it on the bag.** *Each hardware bag to be labelled with model code, hardware part number, revision and a line-by-line part schedule. Parts to be identified by part number; descriptions are not to be used for identification.* The bag label is the last point at which a revision can be checked against the schedule, so it has to carry the identity.
• **Forbid substitution without written approval.** *No fastener, dowel or cam may be substituted, and no dimension may be shortened, without prior written approval from the buyer. A substituted part is a non-conformance even where it fits and even where the quantity is unchanged.* That single clause removes the bin-picking mechanism and the shortened-screw mechanism together.
• **Close out revisions in writing and confirm the withdrawal.** *On receipt of a revised hardware schedule the buyer requires written confirmation that all previous revision kits have been withdrawn from the packing bench and stores, and that remaining stock of obsolete parts is destroyed or returned.* A revision notice without a withdrawal instruction leaves two live kits and no way to tell which is which.
• **Build one unit per model from production packaging before stuffing.** *One production-packed unit per model to be fully assembled and the result recorded before container closure. Cost to be borne by the manufacturer.* Eleven minutes per unit is the cheapest test in the entire programme.
• **Audit bags against the schedule before stuffing, at a stated rate.** *Five hardware bags per model to be opened and reconciled line by line against the schedule, plus every bag produced from a kit built in a different production run. Result recorded as a count per model.*
• **Supply spares with the goods.** *Two per cent additional hardware bags per model, packed separately and listed on the packing list.* A retailer that can complete a unit from a spare bag does not send a customer to a technician — and the spare bags are themselves the cheapest possible audit of whether the primary bags were right.