Every seam on the sectional measured 3 mm and the seat still stepped 6 mm where the modules met
Illustrative example This case is written from the patterns we see in furniture export orders, and the figures are representative rather than from a specific client file. It shows how a check works and what it finds. We will replace it with a real, anonymised case as soon as we can publish one with the buyer's permission.
The order, and the line that could not find it
A European homeware distributor ordered 120 sets of a three-piece modular sectional from a factory in Foshan: two armless seat modules and one corner module per set, 360 modules in total. The corner unit and the armless unit are joined by hook-and-latch steel plates supplied in the hardware bag, and the customer assembles the three modules in the store. The order was $580 a set, $69,600 in total, 30 per cent paid as a deposit and the remaining $48,720 held against the pre-shipment inspection report.
The purchase order had a module alignment section, and it was four lines long. Modules to join flush. Gap between modules not more than 6 mm. Seat height 430 mm, plus or minus 5 mm, floor to top of the seat cushion, uncompressed. Overall height 780 mm, plus or minus 5 mm. Corner unit and armless unit to be built from the same frame programme.
That last line is where the case was won and lost. The corner unit and the armless unit were built from the same frame programme. Same frame height, same top rail section, same 100 mm seat cushion, same connector, same leg. One section of the corner frame was 6 mm deeper at the front, to carry the joint load where three modules meet, and the factory took the same 6 mm back out of the top rail so the silhouette and the overall height did not change.
The consequence was invisible from outside and measurable in exactly one place. The corner unit's seat deck sat 6 mm higher than the armless unit's seat deck. Every external line on the assembled set stayed flush: the arm line, the back cushion line, the overall height, the 3 mm to 4 mm seam where the frames met. The seat plane did not stay flush, and the purchase order had no line about the seat plane.
This is an illustrative example drawn from a recurring pattern in modular seating orders, where the requirement is written on the part and the defect only exists in the relationship between the parts. The figures are representative of the category rather than taken from a named customer file.
Where to go next
→ our sofa order verification scope, including frame and assembly checks
Why a 6 mm step hides behind a 3 mm seam
The connector is the reason. A hook-and-latch plate on a modular sectional is designed with a few millimetres of vertical float, because a hook that engaged on one exact plane would refuse to latch on a frame that is 2 mm out. The float is what lets the hardware close. Push two modules together, the hooks pull, the faces come into contact, the fabric seam draws up to a 3 mm to 4 mm line, and the joint reads as mated.
Nothing in that mechanism looks at height. The connector constrains the modules in plan, side by side, and it holds them together. It does not compare where the seat deck of one module ends and the seat deck of the next begins. A module whose deck is 6 mm high still presents a face the connector can reach, still closes to a 3 mm seam, and still passes a gap check with a feeler gauge.
This is why seam alignment and surface continuity are two different measurements and cannot be used for each other. The seam gap is a horizontal question: how far apart are the two faces, and is the line even along the joint. The step is a vertical question: does the surface you sit on carry across the joint, or break there. A joint can be perfect in the first sense and fail in the second, and no quantity of gap measurement will show it.
The cushion hides what is left. The corner unit carried the same 100 mm cushion as the armless units, so the 6 mm in the deck carried straight through to the cushion top. With the cushions in place and pushed together, the eye reads the cushion seam, and the cushion seam was even. The step was under the cushion, visible from above and across the seat, which is not where anyone stands while checking a sectional.
It becomes obvious the moment somebody sits down, slides a hand across the seat, or sets a laptop on it and watches it travel. Then it is a 6 mm ridge across a 430 mm seat: a hard edge against the back of the legs, and a surface that tips.
The buyer's checklist had no line for any of that. It had a gap line, a seat height line, an overall height line and a colour line. Every one of them passed.
Where to go next
What was measured on site
Two things were done on site that the factory's own final check had not done. The first was to assemble a set before checking it. The second was to take the seat cushions off before measuring the deck.
Assembly came first, because the defect is a relationship between modules and a relationship cannot be measured on a single part. Twenty trial sets were assembled from the production lot and the joint checked on each one. Connector engagement was checked before anything was recorded, because a connector that is not fully latched produces a gap that is an assembly fault rather than a manufacturing fault, and the two have to be separated before a number is written down.
The cushions came off for the structural reading. A 100 mm cushion sits on the deck and partly masks a step, so a 6 mm deck difference measured under a soft cushion reads smaller than it is. The deck was measured bare, with a 900 mm aluminium straightedge laid across the joint resting on the deck of each module, and the gap under the straightedge was read with a feeler gauge set at three stations along the joint: 150 mm in from the front edge, at the centre, and 150 mm in from the rear. The same three stations were then measured again with the cushions on, because that is the surface the customer gets.
Seam gap was read at the same three stations, with the feeler gauge on the vertical face gap, and the spread between the largest and smallest reading along a single joint was recorded as a separate number. An even joint reads 3 mm to 4 mm at every station. A joint that reads 2 mm at the front and 6 mm at the back is a different fault entirely: a racked or twisted frame, not a height difference.
Two supporting measurements were taken because they explain the cause. Leg length from the floor to the bottom of the frame was the same on all three module types. Frame deck height from the floor was not.
| Measured on 20 trial sets sampled from a lot of 360 modules | Armless modules, 240 pcs | Corner modules, 120 pcs |
|---|---|---|
| Frame deck height, floor to top of seat platform | 328 mm | 334 mm |
| Seat cushion thickness, foam and cover | 100 mm | 100 mm |
| Seat height to top of cushion, PO 430 mm plus or minus 5 mm | 428 mm | 434 mm |
| Overall height, PO 780 mm plus or minus 5 mm | 780 mm | 780 mm |
| Back cushion top line relative to the armless unit | flush | flush |
| Seam gap at the joint, front / centre / rear | 3 to 4 mm | 3 to 4 mm |
| Straightedge gap across the seat deck at the joint | 0 mm | 6 mm |
| Difference in seat height between the two module types | reference | 6 mm |
| Leg length, floor to bottom of frame | 130 mm | 130 mm |
One caveat was recorded in the report. The step was measured with the modules standing on their own glides on a level hard floor, the way they will be assembled in the store. On carpet the pile compresses and can create or hide a few millimetres on its own, so a step measured on a showroom floor and a step measured on a factory floor are not the same number.
Where to go next
The finding the report could not fail
This is the awkward part of the case, and it is worth being precise about it. On every line of the buyer's own purchase order, the goods conformed. The seam gap was 3 mm to 4 mm against a 6 mm limit. The seat height was 428 mm on the armless units and 434 mm on the corner units, against 430 mm plus or minus 5 mm. The overall height was 780 mm on all three module types. A report written only against the checklist would have released the balance.
The report could not fail the goods on those lines, and it did not try to. An inspection report can only record a non-conformance against a requirement that exists. Where the purchase order is silent, the honest thing is to name the risk, state the measurement, say plainly that it sits outside the checklist, and hand the commercial decision back to the buyer. The report did that, in one page, with the three measurements and the eight trial sets attached.
What the report could do was record the sampling, because the difference between what was sampled as parts and what was sampled as sets is the whole case. The lot was 360 modules. Under ISO 2859-1, general inspection level II, code letter H, the sample size is 20. All 20 sampled modules were checked individually against the purchase order lines, and 20 trial sets were assembled from them and checked for step and seam. Both numbers were stated, with the method, so the buyer could see exactly what had and had not been tested.
It also recorded what it had not verified, because there was something worth not verifying. The structural consequence of a deeper corner front section could not be settled with a tape measure. Whether the corner frame needed a destructive test again at the new section height was the factory's engineering question and the buyer's to ask, not the inspector's to answer.
Where to go next
The two ways to fix it
The factory proposed the cheaper fix and the correct one, and the buyer took a third path, which is worth describing because it is where these cases usually go wrong.
The correct fix was the corner frame. Take 6 mm out of the deeper front section so the deck comes back to 328 mm like the armless units. That means re-machining and re-covering 120 corner units, and it means the corner's load path changes, so the frame drawing goes back to the buyer's engineer and the destructive test on the corner frame has to be repeated. Nobody in the factory wanted to reopen a frame approval four weeks before shipment.
The cheap fix was the cushion. A 94 mm corner cushion in place of the 100 mm one puts the corner's cushion top at 428 mm and the step disappears, with no change to the frame and no re-test. It changes the seat feel on the corner, so it needs a new approved sample, and it leaves the corner unit no longer identical to the armless unit in the one place the buyer had specified that they would be.
What the buyer did was neither, yet. The corner units were re-worked at the front section, the frame drawing was re-submitted and re-approved, and the corner frame destructive test was repeated on the revised section. The inspector verified the corrective action on 8 new trial-assembled sets. The step came out at 0 mm to 1 mm on every joint, and the balance of $48,720 was released after that verification, not before it.
Where to go next
How the specification should be written
None of the lines below cost the factory anything to comply with, and most of them cost nothing to check. They are written as they would have appeared in the purchase order, and each one exists because a specific measurement was missing from the version that was signed.
- Seat deck height, floor to the top of the frame deck, stated per module type: 328 mm, plus or minus 2 mm. The same number for the corner unit and the armless unit. If the two module types are not the same height, the difference has to be stated and justified rather than absorbed silently into the cushion.
- Seat height stays at 430 mm, plus or minus 5 mm, floor to top of the seat cushion, and a relative limit is added on top of it: the seat height difference between any two adjacent modules in the same set shall not exceed 2 mm. The absolute tolerance cannot catch this on its own. Two modules can both sit inside a plus or minus 5 mm band and still be 6 mm apart, and the entire defect lived in that gap between the two.
- Surface continuity across the joint: with the modules connected and the seat cushions removed, the gap under a 900 mm straightedge bridging the joint shall not exceed 2 mm, measured at three stations along the joint, 150 mm from the front, at the centre and 150 mm from the rear.
- The same check repeated with the seat cushions in place, because the cushion partly masks the step and the customer feels the masked version rather than the structural one.
- Seam gap not more than 6 mm at the front, centre and rear, and the spread between the largest and smallest reading along a single joint not more than 2 mm. A constant 4 mm and a gap that runs from 2 mm to 6 mm are different faults, and the checklist has to be able to tell them apart.
- Modules must be connected before the set is checked, and the assembly check is part of the lot rather than a bonus. Sampling is stated in sets: a lot of 360 modules is code letter H, sample size 20 under ISO 2859-1 general level II, and the 20 are assembled into trial sets for the joint checks.
- Connector specified by type, part number and vertical float, not as the approved sample. If the hardware is allowed to absorb 3 mm of vertical difference, that has to be a number in the purchase order, because it is the number that decides whether a 6 mm step can be hidden.
- Assembly instructions and hardware supplied with the set, with the number of connectors per joint, and the requirement that a joint must latch without force. A joint that has to be levered closed is concealing a height difference.
- One line about the frame, and it is the one that would have caught this before production rather than after: corner and armless modules to be dimensionally identical at the seat deck, and any change to the frame section on one module type to be notified before production, not after the sample has been approved.
- One sample set presented assembled, alongside the unassembled modules, so the set is seen as a set before it ships and the customer receives something that has been looked at in the configuration they will build.
Where to go next
The parts were fine. The set was not
The defect was 6 mm of steel in one frame section and one missing line in a purchase order. It survived a pre-shipment inspection because the inspection was written against a checklist, and the checklist was written against the parts. Every part was right. The set was wrong, and the set is what the customer buys.
The one thing worth carrying to the next order is the second line on the list, because it is the only one that changes what the tape measure is pointed at. A tolerance on each module is not a tolerance on the assembly. To measure a set, you have to build the set.
Where to go next
→ our sofa order verification scope, including frame and assembly checks
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.