Every mattress had three spring layers instead of four, the height was exactly right, and the hardness label matched
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 retailer ordered 180 pocket-spring mattresses from a factory in Foshan: 1,500 x 2,000 mm, medium-firm, quilted cover, taped edge. The unit price was $320 and the order value $57,600. Thirty per cent, $17,280, was paid as a deposit, and the remaining $40,320 was held against the pre-shipment inspection report.
The construction section of the purchase order was short. Mattress height 280 mm. Comfort foam 50 mm. Base foam 30 mm. Spring unit 200 mm total. Perimeter encasement 60 mm. Cover: as the approved sample. Hardness: medium-firm, label to match the approved sample. Pocket spring construction: as the approved sample.
Two things were missing from that list, and they were the same thing. The number of pocket spring layers was never written down, and the approved sample was never opened. A pocket spring unit with 200 mm of spring can be built in four 50 mm layers or in three 67 mm layers, and the order specified only that there be 200 mm of spring in it. Both builds came out at 280 mm tall.
The buyer's incoming check did what incoming checks do. It read the label on the cover, confirmed the height with a tape, pressed the surface against the approved sample and wrote that the softness conformed. Every one of those checks was correct about what it looked at. None of them could see inside, and nothing in the order told them to.
This is an illustrative example drawn from a recurring pattern in mattress orders, where the requirement is written on the outside of the product and the construction is only visible in a cross-section. The figures are representative of the category rather than taken from a named customer file.
The answer was at the end of the roll
Mattresses of this type are roll-compressed before packing, so a finished mattress arrives as a cylinder about 1,500 mm long, with the cover wrapped around the outside and the whole thing inside a plastic film. At each end of the roll the cross-section is visible: cover fabric on the outside, then the foam, then the pocketed coils stacked in distinct bands.
The inspector who found it was not looking for springs. They were checking carton marks against the packing list at the loading dock and had stopped at a stack whose film was torn, which happens when a roll is handled roughly. Through the tear the end of the roll was visible, and the banded cross-section was there to be counted: three layers of pocketed coils.
The next question was how many of the mattresses in the container were the same. Cartons were labelled by model, the film covered both ends on some rolls and only one end on others, and a roll with both ends covered cannot be counted without being opened or cut. On the 20 sampled mattresses, 13 had an exposed end that could be counted, 6 had both ends covered, and 1 was cut open to measure the layer heights on a proper section.
This was the cheapest discovery in the whole case. It cost one torn corner of plastic film and the discipline to look at what was in front of them.
Where to go next
What was measured on site
Four things were established, in this order, and each one answered a question the previous one raised.
The roll ends gave the count. Thirteen units were counted from the visible cross-section, and every one of them showed three bands of pocketed coils where the specified construction called for four. The count is a count, but a compressed cross-section distorts heights, so the layer heights could not be taken from it.
One mattress was section-cut, with the buyer's written approval, at the roll end so that the cut ran through the full 280 mm section. That section gave the layer heights, the foam thicknesses, the coil wire diameter, the pocket pitch and the perimeter encasement, all measured on a real cut face rather than inferred from a compressed edge.
Every mattress in the sample was measured for height, the cover label was read on all of them, and the ticking and quilting were compared against the approved sample. The factory's goods-in record for the spring unit was then requested, which is the step that turns a sample finding into a lot finding.
| Measured on 20 sampled mattresses from a lot of 180 | Approved sample and purchase order | Delivered lot |
|---|---|---|
| Mattress height at centre, uncompressed | 280 mm, plus or minus 5 mm | 278 to 281 mm |
| Comfort foam above the spring unit | 50 mm | 50 mm |
| Base foam below the spring unit | 30 mm | 30 mm |
| Total spring unit height | 200 mm | 200 mm |
| Pocketed coil layers in the spring unit | 4 | 3 |
| Coil layer heights, bottom to top | 50 / 50 / 50 / 50 mm | 67 / 67 / 66 mm |
| Coil wire diameter, measured on the cut section | 1.9 mm | 1.8 mm |
| Pocket pitch across the usable width | 25 mm | 30 mm |
| Coils per mattress, calculated from the measured pitch | 18,096 | 9,360 |
| Perimeter foam encasement height | 60 mm | 60 mm |
| Hardness label on the cover | Medium-firm, as the approved sample | Medium-firm, as the approved sample |
| Surface press against the approved sample | conforms | conforms |
Height, foams, encasement, label and cover all conformed. The report records them as conforming, because a report that lists only what failed gives the reader no sense of what was actually in the container.
What the hand test and the tape measure can and cannot do
Two of the buyer's checks were good checks. On their own terms, they were also the wrong two.
The hand press. Pressing the surface and comparing the feel against the approved sample is the fastest and most trusted check on a mattress, and it verifies the cover, the quilting and the top comfort foam, which is the layer a customer actually touches. Its boundary is that it is a judgement about the top of the mattress. A 50 mm comfort foam over four 50 mm coil layers and the same 50 mm foam over three 67 mm coil layers feel close enough to be called the same by hand, because the hand is reading the top 50 mm. The number of layers below it is not what the hand reads.
The tape measure. Total height was exactly right at 280 mm, and the spring unit height was exactly right at 200 mm, because the substituted unit was 200 mm of spring as well. A tape measure measures totals, and a total is satisfied by any number of layers that add up to it. If a specification gives only a total height, then no measurement of that total can detect a change in the build underneath it.
Neither method is wrong and neither should be dropped. The lesson is that they answer different questions, and the layer question was not one of them. A mattress supplier verification that reads the label, presses the surface and measures the height has verified three things, all of them correctly, and has still not established what is inside the mattress.
Where to go next
How the report was written, and what it did not say
The report was issued the same evening, inside the 24 hours promised. Its structure on this attribute mattered more than usual, because the buyer was going to argue about the lot and the report had to survive that argument.
It recorded the lot, the sampling basis and the character of the defect. Lot of 180 mattresses. Attribute sampling under ISO 2859-1, general inspection level II, code letter H for a lot of 180, sample size 20. Layer count treated as a critical characteristic, accept 0, reject 1. Under that treatment a single confirmed failure rejects the lot, and 14 were confirmed.
It separated what was measured from what was inferred. The report said that 14 of the 20 sampled mattresses were confirmed at three layers. It did not say that all 180 were three layers, because that was not measured. What it did say was that the factory's own production record showed one purchase of the substituted spring unit against this order, that the lot was built in a single run between two dates, and that the 6 unconfirmed units came out of that same run, so the inference was that the whole lot shared one construction. That inference was the buyer's to accept or reject, and the report put it in its own paragraph rather than in the findings.
It refused the wrong conclusion. The label was not false and the hardness was not wrong: the cover said medium-firm and the surface was medium-firm. Calling it mislabelling would have been a different case with different remedies, and it would have been wrong.
It recorded the passing characteristics with the same specificity as the failing one: height, foam thicknesses, encasement, cover, ticking, carton marks, and the count of 180 in the container. A reader who wants to know whether this container was worth shipping should be able to get that from the report as well.
| How the layer count was established | Units | Result |
|---|---|---|
| Roll end exposed, counted on site | 13 | 3 layers on 13 of 13 |
| Section cut from one mattress, buyer's written approval | 1 | 3 layers confirmed, heights 67 / 67 / 66 mm |
| Roll end covered, no section taken | 6 | layer count not established |
| Total sampled | 20 | 14 confirmed at 3 layers against a specified 4 |
Where to go next
How the specification should be written
The lines below 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. None of them costs the factory anything to comply with, and most of them cost nothing to check.
- Pocket spring layers: four layers minimum, written as a number. Neither the phrase pocket spring nor the phrase as the approved sample states how many layers are in the unit, and a total spring height cannot substitute for the count.
- Layer height: each pocketed coil layer 50 mm nominal, and total spring unit height 200 mm. Both numbers, because a total alone is met by any number of layers that add up to it.
- Coil wire diameter 1.9 mm nominal, measured on a cut section rather than read from a catalogue.
- Pocket pitch not more than 25 mm across the usable width, which fixes the coil count without requiring anybody to count 18,096 coils.
- Perimeter edge support: 60 mm foam encasement on all four sides, stated separately from the spring unit, so that a change to the spring unit cannot be absorbed quietly by taking millimetres out of the edge.
- Hardness: the label wording to match the approved sample, with an explicit note that the label is a description and is not evidence of the internal construction.
- Approved sample: to be supplied as a whole mattress and as a section specimen, so the internal construction is visible and comparable instead of sealed inside a plastic bag.
- Verification: layer count to be verified from the roll end on sampled units and confirmed by one section cut per production lot, with the method and the readings recorded in the report.
- Sampling: layer count a critical characteristic under ISO 2859-1 general level II, accept 0, reject 1, with the sample size for the lot stated on the purchase order.
- Production record: the factory to state the spring unit part number used for the lot in writing before the balance is released, and to confirm that the same unit was used for the whole quantity.
Where to go next
What happened next, and where the delay came from
The report reached the buyer the same evening. The balance of $40,320 was withheld the next morning, and the factory was asked two questions: which spring unit was used on this order, and what the approved sample had actually been.
The factory's records answered both. The goods-in note showed one delivery of the three-layer unit against this order, and the internal production sheet showed the lot built in a single run over eleven days. The substituted unit had been bought at $6.80 a mattress less than the specified unit, a saving of $1,224 across the 180. The buyer's own technical sheet, which the factory built to, had said only pocket spring.
What the records did not settle was the six mattresses in the sample whose layer count had not been established. The factory's argument was that those six might be conforming, since they had never been cut. The buyer's answer was in the production record: all 180 mattresses came out of the same run on the same unit, so there was nothing about those six that could differ. The report had deliberately not claimed them, and that restraint cost nothing once the records arrived.
The lot was then settled between the two parties, either by rebuilding to the specified four-layer construction with the mattresses re-compressed and re-loaded, or by a credit. The numbers in that settlement are the buyer's. The part worth recording is where the delay came from: not from the inspection, which found the problem in one torn corner of film, but from eleven days of production that then had to be unwound.
Where to go next
The sample was never a specification
Nothing on this order was fabricated, mislabelled or concealed. The factory built 200 mm of pocket spring and a 280 mm mattress exactly as instructed, and the label described what the surface felt like, which was true. What was missing was a number in the purchase order, and a number missing from a purchase order does not get filled in by a good-faith factory.
The one thing to carry to the next mattress order is the seventh line on the list. A reference to an approved sample is a reference to a sealed plastic bag. It cannot tell anyone what is inside, and everyone downstream assumes it does: the factory, the inspector and the buyer alike. A specification that says as the approved sample for a construction detail has outsourced that detail to whoever guesses best.
For mattress orders specifically, the practical version is short. Get a section of the approved sample, count what is in it, and write the number down. Everything else in this case followed from a single line that said pocket spring where it should have said four.
Where to go next
→ our mattress and bed verification scope
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.