The seat cushions measured 28 kg/m³ against a specified 35. Sitting on them told the buyer nothing.
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 situation
A European homeware importer ordered 300 upholstered three-seat sofas for a retail range. The purchase order carried a material specification for the seat cushions: flexible polyurethane foam, 35 kg/m³, tolerance ±10%. It was one line in a two-page specification, and the factory accepted it without comment.
The sofas arrived looking correct. The covers were cut cleanly, the seams were even, the cushions were plump and uniform across the batch. The foam supplier's certificate travelled with the order, showing 34.6 kg/m³ for the lot.
Sitting on them proved nothing. That is what makes foam density different from almost every other measurable in furniture.
Why a showroom sit test cannot find it
A cushion 20% below the specified density still looks full, still feels firm under a hand, and still passes a sit test in a showroom. Density is not a surface property. What a sitter feels on day one is governed by the foam's surface stiffness and by the cover, not by its mass.
What low density changes is what happens after a year of use. The foam compresses under a sustained load and does not recover fully, because there is less material in the cell structure to push back. The cushion ends up 10 mm thinner than on delivery and never quite comes back.
**This is why it surfaces in returns data rather than on the loading dock.** The batch that ships fine is the batch that produces complaints four quarters later, when the failure is attributed to the product rather than to the foam.
What the inspection found
Five cushions were cut from sofas drawn from the front, middle and rear of the batch. Each was measured for density at the centre core and again 15 mm below the mould skin. A separate compression set test was run to ISO 1856.
Cutting into a cushion is destructive, so this is not a check that routine sofa inspection china work includes by default. It has to be written into the order, or it does not happen.
| Sample | Reading point | Measured kg/m³ | Against 35 kg/m³ |
|---|---|---|---|
| Cushion 1 | Centre core | 28.2 | **−19%** |
| Cushion 2 | Centre core | 27.6 | **−21%** |
| Cushion 3 | Centre core | 28.9 | **−18%** |
| Cushion 4 | Centre core | 28.4 | **−19%** |
| Cushion 5 | Centre core | 27.9 | **−20%** |
| Cushions 1–2 | 15 mm below mould skin | 30.8–31.2 | −11 to −12% |
| Foam supplier certificate | Lot test report | 34.6 | Within specification |
| Compression set, 22 h at 70 °C | ISO 1856, ≤10% | 14–19 | **Out of tolerance** |
Two findings, and they carry different weight. The density result is a material shortfall against a written specification. The compression set result is a durability failure against a standard limit, and it is considerably harder to argue with, because a lot number does not explain it.
The sampling point decides the number
Compare the two reading points. Density in moulded foam is not uniform: the skin formed against the mould during filling cools faster than the core and packs tighter, so a reading taken just below the surface returns a higher figure than the same cushion read at its centre.
**A foam check performed at one point is not a foam check.** On this batch the skin readings came 2.5 to 3 kg/m³ above the core readings. That is enough to move a cushion from a clear failure to a marginal one — and enough that a single-reading inspection would have passed two of them.
- **Where to cut.** The centre of the cushion, clear of the seam, the wrap and the mould skin. Both GB/T 6343 and ISO 6079 specify specimen preparation for exactly this reason.
- **How many.** One cushion is an anecdote. Five or more drawn from across the batch is a finding that survives a dispute.
- **What tolerance applies.** The buyer's PO gave ±10%, which puts the accept band at 31.5 to 38.5 kg/m³. Every core reading fell below it. Had the PO said only *35 kg/m³* with no tolerance, there would have been no pass or fail at all — the same position reached when a dining table arrived 50 mm short of the ordered length.
- **Density and compression set together.** Density alone does not predict how the foam will perform in service. Compression set does. Where the order justifies it, specify both.
When the certificate and the field measurement disagree
The factory's position was straightforward. The certificate says 34.6 kg/m³, it comes from the foam supplier's own laboratory, and the material in the sofa came from that lot. The inspection found 27.6 to 28.9 kg/m³ in the same sofa.
Both numbers can be genuine, and the reason is worth stating precisely. **A certificate describes the foam block the supplier tested, not the foam in the finished cushion.** They come apart in ordinary ways: a second lot was pulled when the first ran out and the certificate was never updated; a lighter blend was used through the core while the surface stayed on certified material; the cutting plan mixed blocks from two deliveries. Any of those produces a valid certificate and a sub-specification sofa at the same time.
So the on-site measurement governs the report, because it is the only figure that describes the product actually shipping. The report records both, together with the lot numbers on the certificate — and does not call the certificate false. **How the discrepancy arose was not established, and the report did not claim otherwise.**
How the deviation was recorded
Every density entry carried five things: the reading point, the measured value, the specified value, the percentage deviation, and the number of samples behind the figure. The compression set result was reported separately with its standard reference, so it could not be folded into a single pass or fail verdict.
The report did not state that the foam was counterfeit, that the certificate was falsified, or that the factory had substituted material deliberately. None of that was supported. It stated that five cushions sampled across the batch read below the specified tolerance at the centre core, that the mould-skin readings were also below it, that compression set exceeded the standard limit, and that the certificate on file did not correspond to the foam in the product.
The factory accepted the finding without argument and opened an investigation into the foam lot. **A finding the other side does not contest is worth more than a finding that wins an argument**, because it can be closed out before the container ships instead of during the demurrage window.
What to put in the specification
Foam is specified in one line in almost every sofa order, and that line is nearly always too short. Four additions, all of which fit on the same line:
- **Density with a tolerance and a reading point.** *35 kg/m³ ±10%, measured at the centre core clear of the mould skin* is one clause, and it removes the entire argument above.
- **A sample size.** *Five cushions drawn from the front, middle and rear of the batch* turns a spot reading into a batch finding.
- **Compression set with its standard and limit.** *≤10% at 22 h / 70 °C per ISO 1856* is the check that predicts the first-year sag, and it is the one most often missing.
- **The status of the certificate.** State whether a supplier certificate is acceptable evidence, or whether on-site measurement is required. Leave it unstated and the certificate wins by default, because it is the document that already exists.
Where to go next
→ our sofa and upholstery verification scope
→ what a report records when the evidence disagrees with itself
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.