120 hotel lobby chairs arrived with a woven tag sewn under every seat cover reading B1. The file that came with them held a genuine textile report and a genuine foam report, both correctly graded, both about the two materials the chair was made from. Not one document in the file was about a chair.

The order, and the two documents that came with it

A hotel group was refitting the lobby and the lobby bar of one of its properties in Guangdong, and the order went to a factory in Foshan for 120 upholstered chairs. Eighty were Model A armchairs with a shaped back, forty were Model B side chairs with a flat back. All of them were four-leg, all of them had an upholstered seat and an upholstered back, and none of them had arms. The purchase order carried one line on fire performance. All upholstered seating in the lobby and lobby bar shall meet the flame-retardant requirements for products used in public places, grade 1, and the grade shall be marked on the product in accordance with the standard.

The chairs were finished and a week from loading when the buyer's fire-safety contractor asked for the documentation. The factory sent two documents. The first was a textile report for the two cover fabrics, graded flame retardant grade 1, with real numbers in it. The second was a report for the polyurethane foam, also graded grade 1, also with real numbers in it. Both came from an accredited laboratory. The factory also sent a photograph of the label sewn under the seat cover of a finished chair, because that is what the contractor had asked to see.

What the factory did not send was a report about a chair. Nothing in the file described the article that was actually going into the building. Everything in the file described the two materials it was made from, and both of those materials were, on their own terms, correctly certified.

What the chair label actually said

The label in the photograph was a woven tag, 30 mm by 20 mm, sewn into the seam under the seat cover of every chair in the order. It carried a mill's name, the characters for flame-retardant grade B1, and the words 符合 GB 8624-2012. There was no report number on it, no product category, and nothing that pointed at either of the two documents in the file.

Two things are wrong with a label like that, and only one of them is visible on the chair.

The first is grammatical, and it is a requirement rather than a preference. GB 8624-2012 does not grade a material and stop at the letter. A classification is written as a grade plus the category of the product: B1 for a soft upholstered chair is written as B1 for soft upholstered furniture, and B1 for a piece of cloth is written as B1 for decorative fabric, qualified by how many washes it survives. That standard requires the grade and the product category to be stated on the product or its packaging. GB 20286-2006 makes the same demand in its own vocabulary, requiring the flame-retardant grade of a product used in a public place to be marked on the product or its packaging and to be explicitly stated. A tag reading B1 and 符合 GB 8624-2012 satisfies neither marking rule, because the missing half is the half that says what was tested.

The second is that the label travels, and this is the part that made the case, because nothing about the chair gave it away. A certificate cannot move. It names a specimen, a standard, a date, a laboratory, and it stays with that specimen for as long as anyone cares to check. A label is a physical object, and physical objects get moved. The mill had earned a grade on a piece of cloth in a test machine. The factory printed the grade on a tag and sewed the tag onto 120 chairs. The tag then went into 120 chairs, into cartons, into a container and into a hotel. Not one word changed on the way, and the object the claim was about changed completely.

Nobody made an error, and that is worth stating because it is how these cases go wrong. The mill's report was correct. The factory's label matched the mill's grade. The one step nobody took was writing the product category down, and that step is exactly what would have stopped a textile grade from travelling into a chair.

• The grade itself, written the way the standard writes it, not as a bare letter.
• The product category, because the same letter means a different test depending on what it is applied to.
• The report number, so the claim can be walked back to the document that supports it.
• The issuing laboratory, since a grade is only as good as the body that put its name to it.
• The validity period, because a grade for cloth in one fabric batch is not a grade for cloth in the next one.

B1 is not one claim

The reason a bare B1 is dangerous is that the letter is used by more than one route, and the routes are nowhere near each other. GB 8624-2012 grades by product category, and the two categories that matter for a hotel lobby chair are tested on different machines against different specimens with different failure modes.

Now look at what the file actually held. The textile report was a grade 1 result for the cover fabrics, with an oxygen index of 32.0 or above, a damaged length of 150 mm or less, after-flame of 5 seconds or less, after-smoulder of 5 seconds or less, no flaming of the cotton pad below the specimen, a smoke density rating of 15 or less and smoke toxicity not below ZA2. Every one of those numbers is a real number, measured on cloth, for cloth. And the label on the chair said B1.

That is the trap. The one word on the label was the one word in the file, so any check that compared the label against the certificate and looked for a matching grade passed. The check that mattered was whether the certificate covered the object the label was attached to, and it did not. One of them was about cloth, the other was about a chair, and they met only at the letter.

There is a second mismatch inside the textile route, worth knowing before anyone writes a specification. A textile grade can be qualified by a wash count, so a grade for fabric that survives twelve water washes is marked differently from a grade for fabric that is not washable at all. A chair cover is never washed. So even if the chair label had legitimately carried the fabric's grade, it would have been carrying a claim about a textile that nobody could ever verify on the chair, about a product whose fire performance is decided by the assembly and not by any of its parts.

The route that puts the chair in the machine

For a soft upholstered chair in a public place, GB 20286-2006 does not ask for a fabric grade at all. Clause 5.6 and Table 4 send furniture and components used indoors in public places to Appendix B for soft upholstered furniture, citing GB 17927, and set four criteria for grade 1.

The specimen is the point. Appendix B applies to soft upholstered furniture or a full-scale model of it, and the machine is a 250 mm square burner on a 3 m by 3 m collection hood burning propane or natural gas, with the specimen on a weighing platform and the exhaust flow measured through the duct. The chair goes in as built. A cut sample of cloth does not go in, a block of foam does not go in, and a sandwich of the two does not go in either, because the standard offers no route where the result for a chair is assembled out of the results for its parts.

The table also shows why the two documents in the file could never have been made to work. The standard keeps the unit and the component on separate rows with separate criteria, and the component row carries three indicators where the furniture row carries four, the fourth being the requirement that the specimen itself does not ignite and does not smoulder. A component result stays a component result. Combining a textile grade with a foam grade does not produce a chair grade, because there is no arithmetic for it and the standard does not offer one.

The foam in the chair was a separate requirement in its own right, which the factory's foam report did satisfy. GB 20286 Table 3 puts foam used in seating, sofas and mattresses on its own row, with its own methods: GB/T 16172 for heat release rate, GB/T 8333 for the burn behaviour, GB/T 8627 for smoke density and GB/T 20285 for smoke toxicity. That report was fine. It was also, and only ever, about the foam.

How it was found

The first thing that happened was not a measurement. The buyer's fire-safety contractor photographed the label on a chair in the delivered carton and asked which report stood behind it. That question needs no factory, no machine and no sample size. It needs one document. The factory replied with two documents and the photograph, and neither of the two carried the word chair anywhere in it.

The second step was to read what the two documents were built on rather than what they concluded. The textile report was built on GB/T 5454 and GB/T 5455, which are the methods for cloth. The foam report was built on GB/T 16172, GB/T 8333, GB/T 8627 and GB/T 20285, which are the methods for foam. Neither was built on Appendix B, and neither was issued for a chair.

The third step was to ask the only question here with a countable answer: how many distinct chairs are in this order that need a result of their own. Two. Model A and Model B are different articles with different covers, and Appendix B grades the specimen that is supplied, so each build needs its own result. The file had no Appendix B result for either of them.

Two of the four documents this order needed were in the file, and the two that were missing were the two that the label on every chair claimed. The label was not unsupported because a supplier was careless. It was unsupported because the documents that would have supported it had never been commissioned, and nobody had noticed that a label sitting on 120 chairs is a claim rather than a decoration.

What the label check on twenty chairs found

The goods were inspected as well, and the inspection is where this case became uncomfortable, because the inspection passed the labels. A lot of 120 falls in ISO 2859-1 code letter F, which covers 91 to 150 units, and at general inspection level II the sample size is 20. At AQL 2.5 the acceptance number is 1 and the rejection number is 2.

The label check as it was written on the inspection sheet asked three things. Is the label present, is it legible, is it attached to the right chair. On all twenty sampled chairs the answer was yes. Every one of the 120 chairs had a clean, legible label on it, correctly sewn. Zero defects, and the lot passed.

No sample size would have helped, and the reason is worth being precise about, because it is not that the sample was too small. Nothing about the chair was wrong. The label was correctly printed, correctly attached and correctly worded as far as its appearance goes. What was wrong was the relationship between that label and a document that had never been issued, and a relationship is not an attribute of a unit. There was no unit to reject and no defect to count. AQL answers how many units may fail at a given rate, and this failure had a rate of 100% and was not a unit failure at all, so the plan had nothing to work with.

Tightening the plan to AQL 0 would not have produced a finding either, because at AQL 0 the first chair is rejected and the reason written down would have been that the sample did not match the requirement, which is not a reason. What the piece-by-piece check could confirm is worth confirming: that the label was on the chair, that it read the same on every chair, and that the 120 chairs split into the two builds the order called for, 80 and 40. Those took a few minutes and they are in the report. The question that mattered was not a per-unit question. It belonged in the document review, answered once, before a single carton was counted.

This is the part of a hospitality furniture QC china assignment that most often gets written off as a paperwork detail, and in a hotel fit-out it is the part the buyer's own fire-safety contractor will look at first, because that is the document they have to sign against.

• Whether a label is present on the unit and legible on the unit, which is a per-unit check and belongs in the sample.
• Whether every unit carries the same label, which is a per-build check and can be done on one chair per build.
• Whether the units split into the builds the order called for, which is a count and costs nothing.
• Which report number the label refers to, which is not a per-unit check at all and belongs in the document review.
• Which product category the label names, which is also a document review and takes about a minute.
• Whether the report behind the number was built on the method that applies to the product, which is the document review again.

How the report recorded it

The report was written around the requirement in the purchase order, because the buyer would read it against that requirement and nothing else. Fire performance was graded as its own line, ahead of the physical checks, because a missing flammability result is a different kind of finding from a scuffed frame. One is closed by a document before loading and the other is closed by a rework instruction on site, and putting them in one list is how both get missed.

The recommendation was to hold the fire documentation, not the goods. Two conditions had to be closed before the container was sealed. First, an Appendix B result on each build as supplied, with the cover fabric, the interliner if there is one, and the seat filling named in the specimen description, so that the result is tied to the chair rather than to a description of it. Second, a re-issued label on each chair stating the grade, the product category, the standard and its year, and the report number of the result that covers that build.

None of this required the factory to stop work. It required a laboratory, one chair per build and about three weeks, all of which could have been arranged in the twelve weeks before production finished.

What it cost to close this before loading

A third-party Appendix B test on a chair in this category runs in the region of $320 and destroys the specimen. Two builds, two specimens, so $640 of testing and two chairs out of 120 at the order price of $118, which is $236 of goods. $876 in total, and that is 6.2% of the $14,160 order value. It buys the two documents that were missing, before the container was sealed.

The alternative was to find out at the door. A hotel's fire-safety contractor cannot sign off a lobby on a fabric certificate. The order was $14,160, with 30% taken as a deposit of $4,248 and the balance of $9,912 due against loading. The chairs were already installed in a working lobby. Either two of them came back out of a finished room for destructive testing while 118 sat there of unknown standing, or the balance stayed with the buyer for three weeks of laboratory time.

The verification that found all of this was two man-days of full verification at $299, so $598, which is $4.98 a chair across 120. The inspection cost $598, the two missing documents cost $876, and together $1,474 is 10.4% of the order value for a lobby that was about to open. The same $1,474 spent after the installation is not a rounding error, because then it buys a room with chairs taken out of it.

Seven clauses for the next purchase order

The order was not badly written in intent. It named the right standard for the right product and asked for a mark on the chair. What it did not do was close the gap between the two, and the gap is where the whole case lived.

• Name the standard, the grade, the category and the method in one line. Write GB 20286-2006, grade 1 for soft upholstered furniture, tested to Appendix B with the chair as the specimen, and never write a grade on its own.
• Ask for one result per build. Say how many distinct chairs the order contains and name each build, so the factory counts the results the same way it counts the chairs.
• Require the document to carry a report number, the issuing laboratory, the product category and the validity period, and require the label on the chair to carry the same report number.
• Specify the label wording yourself and approve it before it is printed. Grade, product category, standard and year, report number. A label drafted by the factory from the mill's certificate is a label drafted from the wrong document.
• Require the specimen description to name the cover fabric, the interliner if any, and every filling with its grade and density, so that a later change of fabric voids the result explicitly rather than by argument.
• Write that any change to fabric, foam, interliner or structure voids the result and requires a new test before shipment. A chair is a composite and a new composite is a new chair for this purpose.
• Require the marking to be on the unit and not on the carton alone, because the mark that is inspected is the mark that is on the chair in the room.

Three questions to ask before the next delivery

Which report number is on the label of the chair in the photograph, and which build does that report cover.

Does the report behind the number state a product category, and does the category match what the label claims.

For each distinct build in the order, is there a result on the chair itself rather than on one of its materials.

None of these three questions needs a factory visit, a sample size or a measurement. They need the file and a photograph of the label. This order produced no fraud and no false document. Every report in it was a genuine report of a genuine test on a genuine material, and every person who handled it was being reasonable about a question nobody had put to them. The failure was that nobody asked what each piece of paper was about, and that a label, unlike a certificate, does not stay where it was written.