600 folding terrace chairs passed the hinge clearance check on every sampled chair, and eleven days into service 23 of them jammed in the pivot. The three degrees of backrest angle left behind after the jam was the only part of the failure that anybody could measure.
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 four lines in the specification
A European terrace and poolside operator ordered 600 folding stacking outdoor chairs from a factory in Foshan. Steel tube frame, powder coated, folding backrest on a bush-and-pin pivot, contract grade for daily commercial use. The unit price was $54, the order value $32,400, thirty per cent paid as a deposit, and the remaining $22,680 released against the pre-shipment inspection report.
The specification covered the chair in four lines. Steel grade and tube section. Coating system and colour. Seat and backrest dimensions, with the seat-to-back angle at 100 degrees, plus or minus 1.5. And the hinge, in one line: clearance at the pivot not more than 0.30 millimetres, measured with a feeler gauge at the tightest point.
That last line is a good line, and it is the reason this case is worth writing down. It names a number, it names an instrument, and it names the place on the chair. A clause like that gets signed off with the sentence that it is testable, and it is.
What it does not say is how many times the chair is going to be folded. That number decides whether the hinge survives, and nobody had written it down.
Where to go next
What the inspection measured
The pre-shipment inspection sampled 80 chairs out of 600 under ISO 2859-1 at general inspection level II, code letter J, AQL 2.5, accept 5 and reject 6. On 30 of the 80, the hinge was checked, and the check was done exactly as the specification asked: a feeler gauge at the pivot, tightest point.
The readings came back between 0.28 and 0.33 millimetres, a spread of 0.05 millimetres across all 30 chairs, every one of them inside the drawing limit. The report recorded hinge clearance conforming, backrest angle conforming, locking and release conforming, and the folding action working. The report passed, the container sealed, and the balance was released.
There is nothing wrong with that report. Every measurement in it was made properly, on the chair, against the number the buyer had written down. The defect that ended this order was not present on any of those 30 chairs, because at the moment they were inspected it had not been created yet. A stacking chair inspection working to this specification had no reason to expect a jam on day eleven, because the specification did not contain the number of cycles that would produce one.
| Check | Method | Result on the sample |
|---|---|---|
| Hinge clearance | Feeler gauge at the pivot, tightest point | 0.28 to 0.33 mm, 30 of 30 conforming |
| Seat-to-back angle | Protractor against the seat datum | 99.8 to 100.4 degrees, conforming |
| Lock and release | Fold and unfold by hand, five cycles | Working on all 30 |
| Static load | Seat and backrest loaded to the contract figure | Conforming, no permanent set |
| Coating | Visual, plus the coating certificate on file | Conforming |
Where to go next
Eleven days into service
The chairs went to twelve sites. Each site stores them folded, in stacks of fifty, in a service store, and puts them out in the morning and folds them again in the evening.
On the eleventh day a chair would not fold. Staff pulled it, and it came open with a bang and a three degree error in the backrest angle. Over the following week, twenty-three chairs across nine of the twelve sites jammed in the same way. Every one of them was in the bottom three of its stack. Nothing at position four or above failed at all. At three sites the stack leaned and had to be taken down and restacked by two people.
And the staff did what staff do with a chair that will not fold: they shouldered it, put a foot against the frame and levered it open. That is the moment a jam became a permanent three degrees.
Where to go next
The three degrees nobody could see
The drawing calls for a seat-to-back angle of 100 degrees, plus or minus 1.5, which puts the lower limit at 98.5. On the chairs that had been forced open, the angle came back between 97.0 and 97.4 degrees. That is 2.6 to 3.0 degrees out, and it puts every one of them outside the drawing limit by between 1.1 and 1.5 degrees.
The backrest is 420 millimetres from the pivot to the top edge. A three degree error over 420 millimetres is 22.0 millimetres at the top. So what the factory shipped and what the operator received was a 22 millimetre error in a chair that looks right, on a frame with sound welds and a coating that passed a salt spray test.
And nobody on the ground could have found it. There is no protractor on a terrace at seven in the morning. The person using the chair either feels it or does not: a three degree change in the seat-to-back angle moves the effective seat depth at the back by 22 millimetres, so the chair sits differently under a person. It is a comfort fault, and comfort faults do not get reported.
Which is why the jams got reported and the 22 millimetres did not. A jam is loud. An angle is quiet. The failure the buyer knew about and the failure that actually cost them were not the same failure.
Where to go next
→ our dining chair verification scope, including folding and stacking checks
Why a tolerance that passed could not have caught it
The buyer's first instinct was to argue that the factory had built chairs outside the drawing. It had not. The drawing said 0.30 millimetres at the pivot and the chairs were at 0.30 millimetres. The tolerance was met.
Here is why that sentence is true and useless at the same time. A tolerance is a bound on a measurement taken at a moment. What failed here was not a dimension, it was a rate. A joint that is inside its clearance on day zero and outside it on day eleven has not violated anything, because nothing in the drawing said where that joint would be on day eleven. A static bound cannot govern a process, and a process is what failed.
There is a second problem, and it is the one that should worry anyone writing hinge clauses. Those 0.28 to 0.33 millimetres were measured by pushing the leaf of the hinge against the feeler gauge. Do that on a pivot with swarf in it and you push the swarf aside and measure the metal underneath. The reading was accurate. It was a reading of a clean joint.
The joint as it actually works is the gap the leaf sits in when nobody is pushing anything. On the jammed chairs that gap was between 0.02 and 0.07 millimetres. Under 0.05 millimetres of free travel a leaf has stopped sliding and started to wear. A chair in that condition does not fold loudly either. It folds the way a hinge that is about to fail folds, which is to say it does not come back on its own.
And a third problem, the one nobody writes down. Stackability is not a property of a joint. It is a property of a system: 600 joints, twelve stacks, fifty chairs high, a floor that is wet in the evening, and salt in the air. The drawing specified one joint on one chair, and the failure was in the system.
Where to go next
What was actually in the pivot
The twenty-three jammed chairs were opened at the factory. The pivot is a pin in a moulded bush with a steel leaf each side, and the clearance is the space between the leaf and the bush where the two rub.
In the bottom chairs that space held ferrous swarf: 80 to 150 milligrams a pivot, weighed on a 0.1 milligram balance after disassembly. In the top chairs of the same stacks it held 3 to 8 milligrams. The swarf at the bottom was wet and rust-stained; at the top it was dry.
Where it comes from is ordinary. The hinge leaf is laser cut and the cut edge keeps its burr. The frame is powder coated and the coating line grinds the tube ends. The pivot is exactly where all of that ends up, and a folded chair on a terrace gets dragged, tipped and rained on. Every cycle works a little more coating off that edge and a little more swarf into the gap.
| Position in the stack | Swarf in the pivot | Free travel at the leaf | Chairs that failed |
|---|---|---|---|
| Bottom, position 1 | 130 mg average | 0.02 mm | 9 of 12 |
| Position 2 | 96 mg | 0.04 mm | 8 of 12 |
| Position 3 | 61 mg | 0.07 mm | 6 of 12 |
| Position 4 and above | 5 mg | 0.28 mm | 0 of 12 |
The free travel column is the one to read twice. It is the gap the leaf actually sits in, measured without forcing the leaf closed, and it is the number the specification never asked for. On the bottom chair of a stack there is almost nothing left: 0.02 millimetres, which is less than the swarf is thick, so the leaf is resting on debris and grinding it in. At the top of the same stack the figure is 0.28 millimetres, which is the metal clearance with nothing in it, and not one of those chairs failed.
Where to go next
How the gap and the jam have to be measured
The pivot on a folding chair is not one measurement, it is five, and the order they are taken in matters.
- Free travel, as found. Put the feeler gauge at the leaf without forcing the leaf closed, at three positions, and take the tightest. This reading includes the swarf bed, because nothing has pushed the bed out of the way. On this lot it read 0.02 to 0.07 millimetres on the bottom chairs and 0.28 on the top.
- Cleaned clearance. Now clean the pivot, with compressed air and then a lint-free wipe and a solvent, and read again. That is the metal-to-metal clearance the drawing is actually about. It came back between 0.28 and 0.33 on every chair in the lot, jammed or not.
- Swarf mass. Open the pivot, collect what is in it, and weigh it on a 0.1 milligram balance. The drawing has no number for this and it should: 130 milligrams is a bed, and 5 milligrams is a floor.
- Peak folding moment. A torque wrench on the backrest, or a force gauge at a stated distance from the pivot, reading the peak moment needed to lift the backrest from folded to open. A fresh chair takes 3.0 to 3.6 newton metres. The jammed chairs needed 11 to 14. The chairs at the top of the stack, with 5 milligrams in them, took 3.4.
- Residual angle. The seat-to-back angle on the same chair at cycle zero, then again after the cycle count, with a digital angle gauge. Drift above half a degree means the pivot is moving, whatever the clearance said that morning.
The fold moment is the easiest of the five to add to a scope, because it needs nothing but a torque wrench. A fresh chair folds on 3.0 to 3.6 newton metres, which is what the backrest's own weight gives: 1.2 kilograms with its centre of gravity 280 millimetres from the pivot is 3.30 newton metres. The jammed chairs needed between 11 and 14, which is three and a third times that, up to four and a quarter times. That ratio is what a member of staff feels in their shoulder when they pick the chair up, and it is why they used a foot instead of a hand.
Where to go next
Why a bigger sample is not the answer
The buyer's next question was whether the inspection should have used a larger sample. The arithmetic says that question does not matter.
Lot 600 at level II is code letter J, sample 80, AQL 2.5, accept 5, reject 6. The defect rate that actually materialised was 23 in 600, or 3.8 per cent. In a sample of 80 from that lot the expected number of defective chairs is 3.1, and the probability of drawing five or fewer, which is a passing sample, is 91.3 per cent. The lot would have passed with better than nine chances in ten.
But that is not why the sample missed it. The reason is that 3.8 per cent is the rate at cycle eleven, and the chairs were inspected at cycle zero, where the rate is zero. Doubling the sample to 160, moving to level III, or inspecting every chair in the lot would all have produced exactly the same result, which is nothing. A hundred per cent inspection of a static characteristic, on a defect that does not yet exist, finds nothing at a hundred per cent efficiency.
That is an unusual conclusion to write down, so it is worth being plain about it: the answer to this failure is not a bigger sample. It is a different test.
Where to go next
What the standards cover, and what they do not
EN 581-2 is the standard that governs the mechanical safety of outdoor seating, and it is thorough. At contract grade it wants 2,000 newtons on the seat and 560 on the backrest, static load on the seat front edge and on the armrests, 50,000 cycles of combined seat and backrest durability, an impact test, and stability.
And it contains no stacking test. It contains no folding cycle count for a folding chair either. That is not an oversight in the 2015 edition. The 2026 revision, approved by CEN in May 2026 and in force from 30 November this year, adds deck chair latch fatigue, latch detachment before and after that fatigue, backrest impact and backrest vertical loading, and still no stacking requirement.
The Chinese standard for folding furniture, GB/T 3324, has a folding test and it is a good one: the chair is tilted to 70 degrees, and then loaded with 50 kilograms on a 10 degree inclined platform, with the folding device in its least favourable position, for five minutes. But it is one cycle and one load. It says nothing about a chair at cycle five thousand.
The number that does exist is in QB/T 4454 for beach chairs, which requires the folding mechanism to survive 5,000 open and close cycles without failure and to stay free of jamming. A product standard for beach furniture already contains the clause this order needed, and nobody copied it into the purchase order.
One safety point deserves its own paragraph, because it is easy to miss. EN 581-1 permits shear and squeeze points in a folding chair, on the reasoning that the person setting the chair up is in control of the movement and can stop when it hurts. That is the basis of the 7 to 18 millimetre band and of the 7, 12 and 18 millimetre probes used to assess it. A hinge that binds and then releases without warning removes exactly that control. A joint that does not close cleanly is outside what the clause assumes, which is another way of saying that the jam and the three degrees are a safety story and not only a comfort one.
Where to go next
What the report recorded, and what it could not
The report on the reworked lot carried five measured values per sampled chair and the counts by stack position. It also carried the things it could not establish, and those lines matter more than the numbers.
It could not establish that the reworked chairs would not jam on the eleventh day, because the re-inspection ran to 50 cycles on the sampled chairs while the factory's 5,000 cycle run was on three chairs that we witnessed and verified but did not carry out. The report says so in those words.
It could not establish how many of the 600 chairs had already accumulated swarf in the pivot, because the factory reworked all 600 rather than sampling, on the grounds that the cause was systemic and not a matter of degree. That is a decision the buyer took with the factory, and the report records it as a decision and not as a finding.
And it could not certify the chairs against EN 581-2 for stacking, because EN 581-2 contains no stacking requirement. Any report that says a folding chair is compliant with EN 581-2 and stops there has answered a question nobody asked.
Where to go next
What it cost, and what one man-day would have cost
The operator did not spend nothing. Here is the invoice for finding out.
| After the first jam | Amount |
|---|---|
| 23 chairs out of service at nine of twelve sites, hire replacements at $9.50 a chair a day for 17 days | $3,715 |
| 600 pivot assemblies reworked at the factory at $4.20 a chair | $2,520 |
| Re-inspection of the reworked lot, 80 chairs sampled, two man-days at the Level 3 rate | $598 |
| Verification of the factory type test, one man-day at the Level 3 rate | $299 |
| Air freight and re-delivery of 23 chairs to the nine sites | $640 |
| Total | $7,772 |
All 600 pivot assemblies were reworked, not the 23 that jammed, because the swarf was coming out of the cutting and coating lines and the only honest position was that every chair in the lot was carrying some of it. The failure had been visible in service on 23 chairs and the decision was taken on 600.
Against that, the check that would have caught it was two man-days at the Level 3 rate, $598, to sit in the factory, run the five measurements on the sampled chairs, and read the folding moment readings off the factory's own cycle run as they were taken. That is thirteen to one.
But the honest comparison is not the $598 against the $7,772. It is the $22,680 that had already left the buyer's account when the balance was released against a report saying hinge clearance conforming. A number measured on the wrong property is not cheaper than a number measured on the right one. It is only paid later.
Where to go next
What the specification should have said
Nine clauses. None of them needs a laboratory, and four of them are numbers.
Hinge clearance shall be 0.30 millimetres, plus or minus 0.10, measured with a feeler gauge at three positions on the leaf, without forcing the leaf closed, and the reading shall be recorded before and after the pivot is cleaned. The difference between the two readings is the swarf bed and shall not exceed 0.05 millimetres.
Free travel at the leaf, measured with the leaf unforced, shall be not less than 0.20 millimetres on a new chair and not less than 0.15 millimetres at the end of the cycle test.
The pivot shall be free of ferrous swarf and of loose coating particles, and the shallower of the two clearance readings shall be the one used for acceptance.
The folding mechanism shall complete 5,000 open and close cycles under a load of 50 kilograms at the seat, at a rate of not more than six cycles a minute, on three chairs from the production lot, with the peak folding moment recorded at cycle 1, cycle 100, cycle 1,000 and cycle 5,000.
Peak folding moment shall not exceed 4.0 newton metres at any cycle, and shall not increase by more than 0.5 newton metres between cycle 1 and cycle 5,000. A jammed chair on this lot needed 11 to 14 newton metres, so this clause would have caught it with a wide margin.
The seat-to-back angle shall be 100 degrees, plus or minus 1.5, on every chair at cycle zero and at the end of the cycle test, and shall not drift by more than half a degree over the test. Any drift indicates a pivot that is moving and is a rejection on its own.
The cycle test shall be witnessed. A certificate from the factory is not acceptance: the buyer's inspector attends the run, reads the moment readings as they are taken, and takes the record with readings at every milestone rather than a summary.
A jam during the cycle test, a folding moment above the limit, or a drift in the backrest angle is a critical defect with zero acceptance, and the lot is not accepted on a pass certificate for the balance of the order.
The purchase order shall state the number of cycles and the acceptance numbers, and shall name the standard it draws them from. Where EN 581-2 is the governing standard, the order shall say in terms that EN 581-2 contains no stacking or folding cycle test, and that the cycle requirement in this order applies in addition to it.
Nine clauses is not an unusual amount of specification to add to a folding chair order. It is less specification than most people put into a finish colour. The difference is that these nine can all be checked on a factory floor by one person with a feeler gauge, a torque wrench and an angle gauge, and every one of them is a number that a report can defend.
Where to go next
Where the balance ended up
The factory did not argue. The swarf came out of its own cutting and coating lines, so the pivot rework, the air freight on the 23 replacements and the re-inspection were all settled at the factory's cost, $4,057 of it. The buyer's own $3,715 of hire furniture was recovered by agreement four months later, which is the part nobody puts in a cost table: the money is only back when the claim is settled.
We took the reworked lot at Level 3, sampling 80 chairs, and ran the five measurements on each: free travel as found, cleaned clearance, swarf mass on a dismantled pivot, folding moment and residual angle. Every reading came back inside the new clauses, and the 80 chairs were pulled from the middle of the run rather than from the pallets nearest the door, which is a habit worth having.
We also attended the factory's 5,000 cycle run on three chairs and read the folding moment at each milestone as it was taken. That was the part that mattered most, and it came from the buyer's own addition to the scope: pull test chairs from the bottom of a trial stack as well as the top. The bottom of a stack is a different place from the top, and a test that only uses chairs out of a carton will never know it.
Where to go next
The three questions to put in the next enquiry
Before asking what a check costs, ask what the number in the specification is attached to. Three questions, in writing, before anything is cut:
How many times does this chair get folded and unfolded in a year, and what is one cycle in your definition? Then ask how many cycles the hinge was run to in the test, and for the folding moment readings at each milestone rather than a certificate.
What is in the pivot when the chair is packed, and how do you know? A number for hinge clearance is not a number for a hinge that has never been folded, and we would like to know what the joint looks like after 5,000 cycles.
If we asked you to run three chairs through 5,000 cycles in front of us and read the readings as they are taken, how long does that take and does it change your price?
This defect was not going to be found by looking harder at a chair. It was going to be found by asking a question the agreed scope had no room to answer, and then by writing the number into the order before the goods were made. A stacking chair inspection that stops at the clearance reading, however carefully that reading is taken, will keep passing this lot.
Where to go next
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.