240 restaurant tables, a coating check that passed, and rust at the leg joint eleven weeks after delivery. The report was accurate and the reading was taken in the wrong place, because the purchase order never said where to 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 that describe the coating
A restaurant group was refitting twelve sites and ordered 240 tables of one model from a factory in Foshan. A 1800 by 900 millimetre top at 750 high, on a welded square tube steel frame, powder coated in RAL 7016 anthracite grey, packed four to a carton. The unit price was $118 and the order was $28,320, with thirty per cent paid as a deposit and the balance released against a pre-shipment inspection report.
The coating was specified in four lines. Frame powder coated, polyester, to a minimum dry film thickness of 70 microns. Salt spray to 100 hours under GB/T 10125. Cross-cut adhesion class 1 to GB/T 9286. Coating thickness to be verified on site before loading.
Every one of those four lines is checkable, and every one of them was checked. What they do not say is where. There is no measurement point in the specification, no minimum individual reading, no separate requirement for the corners and welds of the frame, and no rule for what a coating thickness number is allowed to be an average of. A thickness requirement without a measurement point is not yet a requirement, and this is a restaurant table inspection china order that turned on exactly that gap.
This is an illustrative example built from the pattern behind coating disputes on commercial table orders: the readings and the arithmetic are representative of the category rather than of any particular order.
Where to go next
→ the tables scope, for commercial tables checked against contract figures
What the pre-shipment inspection measured
The lot was sampled under ISO 2859-1. Two hundred and forty tables fall in code letter G, general inspection level II gives a sample of 32, and at AQL 2.5 the acceptance number is 2 and the rejection number is 3.
On each of the 32 sampled tables the coating was measured with a magnetic induction gauge on the flat outer face of the top rail, which is the flattest, largest and easiest surface on the frame. Readings ran from 88 to 94 microns. The mean of the flat rail readings across the sampled tables was 91 microns.
The report recorded the coating as conforming. The lot passed. The container was booked.
Where to go next
What the estate manager found eleven weeks later
Eleven weeks after delivery, the buyer's estate manager sent photographs from four of the twelve sites. They show a narrow orange-brown band at the junction where each leg meets the apron that carries the tabletop, running the full width of the weld. On two of the tables the bloom is heavy enough to have marked the underside of the top.
The photographs came with a note. The frames are rusting at the leg joints, the tables are in dining rooms, and the chain is not going to take them out of service.
The factory replied with its pre-shipment report, which showed ninety microns. Both documents were accurate. This is the part of a restaurant table inspection china file that gets argued about, because each side is holding a true number and the two numbers are about two different places.
Where to go next
→ what to confirm before the balance payment leaves
→ the eight parts of every verification report, and which two decide the outcome
Two true numbers about two different places
This is where the case turns, and it is worth being precise about it, because the instinct is to treat it as a dispute about who measured correctly. Nobody measured incorrectly. The gauge was put on the flat face of the top rail because that is a flat face, the number it returned was 88 to 94 microns, and that number describes the coating on the top rail accurately.
What the number does not describe is the coating where the leg meets the apron. Those are different surfaces with different physics, and the specification named neither of them. Ninety microns on the rail and forty-five microns at the joint are not two versions of the same fact. They are two facts about two pieces of steel, and the report quoted the one that was easy to quote.
A frame that is ninety microns on the rail and forty-five microns at the joint is not a frame with a defect. It is a frame with two coatings in one part number, and until somebody says which surface the requirement applies to, the number on the rail is the only number that can be compared against the specification.
Where to go next
Why the joint could not be measured with the gauge
The obvious move is to put the gauge on the joint and read it. That is where the standard stops you, and the way it stops you is the useful part of this case.
ISO 2808 is the method standard for film thickness. GB/T 13452.2-2008 is the Chinese adoption of it and is identical to ISO 2808:2007, and ISO 2808 itself was revised in 2019. Clause 6.3.3 of the standard limits the magnetic methods, and the limit is physical rather than procedural. A magnetic induction gauge measures the change in a magnetic field between probe and substrate, and that field is shaped by the geometry of the substrate, its thickness, its material and its roughness.
There is a second limit that bites harder here. Readings taken within 10 mm of an edge are not valid, because the field at an edge is distorted by the edge itself. A welded leg-to-apron junction is a corner with a bead at the toe. It is not flat, and it is not 10 mm from an edge. It is the worst case of both limits at the same time.
Add the surface and it gets worse. A weld bead is not a smooth substrate. ISO 19840 gives the correction for rough surfaces using the surface profile comparator grades of ISO 8503-1, and the correction is subtracted from the reading: 10 microns on a fine profile, 25 on a medium profile and 40 on a coarse one. On a bead that has been ground but not polished, the correction on its own is larger than the difference the buyer is trying to resolve.
- Flat surfaces only. Clause 6.3.3 of ISO 2808 permits magnetic and eddy current methods on flat surfaces, because substrate geometry changes the field.
- Not within 10 mm of an edge. Readings close to a boundary are invalid for the same reason, and a weld toe is a boundary.
- Corrected for profile on a rough substrate. ISO 19840 and the ISO 8503-1 comparators, 10 microns for fine, 25 for medium and 40 for coarse.
- Method and edition have to be written down. GB/T 13452.2-2008 follows ISO 2808:2007 and the current international edition is 2019, and the two are not the same document.
Where to go next
→ how we work through what can be checked where
→ the eight parts of every verification report, and which two decide the outcome
So the check had to change method
The re-check at the factory was commissioned with a single instruction: measure at the junction. The method changed with the instruction, because the gauge was the wrong instrument for that surface and no amount of care at the frame would have changed that.
On each sampled table the inspector took a sacrificial frame, cut the leg-to-apron weld through, mounted and polished the section and measured the coating on the cut face under a microscope, which is the cross-section method of ISO 2808. Nothing was measured with a gauge at the joint. Cross-cut adhesion was then run across the junction rather than on a flat panel, and the toe was examined visually for continuity before the section was taken.
| Where it was measured | Method | Result | Against the specification |
|---|---|---|---|
| Top rail, flat outer face | Magnetic induction gauge to GB/T 13452.2 | 88 to 94 microns | Conforms |
| Leg outer face, flat, 300 mm above the floor | Magnetic induction gauge | 82 to 86 microns | Conforms |
| Leg-to-apron junction, 10 mm from the weld toe | Cross-section under microscope | 45 to 48 microns | Below the 70 micron minimum |
| Leg-to-apron junction, at the weld toe | Cross-section under microscope | Coating absent over a 2 mm band, bare steel with red rust already present | Fails |
| Cross-cut adhesion at the junction | GB/T 9286 | Class 3 | Class 1 required |
Where to go next
→ a sample report you can read section by section
→ the tables scope, for commercial tables checked against contract figures
The number that passes and the number that fails
This is the arithmetic that decides whether a lot is accepted, and it is worth writing out because the wrong version of it is extremely common.
ISO 19840 sets the acceptance rule for a coating system, and the same logic applies to any dry film thickness specification. The arithmetic mean of the corrected readings must be at least the nominal dry film thickness. Every individual reading must be at least 80 per cent of the nominal. The number of readings falling between 80 and 100 per cent of the nominal must be less than 20 per cent of the total. No reading may exceed three times the nominal.
Apply that to a 70 micron specification. The floor for any single reading is 56 microns, because 70 times 0.8 is 56. A reading of 45 is not marginal. It is 11 microns below the floor, and it fails on its own, whatever every other reading on that table does.
Now the average. Six flat face readings at 90 microns and two junction readings at 45 microns give a mean of 78.8 microns across the eight points, which clears 70. The table passes on the average and fails on the individual reading. Both results are correct, and which one governs is decided entirely by whether the specification says so.
80 per cent of 70 is 56. Two readings of 45 sit 11 microns below that floor. Eight readings of six 90s and two 45s average 78.8.
Why a welded corner is always the thinnest place
None of these three is fixed by spraying the corner longer, which is the point. A welded corner cannot be brought to 70 microns by more effort at the gun. It has to be designed, ground and pre-treated differently, and the specification has to say so before the frames are cut.
- Electrostatic shielding. Powder is charged as it leaves the gun and is drawn to the grounded workpiece. Where two grounded surfaces meet at an angle, the field between them screens the incoming powder, so the inside corner of a joint receives less material than the open faces do. This is the Faraday cage effect, and it is why some processes preheat castings before coating to lift the part away from the workpiece.
- Shrink-back from the edge. Once the powder melts, the film is liquid and surface tension pulls it. A sharp edge has nowhere to hold it, so the coating retreats from the corner and leaves a bare band along the toe. That band is not thin. It is missing, and bare steel underneath it starts to rust.
- The radius. Measurements on coated substrates with a 5 mm radius and a 0.5 mm radius edge put the coating in the edge region of the 5 mm radius part at roughly 40 per cent above the coating on the adjacent side face. The geometry of the edge decides the coverage, not the powder.
Where to go next
→ outdoor furniture checks that start from the coating and the metal underneath
→ the tables scope, for commercial tables checked against contract figures
Where the rust starts
Two millimetres of bare steel at a weld toe, in a steel frame, in a restaurant, is not a cosmetic defect. The junction is a crevice, crevices hold water and detergent, and a bare steel edge in a wet crevice turns from grey to red faster than most people expect. The frames were photographed at the factory with rust already present at the toe.
The ninety microns on the rail does nothing for this. It is ninety microns of good coating on a piece of steel that is not the piece of steel that is failing.
For scale on the thickness difference: 45 microns is 64 per cent of the 70 micron minimum, and it is half the thickness on the flat rail. At the toe it is not 64 per cent of anything, because at the toe the coating is absent. A restaurant table inspection china buyer who reads only the flat rail number learns nothing about the part of the frame that is in contact with water every night.
Where to go next
→ what to confirm before the balance payment leaves
→ the tables scope, for commercial tables checked against contract figures
What the salt spray number was actually testing
The specification asked for 100 hours of neutral salt spray under GB/T 10125, and GB/T 3325-2024 for metal furniture asks the same of the coating. The factory held a report showing the coating passed, and the report was genuine.
It was a report on a flat test panel.
A flat panel has no weld, no corner, no radius and no toe, and the coating goes onto it in a way it does not go onto a welded frame. It is also worth saying that AAMA 2604, the specification often written into powder coat clauses for furniture, is an architectural coating specification for aluminium extrusions and panels. Its minimum film thickness is 30 microns and its salt spray requirement is 3000 hours. Citing it for a welded steel furniture frame moves a document from one object to another, which is precisely what happened here.
Two documents, both true, neither of them about a leg-to-apron weld.
Where to go next
How the report was written
The report was written against the four lines of the purchase order rather than around them, because the buyer reads it against those four lines and nothing else. Coating was given its own numbered section ahead of the general physical checks, because a coating result and a dimensional result are closed in different places and in different ways.
The section that carried the most weight was the last one. It stated that no coating thickness was measured by gauge at the leg-to-apron junction, because the method does not permit it, and that the junction results were obtained by sectioning one frame per sampled unit. That sentence is what allowed the buyer to see that the earlier ninety microns and the later forty-five microns were both correct, and that the first report was measuring a different surface.
| Section | What it decided for this order |
|---|---|
| 01 Order identification | Whether the four coating lines name a standard, a method, a minimum and a salt spray duration, and where the measurement points are, which was the finding |
| 02 Reference documents | That the salt spray report covered a flat panel and the AAMA clause covered aluminium extrusions, neither of them the welded frame |
| 03 to 05 Counting, materials, dimensions | The evidence that the rest of the table was sound and that the coating was the only open item |
| 06 Coating and physical check results | The location, method and value of every coating reading, including the sectioned junction at 45 to 48 microns and the 2 mm bare band |
| 07 Findings and recommendation | That the lot failed on the individual reading floor rather than on the mean, and the conditions to close before loading |
| 08 Items we could not verify | That no gauge reading at the junction is valid, that the salt spray result is a coupon result, and that a section is destructive |
What it cost
The re-check at the factory was two man-days, $598. It found the defect before the container was sealed.
The alternatives on the table were worse. Having the frames stripped, the toes re-ground, the coating redone and the goods re-packed at the factory is about $28 a frame, $6,720 on this order, and it is a rework the factory has to absorb. The same job done in the field, with the tables in twelve dining rooms and the buyer paying a contractor and the downtime, is closer to $155 a table, $37,200.
$37,200 divided by $598 is more than sixty times. And the $19,824 balance was never released while the question was open.
Where to go next
Ten clauses for the next purchase order
The order was not badly written in intent. It named the right properties and asked for the right checks. What it did not do was close the gap between a property and a place, and the gap is where the whole case lived.
- Name the measurement points, not the part. Write them as positions a factory can find: the leg-to-apron weld junction at 10 mm from the toe, the toe itself, the underside of the top rail, the leg at 300 mm above the floor, and every other welded transition. On the frame is not a measurement point.
- State the method for each point separately. Magnetic induction gauge to GB/T 13452.2 on flat faces only. Cross-section under microscope at corners, welds and anywhere within 10 mm of an edge. A gauge reading at a joint is not a measurement.
- State the acceptance rule, not just the minimum. Minimum dry film thickness 70 microns; arithmetic mean of all corrected readings at least 70 microns; no individual reading below 56 microns; no reading above 210 microns. Write the 80 per cent floor into the order so an average cannot carry a bad point.
- Set a separate minimum for joints and corners. A welded assembly cannot hold 70 microns everywhere. Ask for 50 microns at any welded transition and no visible discontinuity at the toe, and state how that will be measured.
- Specify the edge treatment. Minimum radius R2 on exposed edges, weld toes ground and rounded before coating, and the radius used declared on the drawing. The radius does more for coverage than any amount of extra gun time.
- Specify the pre-treatment. Phosphate or zinc pretreatment before powder, with a wash and conversion stage that reaches into the joint. Coating a frame whose welds were only wire brushed is coating a scale.
- Ask for the cross-cut at the junction and not only on a flat panel. Cross-cut adhesion class 1 to GB/T 9286 at the leg-to-apron weld, and class 1 on a flat panel as well.
- Ask for the salt spray specimen to be a welded frame. A 100 hour NSS panel cut from the batch and including a leg-to-apron weld, not a flat test coupon from the powder supplier.
- Require a first-article check at the junction before the batch is coated. One sacrificial frame, sectioned, with the readings sent to the buyer. This is the cheapest moment in the whole order to find a 45 micron corner.
- Require a location and a method next to every reading in the report. A number without its place on the object is not a result, and the buyer cannot use it later. Junction at the toe, sectioned, 45 to 48 microns, rather than coating thickness 45 to 48 microns.
Where to go next
Three questions to ask before the next delivery
Which points on the frame was the coating measured at, and was any reading at a corner taken with a gauge or on a section.
Does the specification carry an individual reading floor, or only an average. A floor of 56 microns on a 70 micron minimum catches a 45 micron joint. An average does not.
Was the salt spray specimen a flat panel or a welded frame, and does it include the joint where the legs meet the aprons.
None of these three questions needs a factory visit, a gauge or a sample size. They need the report and the purchase order, and they can be asked before a single container is loaded. This order produced no falsified document and no substituted material. The pre-shipment report was accurate, the salt spray report was accurate, and the coating was thinner than the specification allowed at the one place the specification did not mention. That is the ordinary way these things go wrong. Not through dishonesty, but through a requirement that never said where to look.
Where to go next
→ send us the order details and we will tell you what a check can cover
→ the tables scope, for commercial tables checked against contract figures
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.