Every face on every table top was genuine ash. Underneath it sat 16.8 mm of particleboard, and a specification that never said how much wood there had to be.

The order said solid ash, and the legs were solid ash

A European restaurant furniture importer ordered one hundred and eighty dining tables from a factory in Guangdong. Each table was 1,400 by 800 mm, 750 mm high, with a twenty-five millimetre top and a solid ash frame. The purchase order and the buyer's own written specification both used the same short phrase: solid white ash. Nothing else in the document said how much of it there had to be.

The factory made the legs, the apron rails and the stretcher from solid ash, cut from kiln-dried timber the buyer had approved at the sample stage. The tops were a different thing entirely. They were twenty-five millimetre panels taken from a board supplier's standard line: 4.1 mm of ash veneer on each face over a 16.8 mm particleboard core, finished with a one millimetre edge band. Total thickness twenty-five millimetres, which is exactly what the dimensioned drawing asked for.

Every visible surface of every table top was genuine ash. The colour, the grain, the ray fleck, the sheen under the clear coat, all correct. A substantial share of the order was exactly what was bought, and the part that was not was the part the buyer had described in the most detail.

This is the shape veneer substitution almost always takes. It is not a counterfeit and it is not a bad product. It is a compliant-looking panel in a compliant-looking carton, and it is wrong in exactly one place: the inside, which nobody looks at until somebody cuts it open.

Why the species check passed

The first thing an inspector does with a solid-wood order is identify the wood, because a wrong species is the fastest way to lose an argument with a factory. The inspector took a clean section from the underside return of a top and ran the standard comparison: growth ring spacing, pore arrangement, ray fleck, and the smell of a freshly cut surface.

It came back ash. It was ash. Four millimetres of it on the face and four more on the back, with a chipboard-like core in between that a surface check has no way to see.

This is the uncomfortable part of the case, and the part worth keeping. The statement that the wood is ash was true. As a piece of information it was also close to worthless, because ash is a species and a surface tells you the species. It does not tell you how deep that surface goes, what sits under it, or which parts of the order the word applies to. A face four millimetres deep and a face twenty-five millimetres deep are indistinguishable from above, indistinguishable under a loupe, and indistinguishable to a smell test, because both are a photograph of the same tree.

A wood species verification ash result taken from a face is a statement about the layer you touched. That is a genuinely useful result when the whole part is one piece. It is not a result about the part. So the species check passed on all one hundred and eighty tables and told the buyer nothing whatsoever about the tops.

The buyer's approved sample did not close the gap either. It was a solid ash offcut, and a panel whose face is sliced from the same species matches that offcut in colour and grain perfectly. Sample approval is not construction approval, and a factory that saw the sample had no reason to believe the panel it was ordering was not what had been approved.

One offcut, and what a section shows

The inspector asked to see the panel offcut bin at the cutting table. Cutting a 1,400 by 800 mm top out of a 2,440 by 1,220 mm sheet leaves a strip, and strips from the top runs carry a full section through the panel: face veneer, core, back veneer. This is the standard way to see a panel without cutting into a saleable table, and it is the reason the offcut pile is worth asking for on every visit. The section was measured with a steel rule and read under a loupe.

Only the stack was wrong

The thickness was right. The species was right. The face and the back were right. Only the stack was wrong, and it was wrong in the one direction the specification had left undefined.

The inspector then took the same section from a second offcut, from a different cutting run and a different shift, and from a part that had been cut through for a scrapped table. Same result each time. That still left the question of extent, because a section taken off a cutting table tells you what was in the top of that stack, not what was in the whole order. That question was answered by paperwork, and the paperwork was worse than the panels.

Four cores, one face

A section identifies the construction, not only the species, and it does so from a single cut face. These are the readings that come off the same cut edge, and every one of them starts with a face that looks identical to the others.

One tell that needed no saw at all

Every one of those faces looks the same, and that is the entire problem. It is why the section is not an optional extra on any order where a panel of any kind is acceptable.

The inspector also noted one detail that required no cutting. The panel edges had been banded, and the band returned under the underside edge of the top. At one corner the return had not closed completely, and a sliver of core was visible from below without a saw being taken to the table. Banding is applied in order to hide a core. That is what banding is for.

Two records that agreed with each other

Before the sections were taken, one routine measurement had already put a question mark on the order, and it is worth recording because it nearly did not. The inspector weighed a solid ash top from the buyer's approved sample and then weighed the tops as they came off the line. The sample was a 400 by 400 by 25 mm block of solid ash. At a density of seven hundred kilograms per cubic metre for kiln-dried white ash, that block weighs 2.80 kilograms. The delivered panel of the same size weighs 2.53 kilograms, because eight point two millimetres of it is ash and sixteen point eight is particleboard at six hundred kilograms per cubic metre. The gap was 0.27 kilograms, or nine point six per cent.

A nine point six per cent gap sounds decisive until you notice how easily a tolerance swallows it. A weight tolerance of plus or minus ten per cent, which is not unusual on a wood tolerance line, would have passed every top in the order. The gap only became useful once it was read against a tight number, and the tight number came later, in the specification.

Across eight tops drawn from the full range of the lot the gaps ran between 1.86 and 1.94 kilograms on a specified 19.6 kilogram top, which is the same nine point six per cent every time. Consistency is what turned a curiosity into a pattern.

The second record came from the factory's own production file. The cut list for the top read: twenty-five millimetre ash-faced panel, four millimetre veneer both faces, particleboard core, edge-banded. And the panel supplier's invoice for the 190 sheets the order consumed described the product in the trade's own words: ash-faced particleboard, twenty-five millimetre. The factory had not misread the order. It had built exactly what it had ordered, from a panel whose own name said what it was, against a specification that had never said the core was not allowed.

The lot, the sample, and why one section rejected a hundred and eighty tables

The lot was one hundred and eighty tables of a single model in a single colour. Under ISO 2859-1 at general inspection level two, a lot of 151 to 280 falls under code letter G, which gives a sample of thirty-two units. At AQL 2.5 that sample would accept two and reject three.

Those numbers were never the deciding factor here, and saying so plainly is more useful than pretending otherwise. The finding was graded critical, because the goods did not match the order. A critical finding carries an accept number of zero. One confirmed section rejects the lot, and it did.

What the sample size governed was everything else. The thirty-two units were checked for dimension, wobble, surface finish, hardware fit and carton marking, and the tables passed those checks cleanly. It is worth saying that a factory did not send a bad order. It sent good tables that were not the tables that were ordered.

The extent of the material finding did not come from extrapolation either, because extrapolation is exactly what a verification report should avoid. It came from the cut list and the panel invoice: the same material went into every top in the order. The affected quantity was one hundred and eighty, not an estimate of some fraction of them.

The report, in eight sections

The finding was written into the standard eight-section report structure, and the sections that carried it were the ones that matter most on a material question.

Section two, reference documents, carried the finding that nobody expected. The factory's production folder held a dimensioned outline drawing showing a twenty-five millimetre top and nothing about its construction. The word solid appeared only in the buyer's written specification, in an email attachment that had never been countersigned and never made it into the factory's document set. The report stated that plainly, because a factory working from a drawing that says twenty-five millimetres is not in the wrong.

Section four, material comparison, listed every part of the table against what was specified and what was found, with the method used to determine it. Section five, dimensions, recorded the top at twenty-five point zero millimetres and confirmed it as compliant, which is the sort of line that makes a report believable. Section seven graded the material finding critical, with the section photographs attached.

Section eight did the honest work. The report stated that only three sections had been taken, that the buyer had not supplied a construction drawing, that the edge band had not been verified because the finish covered the return, and that the factory's previous shipments under the same purchase order were outside the scope of this visit and therefore unverified. A report that cannot say what it did not establish is not worth reading.

The specification, rewritten

The buyer's original specification was not a bad document. It named a species, it named a thickness, it named a finish and it named a delivery date. What it did not do was say anything about construction, and in a contract where the difference between twenty-five millimetres of timber and eight millimetres of timber over a chipboard core is the whole value of the order, that is the difference between a document and a wish.

The rewritten specification changes four lines and adds one. The buyer sent the revised version to the factory with the panel invoice still attached, and the factory signed it without argument, which is the tell that it should have been written this way in the first place.

Part by part, not product by part

The second change was structural rather than verbal, and it is the one that generalises beyond this case. The buyer stopped writing one line for the product and wrote one line for each part, with the construction and the verification method attached to that part.

Product-level wording is where veneer substitution gets in. A table is a table, and one line about it invites one check on it. Parts are where construction actually varies, and they are where the cheap material always ends up: the largest flat panel, the part with the thinnest section, the part nobody opens.

The specification that went back to the factory read as follows.

The commercial outcome

The order was worth thirty-three thousand four hundred and eighty dollars at one hundred and eighty-six dollars a table. Thirty per cent was paid as a deposit. The balance of twenty-three thousand four hundred and thirty-six dollars was still outstanding when the report arrived, which is the only reason this case ended as a negotiation rather than a return.

The factory quoted both replacements. A twenty-five millimetre solid ash top for the specified size was seventy-four dollars against forty-three dollars for the panel it had built, a difference of thirty-one dollars a top and five thousand five hundred and eighty dollars across the order. With the balance still in the buyer's hands, the panel tops were accepted at a credit of that figure, the tops were re-declared in the buyer's own paperwork as ash-faced panel, and the sales material that had called them solid ash was pulled.

That last part is the part that usually gets skipped, and it is the part that should worry buyers most. The loss on this order was five thousand five hundred and eighty dollars of material value on a thirty-three thousand dollar shipment, and the panel tops were a perfectly good table that a restaurant could have used for years. The real exposure was on the buyer's own documents, which continued to describe one hundred and eighty tables as solid ash to the end customers who had already been sold them.

Nothing in this case was difficult to inspect, and nothing was expensive to check. One section off a cutting table, one part-level specification, and one weight tolerance written tightly enough to mean something. The factory did not conceal anything. It ordered a panel whose name said what it was, and built it under a specification that never said the core was not allowed.