60 conference tables were banded while the plywood panel was at 12.4% moisture, and 9 of the 20 tables tested had a band that would not stay on

The order, and the two lines that decided it

A European contract furniture supplier ordered 60 conference tables for an office fit-out: a 1,800 by 900 mm top on a black powder-coated steel frame with a cable tray, a 25 mm plywood core, melamine faced on both sides, and a 1.0 mm ABS edge band on all four edges. The unit price was $460, the order was $27,600, 30 per cent was paid on order, and the remaining $19,320 was held against a pre-shipment inspection report.

The purchase order ran to nine lines. Top 1,800 by 900 by 25 mm, plywood core, melamine faced, edge banded on all exposed edges. Edge band 1.0 mm ABS. Melamine face and band colour to match the approved sample. Steel frame, black powder coat, RAL 9005, with a cable tray under the top. Assembly by the buyer. Four tables to a carton. Balance released against a pre-shipment inspection report.

Two of those nine lines decided the case, and neither of them was about glue. The line for the edge band stated a material and a thickness and nothing else: 1.0 mm ABS, all exposed edges. It gave no bond strength in newtons per millimetre, it said nothing about the state of the panel at the moment the band went on, and it did not require a piece of the panel to be kept back so the bond could be tested later. The line for the sample pointed at an approved board with the note as per approved sample, which is a reference to appearance and construction, not to a number that can be read off an instrument.

Everything the buyer actually wanted to know was therefore absent, and the one thing that was present pointed at the wrong document. A lifted edge band is not an appearance question, it is a bond. A bond has a number, and a number missing from a purchase order does not get supplied by an approved sample. But a lifted band can be seen, heard and counted, and those are the things an inspector can do standing next to a table with a steel rule in their hand.

The line in the inspection request

The buyer's inspection request had eleven items on it, and the last one was written by somebody who had been disappointed before. Check the edge band adhesion, particularly at the mitred corners, and report any band that lifts. That request is the reason this case exists at all, because the standard pre-shipment walk around a furniture order does not include a knock on the edge of a table.

The knock test is not a formal standard and it takes ten seconds. The inspector ran the back of a steel rule along the band on the first table, slowly, in the way the request had described. On the long edges the note was solid and flat. At the mitre corner on the left-hand end of the top, the note changed. It went hollow, it stayed hollow for about 200 mm, and at that point a strip of band came away from the table in the inspector's hand, cleanly, along a line where the glue should have been.

It took ninety seconds to find the first defect and four seconds to understand that the lot was in trouble. A lift that starts at a mitred corner and runs along the end band is not a handling injury. Handling injuries are random, they sit on exposed corners, and they are one or two pieces. This was a bond letting go along the line where the glue film is thinnest, and if the first table off the packing line had it, the line was not going to improve by table number forty.

The rest of the lot was walked through the same check, because the cost of stopping at table one is that you learn nothing about how much of the lot is affected, and the cost of not stopping is that the buyer finds out at the customer's office. A conference table inspection that treats the top as a solid surface and never touches the edge will pass this lot, and passing a lot is a report that says less than it appears to.

The knock test, and what it found across 20 tables

The knock test was written down before the lot was opened, in the same form as the request, so that the number in the report could be reproduced by somebody else. A rubber mallet was used for the full sweep and the back of a steel rule for the corner work, and every banded edge on every sampled table was tested along its whole length. A hollow note anywhere on an edge counted as a defect. Nothing was peeled to find out whether the note meant anything, because peeling is the destructive step and it comes later.

The lot of 60 was sampled at 20 tables. Under ISO 2859-1 at general inspection level II, a lot of 60 is code letter F and the sample is 20, and the tables were drawn at random across the lot rather than off the top of the stack, so that a band run early in the week was not treated as a band run all week. Nine of the twenty carried a hollow note on at least one end band, which at AQL 2.5 for a major defect is a rejection on that characteristic, since a sample of 20 accepts one and rejects at two.

The peel test, and the one place it was allowed to happen

The count established that the lot had a problem. What the report needed next was the reason, because a band that lifts and a band that is weak are not the same statement and they have different fixes. A band that lifts at one corner can be a handling and packing problem. A band that peels off along the glue line everywhere it is checked is a bonding problem, and it needs a bond.

The only way to tell them apart is to peel something, and this is where an inspection usually stops, because the goods are the buyer's goods and nobody has agreed to have pieces of them removed. That constraint was real and it was resolved here at almost no cost. The end band on a conference table runs around the 900 mm width of the top, and at the bottom of that end there is a steel leg bracket screwed across it. The inspector asked the buyer for written permission to take three 50 mm test strips, and the buyer approved it, because the request was specific: three strips, from the underside of the end band where the leg bracket covers them, on three tables the factory had already agreed to re-band that morning. The strips were handed back with the report, so nothing left the factory except a piece the buyer already knew about.

The method was the 180 degree peel on a 50 mm strip at 100 mm a minute to the EN 311 method for edge banding, with the force recorded in newtons per millimetre of band width. Both the peak force and the failure mode were written down, because the mode is the part that explains what happened.

What the failure mode told them

The failure mode is the whole answer. On the three tables the glue film let go of the plywood and stayed on the plywood, so the band came away whole and the adhesive was still sitting on the core. That is a wetting failure: the glue arrived at a surface it could not grip. It is not a band that was too soft, not a band that was the wrong colour, not a corner cut badly. On the first article the band itself tore and the glue line stayed intact, which is what a healthy bond looks like.

The factory had that first article on the shelf and offered it as evidence that the process was sound. It had been banded six weeks earlier, from a different roll, on panels that had been in the factory for a month, and once that was pointed out the factory withdrew the document. This is the ordinary shape of these cases: the factory is not hiding anything, it is holding up the piece of evidence that happens to be about a different batch of panels.

Why a band peels off a table that was built correctly

A conference table with a plywood core, a melamine face and an ABS band is a straightforward assembly, and there is nothing difficult about making one that holds. The reason this lot did not hold is the order the steps happened in, and the panel was in the wrong state when the band went on.

The factory keeps a stock log for its panel store, and the log records the moisture content of incoming plywood twice a week, measured with a resistance meter on the panel face. The reading for the week this lot was banded was 12.4 per cent. The same meter, used the same way on the same panels in the factory warehouse on the day of the inspection, read 8.6 per cent. The panels had moved 3.8 points of moisture in 22 days.

The report did not calculate a shrinkage figure from those two numbers, and the reason is worth stating because it is a rule rather than an admission of weakness. Panel movement across a four point moisture change depends on the grain direction of the face veneer, on the layup of the core, and on which way the panel was cut, and a number produced from none of those is a number that will not survive a dispute. What the report did instead was record both readings, record the failure mode, and let the mechanism carry itself: the panel lost moisture, it got thinner across its edge, and the band had been bonded at the thickness the panel had while it was wet.

A band that is snug against a wet edge ends up proud of a dry one. The band does not shrink with the plywood, because ABS is dimensionally stable and the plywood is not, and the difference shows up as a stress at the corners, because the mitred corner is the one place on the table where the band changes direction and where the glue film is thinnest, since a mitre is a butt joint and a butt joint carries less adhesive than an overlap. So the sequence runs: the panel dries, the band stands proud at the corners, the mitre joint opens first, and the loose strip then peels along the end band away from the corner. That is why nine of twenty tables showed the lift at the corner rather than at the middle of the edge.

It is also why the long edges were sound. A 1,800 mm band on a panel that is shrinking across the 900 mm width has length to give, so it flexes rather than peels. The 900 mm end band has nowhere to give, so the dimensional change goes into the joint.

What the report could fail, and what it could not

Two findings were failable and two were not, and the split came entirely from the purchase order. The lift itself was failable. A band that lifts, a corner that opens, a mitre with a visible gap and a hollow note along 200 mm of an edge are all things that can be seen, photographed and counted on a table standing in the workshop, so the report recorded the lift as a non-conformance against the line requiring all exposed edges to be banded, and against the reference to the approved sample, and named the corrective action.

The peel strength could not be recorded as a non-conformance, because the purchase order contained no figure. The inspector had a method and three readings: 0.6, 0.9 and 1.1 newtons per millimetre of band width. The order gave no number to fail them against. The report stated the readings, stated the method, stated the failure mode at the glue line, and said plainly that the order was silent and that the decision on the balance belonged to the buyer. Writing a limit the buyer had not specified would have put the whole document in doubt, and a report that fails goods against a standard it supplied itself is a report nobody can rely on.

The moisture reading could not be recorded as a non-conformance either, for the same reason and a second one. The order said nothing about the state of the panel when it was banded, and the 12.4 per cent figure came out of the factory's own stock log, which is a document the factory controls. The inspector quoted it, dated it, and put it in the report as the factory's own record. The second reason matters more: the inspection happened after the panels had already dried, so the inspector could measure 8.6 per cent on the day and could only infer the 12.4 from a log. A number that comes out of somebody else's notebook is a lead. A number that comes out of the inspector's own instrument is a finding, and the report kept the two apart.

There was also a piece of paper the buyer had been shown and believed. The factory's edge banding specification sheet for that line said the band was applied with a standard hot-melt adhesive at a pot temperature of 180 degrees and that the substrate was to be clean and dry before banding. Dry. One word, and it is the word the order should have carried, because it is the word that would have stopped this lot at the bander in week one, with sixty tables still sitting in the workshop and no rework to pay for.

Nine days to re-band sixty tables

The factory's first proposal was to re-band the nine affected tables and ship. That would have been the wrong call, for a reason the moisture numbers make obvious. Nine of twenty tables failed the knock test, and the knock test only detects a band that has already opened. A band that is still holding on a table that was banded at 12 per cent moisture is bonded to a core that has since lost 3.8 points, so it is carrying load that has not shown itself yet. Shipping a lot where nine tables are known bad and forty have never been checked is not a rework, it is a delay with extra steps.

The lot was conditioned instead. The 60 tops were moved into the covered area of the factory, stacked with the edges exposed and timber spacers between the stacks so air could pass across the banded edges, and left for four days while the remaining moisture in the panels evened out. The tops were then re-banded over five days at about twelve tables a day: the old band stripped, the edges sanded back to remove the old glue film and the sanding dust, wiped, and new 1.0 mm ABS band applied with the primer the bond needs. Getting the right band onto the line took one decision, and it is worth reporting because it was the closest this order came to a second non-conformance.

The re-inspection was not a re-run of the knock test on a few tables. Every banded edge on all 60 tables was knocked, which is 240 edges, and the inspector did it himself rather than delegating it, because the whole argument of the case rested on the test being done the same way twice. Five witness strips were peeled at 1.8, 1.9, 2.0, 2.2 and 2.3 newtons per millimetre, and in all five the band itself tore with the glue line intact, which is the result a healthy bond gives. The balance of $19,320 was released after that and not before it.

The rework cost the factory nine days of line time and one roll of band. It cost the buyer nine days of fit-out programme, and it nearly produced a second defect of exactly the same kind: sixty tables were re-banded, and the inspector checked that the new band went onto a panel at 8.6 per cent rather than a panel at 12, by having the factory log the moisture on the day of re-banding. That log is the buyer's protection in the next order, not a curiosity in this one.

The band material decides which failure you get

Band material is often treated as a visual decision, and it is not. The three common choices fail in three different ways, and the difference matters because the failure mode tells you where to look. The list below is the part of this case worth keeping even though the band on these tables was the specified one, because the next order is the one that gets the wrong band.

• ABS, 1.0 to 2.0 mm, the default for contract office furniture. Tougher than the panel, takes paint, takes a solvent without marking, and holds a hot-melt bond on a clean porous substrate. It does not follow the core across its thickness, so it is unforgiving of a panel that is banded wet.
• Polypropylene, 0.8 to 2.0 mm, cheaper and softer, and the one most often proposed when a roll is on allocation. PP has a low surface energy and does not wet with a hot-melt glue without a primer or corona treatment, and it moves with heat and moisture more than ABS does. On a panel banded wet, PP is the combination most likely to let go at the corner.
• Melamine paper, 0.5 to 0.8 mm, thin and cheap, and the right choice for a melamine faced panel on a budget. It fails differently. The paper is weak, so the band tears and the failure looks like a crack along the edge rather than a lifted strip, and a corner that has opened slightly shows a hairline instead of a gap.

Why the material belongs in the order

That decision was the band itself. The ABS roll the line had been using had run out, the delivery was three days out, and the factory proposed re-banding the lot with 0.8 mm polypropylene instead of 1.0 mm ABS.

The inspector asked the buyer for a decision rather than agreeing on the floor. The answer was that the substitution would not be accepted without written approval and a test strip first, and the ABS roll was found in the same warehouse in a different size.

The band on the delivered tables was confirmed as ABS. The goods-in label for the roll said 1.0 mm ABS, a caliper on the edge read 1.0 mm, and the band took paint without lifting, which a polypropylene band of the same thickness would not have done. So this case is not a substitution case. It is a bonding case, and the substitution only entered it at the rework, where a buyer who had written the material into the order had nothing left to argue about.

How the specification should be written

None of the lines below cost the factory anything to comply with, and every one of them exists because a specific measurement was missing from the version that was signed. They are written as they would have appeared in the purchase order.

• Edge band: 1.0 mm ABS, applied to all four exposed edges, colour to match the melamine face. A different material or thickness is a substitution and requires the buyer's written approval before the band is applied, not after the lot is banded.
• Bond strength: the band shall have a peel strength of not less than 1.5 N/mm of band width, determined by a 180 degree peel on a 50 mm strip at 100 mm a minute to the EN 311 method for edge banding. The figure applies to the finished table, not to a first article.
• Failure mode: the peel shall break in the band or in the panel face. A peel that separates cleanly at the glue line is a reject, whatever the force reading says.
• Substrate condition at banding: the moisture content of the panel at the moment the band is applied shall be not more than 10 per cent, measured with a resistance meter on the panel face, and the reading shall be logged for each lot of panels banded on that shift. Panels above 10 per cent are to be conditioned and re-read before banding.
• Edge preparation: the panel edge is to be sanded and wiped clean of dust, oil and glue residue before banding. The edges of a melamine faced panel are to be lightly sanded to break the surface film. No filler or putty on the edge unless the purchase order says so.
• Corner treatment: the band is to be mitred and glued at the corner, with adhesive applied to both faces of the joint, and the mitred joint is to be reinforced from behind where the construction allows. A butt mitre with adhesive on one face only is not acceptable.
• Witness pieces: for each production shift, three pieces of the panel being banded are to be banded, marked with the date, the shift and the panel lot number, and retained by the factory for 90 days, to be made available to the buyer or the inspector on request. These are the only way to test the bond without cutting into the goods.
• On the goods: the band shall not lift at any point, and the mitred corner shall show no gap greater than 0.3 mm when probed with a 0.2 mm feeler gauge.
• Delivery condition: the tables are to be delivered with a moisture reading for the lot recorded on the packing list, and the panels are to have been conditioned, not merely stored, before banding.
• Sampling: the edge band check is a major characteristic at AQL 2.5 under ISO 2859-1 at general inspection level II, on a sample drawn from finished tables rather than from cartons, and the knock test is applied to every banded edge of every table in the sample.

What to carry to the next conference table order

Nothing on this order was fabricated and nothing was mislabelled. The factory built 60 conference tables to the order, banded them with the material that was on the line, and packed them in cartons that were counted correctly and marked correctly. The defect came from a panel banded while it was wet and a mitred joint that was never given adhesive on both faces, and neither of those two things was a line in the purchase order.

The thing worth carrying away is that edge banding is a bond and not a trim. It has a number, the number is easy to measure, and the measurement costs one steel rule and ninety seconds. A buyer who writes three lines about the state of the panel, the witness pieces and the peel strength will not have this conversation again. A buyer who writes only edge band 1.0 mm ABS has asked for a part and left the joint up to the factory.

And keep the two things apart. The bond was weak because the panel was wet, and the corner opened because the mitre had adhesive on one face only. Fixing the moisture would have made the band hold, but the corner would still have been the weak point in every table. Fixing the corner would have saved the nine tables that failed, but the fifty-one tables that passed the knock test were bonded to a panel at 12 per cent and had not finished moving yet.