In one hundred and eighty of two hundred and forty hotel rooms, the guest chair could not be pushed under the writing desk. The desk was exactly the height on the drawing and the chair was exactly the height on the drawing. The number that decided it was in neither.

Two hundred and forty room sets, and the chair would not go under the desk

A hotel group was fitting out a new property and ordered guest room furniture for two hundred and forty rooms. Each room got a writing desk and a guest chair. The desks were 1,100 by 550 millimetres and seven hundred and twenty millimetres high, with a twenty-five millimetre top, a drawer pedestal on the right and a modesty panel at the back. The chairs had a four hundred and sixty millimetre seat, a seven hundred and sixty millimetre overall height, and arms.

Both products matched their drawings. That was not the problem. The problem was the set. In one hundred and eighty of the two hundred and forty rooms, the chair could not be pushed under the desk at all. The leading edge of the armrest met the front edge of the worktop and stopped, five millimetres of interference. The seat stayed outside the desk footprint, the guest's knees met the front edge of the desk, and the chair looked permanently pushed out.

The other sixty rooms were the accessible rooms, and they had an armless version of the same chair. An armless chair slides under a desk. A chair with arms does not, unless the arms are low enough to clear the underside of the worktop. The buyer had specified both chairs, approved both, and had never once put either one under either desk.

Nothing on this order was made wrong. That is worth saying plainly, because the reflex in a case like this is to look for a factory that cut a corner, and there was not one here. The drawings were right, the production was right, and the set was wrong. No document in the file described the set.

What the trial on the shop floor showed

The inspection was booked as an Order plus Quality Check, two man-days at one hundred and ninety-nine dollars a day. The buyer's specification gave the desk height and the chair height. Neither number told the inspector whether one went under the other, and neither was a dimension the specification asked anybody to measure.

So the inspector took a desk and a chair from the same lot, stood them together on the workshop floor, and tried the fit. The chair stopped at the worktop edge. Then the inspector did the part that turns an observation into a finding, and measured everything instead of describing it.

How the finding was graded, and what it was worth

Every one of those readings is a measurement. None of them is an opinion, and the third and fourth rows are the two numbers the order never contained.

The finding was graded critical, because the goods cannot do the job the order placed them in. A critical defect is judged at zero acceptance under ISO 2859-1: one confirmed failure rejects the lot. All thirty-two chairs in the sample failed the trial, so the outcome did not depend on the grading, but the grading is what sets the acceptance number.

The lot was two hundred and forty chairs. At general inspection level two that is code letter G, a sample of thirty-two, drawn across the batch rather than from the stack nearest the door. At AQL 2.5 that sample would have rejected the lot as well, with two to accept and three to reject, and thirty-two defects is thirty-two defects.

The fit trial itself took under an hour. The whole inspection was booked at two man-days, three hundred and ninety-eight dollars. The problem it found was worth ten thousand eight hundred dollars in chairs alone, and the soft opening it put at risk was worth considerably more.

Why nothing in the file said so

The buyer's chair specification said: seat height four hundred and sixty millimetres, overall height seven hundred and sixty millimetres. Those two numbers describe a chair completely, as far as anyone reading the document was concerned. They describe the seat and the back. They say nothing about the armrest, and the armrest at seven hundred millimetres is the only part of the chair that has to go under the worktop. The reference chair from the previous property had arms that sat lower, and nobody had ever needed the number.

The desk specification said: seven hundred and twenty millimetres high, twenty-five millimetre top. Those two numbers describe a desk completely, as far as anyone reading the document was concerned. Subtract one from the other and you get six hundred and ninety-five millimetres of clear space under the worktop, and that number appears nowhere, because it never came up.

So each document was complete on its own terms and the two were jointly silent. Each product was checked against its own line item, signed off against its own line item, packed and shipped. The fit between them lived in nobody's specification, which means the factory had no requirement to check it, the buyer's quality team had no requirement to check it, and the approved sample had no requirement to demonstrate it.

This is the failure mode that a per-product inspection is structurally unable to see. An inspection that verifies each item against its own line item passes this order completely. There is no line item for the relationship, and there is no defect without a line item.

How to measure the clearance under a worktop

If you take one measurement out of this case, take the clearance under the worktop, and take it from the unit standing on the floor rather than from the drawing. The method is simple, but four details decide whether the number is usable.

• Measure from the finished floor to the underside of the worktop at the front edge, in the position where a guest's knees go. Do not subtract a nominal top thickness from a nominal desk height. Stand a steel rule or a depth gauge under the actual unit and read it.
• Take the lowest reading across the width and from front to back, and read it in the knee position rather than in front of the drawer pedestal. A modesty panel, an apron, a cable tray or a pedestal reduces the usable space in one place and not in another.
• Measure the chair at its highest projecting member. On an armed chair that is the top of the armrest, not the top of the back. The back stays in the room; the arm has to go under the worktop edge.
• Leave at least fifteen millimetres of working clearance. A chair that clears by two millimetres is not a chair that fits, it is a chair that has not been tested.

A set is two products and one relationship

A desk and a chair in a guest room are bought as a set and made as two product lines, usually in two workshops, sometimes on different days, and certainly in different cartons on different pallets. The buyer approves a desk sample and a chair sample, normally at different meetings, against different reference documents, and neither meeting had both products in it.

A chair sample gets approved for what a chair sample can be approved for: the upholstery, the foam, the finish, the comfort, the colour. None of those attributes say anything about whether the chair will go under the desk. A chair that has been sat in for twenty minutes in a showroom has usually been sat in without a desk.

Here the answer was sitting in the same file and nobody was looking at it. The accessible rooms had the armless chair, and the armless chair went under the desk in all sixty of them. Both chair drawings were in the specification. What was missing was not information. It was the observation that the information was about the same desk.

The rule that follows is short enough to put in a purchase order. Any order containing a table and a chair is a fit problem as well as a quality problem, and it should be treated as one from the first drawing rather than the last pallet. The two products get checked against each other once, early, physically, and the result is written down as a number that both the factory and the buyer's inspector can measure without thinking about it.

What the report recorded, and what it could not

A finding is only as strong as the clause it is measured against, and this one had no clause. The report had to be written so that it stayed inside what it could prove.

Section two, reference documents, noted that the chair specification gave seat height and overall height and no armrest height, and that the desk specification gave overall height and top thickness and no clearance under the worktop. Section five, dimensions, recorded the measured worktop height, the measured top thickness, the derived clear height and the measured armrest height, with the arithmetic shown instead of a conclusion. Section six, physical check results, recorded the failed fit trial with photographs of the chair stopped against the worktop edge.

Section seven graded the finding and recommended holding the release on the one hundred and eighty affected sets. Section eight carried the caveat that mattered most: whether the arms were bolted on or permanently fixed could not be confirmed from thirty-two chairs, and if they were bolted then a remedy exists that nobody had thought of. The factory was asked to confirm it in writing before the buyer committed to anything.

Three options, and the arithmetic

The order was forty-nine thousand two hundred dollars. Two hundred and forty desks at a hundred and forty-five and two hundred and forty chairs at sixty. A thirty per cent deposit of fourteen thousand seven hundred and sixty had been paid, and the balance of thirty-four thousand four hundred and forty was due against the bill of lading. The affected chairs were worth ten thousand eight hundred.

There were three ways out, and each one costs a different thing.

How to write the specification so this cannot happen again

The fix for this case is six lines of specification, and not one of them is about workmanship or quality.

• State the clearance under the worktop as a minimum measured dimension: clear height from finished floor to the underside of the worktop at the front edge in the knee position, not less than 690 mm, measured on the finished unit standing on the floor. The top thickness is already inside that number.
• State the maximum armrest height as a number: top of armrest not more than 660 mm above the finished floor. A seat height and an overall height do not constrain an armrest.
• State the working clearance between the two products: the chair shall clear the desk underside by not less than 15 mm when pushed fully under.
• State the seat height as a range rather than a point: seat height 440 to 470 mm.
• Name the sets. List the desk and the chair together in one line of the specification so that a change to either one triggers a re-check of the pair.
• Add a fit clause: each desk and chair pair shall be physically trialed together before shipment and the trial recorded in the inspection report.

The part that is easy to leave out

A guest room set is a small thing to get wrong and an expensive thing to get wrong. The desk was not broken. The chair was not broken. The chair simply did not go under the desk, and nobody had written down the number that would have told them.

A guest room furniture inspection that checks each product against its own drawing passes this order, and it passes it every time, because there is nothing in the order to fail. The only check that finds this is one that puts the two products together, and the only way to guarantee that check happens is to write it into the contract as a dimension and as a trial.

So the reporting convention that follows from that is the one we use on every coordinated set: state the relationship as a number, measure it on the finished unit standing on the floor, and record the measurement as a measurement rather than as an opinion. If the arms come off, the buyer finds out before the container is sealed. If they do not, the buyer finds out before the bill of lading is signed. Either way the finding arrives while it is still a production decision.