Six conference tables stood in a row in a meeting room and one of them sat 4 mm low at the joint, a step you could see from across the room. All 240 tables in the order were inside the height tolerance. The order specified a table, and what the buyer needed was a set.
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 criterion that was missing
An office fit-out: 240 conference tables for a corporate training campus. Each meeting room takes six tables butted end to end into one continuous run, 10,800 mm long, with a 25 mm melamine top on an MDF panel, a powder-coated steel frame, and four adjustable glides under each frame. Forty rooms, forty sets of six. The order said so as well: 240 tables in 40 sets of 6, one set per room.
What the specification gave the factory was a table. Height 740 mm, tolerance plus or minus 2 mm, measured from the floor. Top flatness no more than 0.5 mm over 1,800 mm. Melamine and colour matched to the approved sample. Frame square, no rocking, four glides fitted. Every one of those lines describes a table, and every one of them was met.
What the specification did not give the factory was a set. There was no line about the joint between two tables, no line saying the run had to be level, no line saying which table in a run was the reference, and no line saying the six tables of a room had to be chosen, matched and labelled as a set before they were packed. Those four omissions are the whole case.
This is an illustrative example built from the patterns behind furniture export disputes: the goods are representative of the category, the figures are representative of the trade, and no particular buyer, factory or campus is described. It shows what a check finds when the specification describes a unit and the delivery is a group.
Where to go next
→ office and commercial table verification
→ the three levels of check we sell, and what each one covers
What height from the floor actually measures
The factory stood each table on the workshop floor, brought all four glides into contact, and read the height with a digital height gauge at the centre of each long edge. The tallest of the 240 was 741.8 mm and the shortest was 738.2 mm, so every one of them sat inside plus or minus 2 mm of 740. They were flat to 0.4 mm over 1,800 mm, square, and none of them rocked.
Here is the trouble with that criterion, and it is arithmetic rather than carelessness. Height from the floor references every table to a different piece of ground. Table 1 is measured to the floor under table 1, table 6 to the floor under table 6. When six tables are joined, their tops have to be level with each other, and the height of a top over a run is not the sum of two heights above a floor. It is the difference between two heights, and each of those heights has already been measured against its own patch of ground.
So the number the factory measured was a real number and the wrong one. It established that no table was too tall or too short on its own. It could never establish whether the six tables would read as one surface, because that question does not exist until the tables are standing on the same ground at the same time, and no line in the order ever asked it.
Where to go next
→ what an order verification report records and what it leaves out
Three readings of the same six tables
Room 4 is where the complaint came from. The fit-out project manager had a 1.2 m aluminium straightedge in his hand and could not seat it across the joint between table 5 and table 6. The inspector took one datum and stuck to it: the highest point of the floor under the run. The floor under each table and the top of each table were both read from that same point, so the two columns can be subtracted from each other and the third measurement, the straightedge across every joint, can be checked against them.
The tops of the six tables, and the floor under them, came out like this. Every table stood between 739.9 and 740.3 mm above the floor directly beneath it, which is why every one of them passed the order as written.
| Table | Floor under it, from the floor datum (mm) | Top of the table, from the same datum (mm) | Step at the joint to the next table (mm) |
|---|---|---|---|
| Table 1 | 0.0 | 740.1 | 0.4 |
| Table 2 | -0.3 | 739.7 | 0.4 |
| Table 3 | -0.2 | 740.1 | 0.6 |
| Table 4 | -0.4 | 739.5 | 0.2 |
| Table 5 | -0.5 | 739.7 | 3.6 |
| Table 6 | -4.0 | 736.1 | not applicable |
Read the second column against the third. Table 6 stands on floor 4.0 mm below the highest point of the floor under the run, and its top is 4.0 mm below the top of table 1. Those are the same number. The table did not sink. The ground under it was low. The tables did not drift; the ground did. And because every table had been levelled to the floor beneath it, no measurement taken against the floor could ever have shown this.
Where to go next
→ how we choose a sample size and read the acceptance numbers
The same six tables, one setting changed
Before writing anything down, the inspector did one more thing. On site, with the client's own installer, the glides under table 6 were wound up to bring it to the plane of the others, using a laser across the 10,800 mm and the manufacturer's own 20 mm of glide adjustment. The step at the joint between tables 5 and 6 went from 3.6 mm to 0.4 mm, and the largest step anywhere in the run became 0.6 mm, which was the spread that was already there among tables 1 to 5. The floor did not move. The tables did not move. Nothing was replaced.
That is the proof, and it is worth stating plainly: the goods were never the problem. Levelling each table to the ground under it produces six correct tables and one broken run. Levelling the six against each other, with the tops as the reference and the floor left alone, produces a set that reads as one surface. It is worth noticing that a laser plane is a site operation, so a set has to be levelled where it will stand, which is why the floor condition belongs in the specification rather than in the installer.
Notice what the levelling did to the height in the order. Before the adjustment, table 6 stood 740.1 mm above the floor under it, and 4.0 mm below the top of table 1. After the adjustment, its top is on the same plane as table 1, and it stands 744.1 mm above the floor under it, which is 4.1 mm outside the plus or minus 2 mm the order allowed. The two requirements in the order were mutually exclusive on that floor, and the installer could not have satisfied both. He could not have known that either, because the order never asked for the run to be level, so it never told him which requirement to break.
Where to go next
→ how our inspection actually runs, on site and in the workshop
Why a sample of tables could never have found it
The order carried a standard AQL clause: AQL 2.5 at general inspection level II. That is the usual way to say how many units to look at, and here is what it does to a set of tables. A lot of 240 tables is ISO code letter G, so 32 tables are drawn, and 2 defective in the sample is acceptable while 3 rejects the lot. A lot of 40 sets is code letter D, so 8 sets are drawn, and at the same AQL 2.5 nothing is acceptable while 1 rejects.
Same AQL, same inspection level, same 2.5 limit. The only thing that changed is the unit being sampled, and that single change is the difference between a check that cannot see the failure and a check that rejects the goods at the factory. A sample of 32 tables can pass while eleven of forty rooms fail, because the defect is not in the tables and the sets do not exist to be sampled. You cannot sample what you have not assembled.
| What is being sampled | Lot size | ISO code letter | Sample at level II | Accept or reject at AQL 2.5 |
|---|---|---|---|---|
| Tables, as the order was written | 240 | G | 32 | Accept 2, reject 3 |
| Sets, as the buyer needed | 40 | D | 8 | Accept 0, reject 1 |
Where to go next
→ the AQL calculator, with the code letters and sample sizes
The sets were never made
So the inspector went back to the factory and asked for the set records. There were none. Not a set matching sheet, not a run sheet, not a set number on a carton. The goods went out as 60 cartons of 4 tables each, every carton marked with the model, the quantity and the order number, and nothing anywhere that said which six of the 240 tables would end up in the same room.
The quality file held 240 lines, one per table: serial number, height from the floor, flatness over 1,800 mm, colour against the sample, frame squareness, glides fitted, checker initials. There was no field for a set, because there was no set.
- 240 lines of per-table results, and no field anywhere for a set.
- No bench matching, so nothing recorded six tables being stood together and levelled against each other before packing.
- No set number on the cartons, the packing list or the assembly instructions, so the installer had no way of knowing which six tables belonged in which room.
- No levelling method in the carton, so every table was levelled to the floor beneath it, which is what the order's own wording told the installer to do.
- No site acceptance check, so nothing was measured across a joint at any point in the chain.
- One order, 40 rooms, and an assembly sequence that had to be worked out on site by whoever happened to be at the run first.
Where to go next
→ commercial table inspection for contract and office orders
Which datum, and it matters more than it sounds
Choosing a datum sounds like pedantry until a run is 10,800 mm long and one joint is 3.6 mm out. There are five references a specification could name, and only two of them do any work for a joined run.
| Possible datum | What it actually measures | Any use for a joined run? |
|---|---|---|
| The floor under each table | Each table's height in isolation, each against its own patch of ground | No. It references every table to a different point, so it cannot show a set error |
| The floor at one named point | Height above a single spot | Only over a short run on a floor that is flat |
| The top of the first table in the run | Height relative to one member of the set | Workable but arbitrary, and levelling each table to its neighbour lets the error walk down the run |
| A laser plane set across the run | Height relative to one fixed plane | Yes. This is the tool the installer should be working to |
| A straightedge and feeler gauges across the joint | The step at one joint | Yes, and it needs no datum at all, because a step is a difference |
The last line is the one buyers tend to miss. A joint tolerance is a difference between two tops and needs no reference point. A run tolerance is a height and does need one. If a specification asks for the run to be level, it has to say level to what, and that question has only one good answer: the plane the run is set on, established once, in writing, before the first table is placed.
Where to go next
What the report said, and what it did not say
The report was careful about what it could and could not support. It did not say the goods were non-conforming, because they were not: 240 of 240 tables met every line of the order as written, and the inspector held the readings to prove it. It did not say the floor was defective, because a floor is not the supplier's product and certifying a building is outside the scope of any furniture inspection.
It said three things. That the specification did not contain the criterion the buyer was buying, and that no inspection at any price could have found a defect in a criterion that was never written. That the goods were delivered as 240 tables while the order called for 40 sets, and a set cannot be verified when it has never been assembled. And that eleven rooms would fail, which was the inspector's own laser survey of 40 floors and which the buyer can reproduce in an afternoon with the same laser.
It also recorded the re-levelling of room 4 as a site adjustment rather than a change to the goods, and recorded what it revealed: after levelling, table 6 stands 744.1 mm above the floor beneath it, outside the plus or minus 2 mm the order allows. The order as written cannot be satisfied in that room, because it asked for two things the floor made mutually exclusive. Nobody broke the order. The order broke itself.
- Re-match and re-label every set at the factory: stand each set of six on a level reference surface, level the six to each other, mark the set number on all six cartons and record the joint readings. Two man-days of factory time, and the sets leave the building as sets.
- Give the installer a levelling method: set one laser across the run, level the whole run to it, and do not level table to table or table to floor.
- Survey the floors before installation, and correct the 11 rooms with a step over 1.5 mm first. This one is the buyer's own precondition, and it is the line nobody wrote down.
- Replace the per-table height check with a set-level acceptance check: one run per room, all five joints, read with a straightedge and feeler gauges and recorded before the room is signed off.
- Hold the balance until the runs are level. On this order that is $21,504, and the whole argument costs a few hours of a laser and an installer.
Where to go next
What the verification was worth
The order was 240 tables at $128, so $30,720. Thirty per cent was paid on order and seventy per cent was due against delivery, which is $21,504. Re-levelling the eleven affected rooms, once the client had found them, was 11 rooms at 1.5 hours for two installers, so 33 man-hours at $40 an hour and $1,320, plus a fortnight of argument.
The set-level verification at the factory, covering the re-match of all 40 sets, the per-set joint records, the survey of the 40 floors and the report, was three man-days at Full Verification, $897. Against the $21,504 still sitting with the buyer, that is the cheapest line in the file.
And the numbers behind the decision are simple. Eleven of forty runs, 27.5 per cent of the rooms, had a step in the floor over 1.5 mm somewhere along 10,800 mm. In four of them the step was over 3 mm and the run was visible from the doorway. Across the whole order there were 200 internal joints, and the specification permitted a step of up to 4.0 mm at every one of them. The tables never moved and the order never said they should have.
Where to go next
→ verification pricing, from $149 per man-day
→ talk to us about a set-level check before you pay the balance
Seven lines that would have caught this
Here is the specification this order needed and did not have. Nothing in it is exotic. Every competent factory would have read it and asked which of the two lines it was meant to obey, and that is the point: these lines are not advanced. They are simply absent.
- Delivery as sets. 240 tables shall be supplied as 40 sets of 6. Each set shall be chosen, levelled against a common reference and identified by a set number. The set number shall appear on every carton, on the packing list and in the assembly instructions.
- Set level, factory-measured. The six top surfaces of a set, standing on a level surface, shall lie within 1.5 mm of a single plane. The joint readings shall be recorded and signed for each set.
- Joint tolerance. Adjacent tops within a set shall not differ by more than 0.5 mm at any joint, measured with a straightedge and feeler gauges.
- Datum. Where a run height is specified, the datum is the plane set across the run before the first table is placed. Height from the floor is not a datum for a joined run and shall not be used as one.
- Site levelling method. Sets shall be levelled as a run against a laser or a long straightedge. Tables shall not be levelled individually to the floor beneath them, and the height stated in this order does not authorise that.
- Floor precondition. The finished floor shall be flat to within 1.5 mm over any 3 m under a run. The buyer shall survey and confirm this before installation and correct it where it fails. This obligation sits with the buyer and has to be written down like any other.
- Acceptance. A run shall be accepted on the five joint readings, not on the height of the individual tables, and the buyer's inspector may witness the setting of any run.
Three questions to ask before the next order
Three questions, and all three can be answered on a phone call to the factory before anybody signs anything.
- Are the tables going to leave as sets, or as individual tables? If as individual tables, who decides which six go into a room, and how will they be matched to each other?
- What does the specification say about the joint between two tables? If the answer is height plus or minus 2 mm from the floor, then the run has not been specified, whatever the order says about sets.
- How are the sets levelled on site, and against what? If the answer is that each table is levelled to the floor, you will get correct tables and a broken run, and it will not be the factory's fault.
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.