The centre of the mattress was unsupported, and every part of the frame was to spec.

The situation

A European homeware retailer ordered 400 slatted bed bases from a factory in Foshan for its own-brand bedding range. The base was a straightforward piece: a 140 × 200 cm platform with a rubberwood perimeter, ten solid pine slats, four corner legs, shipped flat-packed in a printed carton with an assembly leaflet. 60% of the value was still unpaid at shipment.

The purchase order was short and precise about parts. Slats solid pine, not more than 70 mm clear gap between them. Rails rubberwood. Legs, rails and slats as per the factory's standard model for this size. That last sentence did the damage, and neither party knew it at the time.

The buyer's product team had specified the base for a 200 mm foam and latex mattress, and the mattress maker's warranty terms require a foundation whose slat gap does not exceed 50 mm and which carries a centre support on any base wider than a single. Neither requirement was written into the order, because both of them were, in the buyer's view, part of the mattress rather than part of the bed.

The first 240 bases were delivered to the retail park over three months and sold. The retailer's customer service desk then logged 41 complaints, all of the same shape: the bed wobbles, and the middle of the mattress has dipped. Two mattress warranty claims had already been declined by the mattress maker, on the stated ground that the foundation did not meet the published requirements. The buyer stopped the balance on the remaining 160 and sent an inspector to the factory.

The order was complete about parts and silent about structure

The factory had built what the order asked for. It is worth being precise about that, because the case only teaches something if the manufacturing is not at fault.

Each of the ten slats was solid pine, measured 70 mm wide and 18 mm thick, straight and free of knots at the fixing points. The clear gap between them measured 68 to 69 mm, inside the 70 mm the order allowed and inside the 76 mm that most mattress brands treat as the outer limit for any base. The rails were rubberwood at 30 × 90 mm, the corner legs were square section and fitted with the cam hardware they were designed for, the four leg positions were drilled and plugged as drawn, and every frame arrived with the correct count of slats, rails, legs and fixings.

What the order did not contain was any statement about the span each slat had to bridge, any requirement for a centre rail, any requirement for legs under the middle of the bed, and any statement of how far the deck was allowed to move under load. Legs, rails and slats as per the factory's standard model is a pointer to a shop drawing the buyer never saw, and that drawing contained four legs, because the model had been developed for spring mattresses that bring their own perimeter frame.

The factory's own final check was thorough. Each frame was checked on the assembly line for squareness, for leg plumb, for the count of components and for the slat gap. It was checked on a bench that was flat, unloaded, and seen from above. There is no point on that bench at which a frame with 1,400 mm of unsupported slat span looks different from a frame with a centre rail. The defect was a property of the loaded structure, and the factory had no test that could see it.

Two different numbers: the gap between the slats, and the span between the supports

The check most people run on a bed frame is the wobble test: put a hand on the rail, push, see whether it rocks. It is the wrong test for this fault, and it is worth saying why. A wobble test tells you about the frame as a body. This fault is in the deck as a loaded structure, and the two things a buyer has to verify are separate numbers.

The first number is the clear gap between adjacent slats, measured edge to edge with the slats seated in their holders, not the pitch from centre to centre and not the width of the slat. The second number is the unsupported span between adjacent supports: the clear distance from the inside face of one side rail to the inside face of the other, or to the centre rail, measured at the height of the slats. A base can have a perfectly acceptable gap and an indefensible span, and that is exactly the configuration here: 68 to 69 mm of gap across a 1,400 mm span.

68 mm is not an absurd gap. It is inside the order, inside the general 76 mm rule, and inside what an innerspring mattress would tolerate. It is outside the 50 mm the foam mattress in this order requires, and that is a separate finding with separate consequences, which the report kept separate.

Where the deflection was measured, and why the empty frame looked flat

Deflection was measured on the bare deck, mattress removed, because that is the only way to make the reading repeatable. The frame was stood on a level floor and levelled on its four legs, and the dial gauge was zeroed on the side rails at mid-length so that the reading belonged to the slats and not to the frame underneath them.

The load was 1,200 N, applied through a rigid 1,400 × 600 mm spreader panel so that it landed on the deck as a distributed load rather than as a point on one slat. Readings were taken with a dial gauge reading to 0.01 mm, at three positions along the length of the deck: the mid-length, where the complaints came from; the quarter point; and the foot end. The frame was loaded, held for 10 minutes, read, unloaded, and read again.

The worst position was the mid-length of the deck, at 22 mm of deflection at mid-span, with 2 mm of residual set after the load came off. The quarter point read 14 mm and the foot end 9 mm. On a reference frame of the same design with a centre rail and two centre legs, built from the same timber and the same slats on the same day, the same load at the same position gave 3 mm and no measurable residual. The side rails were read separately and sagged 4 to 5 mm at mid-length over their own 2,000 mm unsupported length, which is a smaller version of the same problem.

There is a second reason this fault survived to the retail park, and it changes what a check on the factory floor has to be. The empty frame was flat. With a 1,200 mm straightedge laid across the deck, the gap under the straightedge did not exceed 1 mm, because a straightedge bridges the slats instead of loading them. Every unit on the line passed that check, and every unit would pass it again. A deck that is flat when you look at it and 22 mm down when a body is on it is not a contradiction. It is what an unsupported span does.

Why the whole run was affected, and what that changes about the remedy

This was not a count of failures. Every frame in the run came off the same jigs with the same bill of materials and the same four legs, so the 22 mm was a property of the setup rather than a defect on some units and not others. The twenty frames sampled from the remaining run of 160 all measured between 21 and 23 mm at the mid-length position. That is a lot-level finding, and it changes the practical answer: the remedy is not to sort, it is to rework the whole remaining run, or to rework none of it and argue about the 240 already shipped.

Rework was not free and not quick. Adding a centre rail and two centre legs to a flat-pack bed is not a five-minute job at the end of the line, because the side rails have to be opened, drilled and re-plugged, and the carton, the assembly leaflet and the hardware count all change. Two weeks of the delivery programme were lost to it.

The order of work matters as much as the cost. Had the inspection happened before the first container, the fix would have been a drawing change and nothing else. The same finding costs a drawing change at the factory and a commercial problem with 240 bases already in a warehouse in Europe.

What the inspection established, and what it did not

What the inspection established: that the slats, the gap, the rails, the legs and the fixings conformed to the order as written; that there was no centre rail and no centre leg on any frame sampled; that the deck measured 22 mm at mid-span under 1,200 N at the mid-length position, 21 to 23 mm across the sample, against 3 mm on a reference frame built with a centre rail; that 2 mm of residual set remained after unloading; and that the empty deck was flat to within 1 mm under a 1,200 mm straightedge.

What it did not establish: why the centre support was missing. The factory's assembly instructions for this model list four legs, and the shop drawing the order referred to lists four legs, so the factory had built what its own documentation said. The report recorded that the drawing referenced in the order was not in the order file and was not provided on request until after the inspection, and it did not conclude that the factory had substituted a cheaper frame, because the frame was the frame the factory designed.

What it could not establish: the behaviour of the 240 bases already shipped, and the 41 consumer complaints behind them. The report's scope was the factory and the remaining run. Whether the mattress maker's warranty position on those 240 bases could be reopened is a commercial matter between two parties, and a report should say that rather than answer it.

One point of method is worth stating, because it is the one most reports get wrong. The 22 mm was an elastic deflection under a defined load on the day of inspection. What the bases do over six months of use depends on creep in the pine, on the fixings working loose and on the mattress wearing into the shape of the dip, and none of those is a fixed number that an inspection can produce. The report gave the load, the positions and the readings, and did not forecast a service life.

What the specification should have said

Six lines. None of them costs anything to write, and each one moves a check from a part that can be verified in isolation to a result that somebody who was not there can repeat.

• **Name the unsupported span, not just the slat.** The clear span of each slat between adjacent supports shall not exceed 700 mm. Any internal width above 1,000 mm shall be fitted with a centre rail running the length of the frame, supported by not fewer than two centre legs in contact with the floor, giving not fewer than 6 floor contact points in total. Where the order does not specify the support arrangement, the buyer approves the shop drawing in writing before the first rail is cut.
• **State the slat gap as a number and a method, and state the mattress's own requirement beside it.** Clear gap between slats shall be 50 mm maximum, measured edge to edge with the slats seated in their holders, at not fewer than 5 positions along the length and on both sides. The frame shall satisfy the foundation requirement published in the warranty of the mattress supplied with it; where the two differ, the tighter figure applies.
• **State the slat section, because it interacts with the span.** Slats shall be solid pine, not less than 90 mm wide and not less than 20 mm thick, straight grained and free of knots at the fixing points, with the number of slats derived from the gap in the previous clause rather than fixed independently of it.
• **State the deflection criterion with the load, the method and the positions.** Deflection of the deck at mid-span under 1,200 N applied through a rigid 1,400 × 600 mm spreader panel shall not exceed 5 mm, measured with a dial gauge at the mid-length, the quarter point and the foot end, after the load has been held for 10 minutes. Residual deflection after unloading shall not exceed 1 mm. The measurement shall be made with the mattress removed, on a level floor, on the bare frame.
• **Make the loaded check a hold point before packing.** Every frame shall be loaded and read on the assembly line before it is packed, and the record travels with the order. The first 20 frames of the run shall be offered for inspection before the remaining frames are assembled, and no frame shall be packed against that record.
• **Say that an unloaded check is not a conformity result.** Conformity of the slats, the gap, the rails, the legs and the fixings does not constitute conformity of the deck under load. The deck is a separate acceptance characteristic, and a frame conforms only when both the parts and the loaded deck conform.