Every buyer has had this conversation. You approve a sample — good weight, clean joints, edges that feel right in the hand, colour that matches your brand. Eight weeks later the shipment arrives and something is off. Not wrong enough to reject outright. Off.
The lid doesn't seat as tightly. The colour is half a shade warmer. The engraving is shallower. A drawer sticks when the room is humid. The factory tells you this is normal. It is normal in the sense that it's common. It is not inevitable.
Nearly every one of these gaps has a specific mechanical cause. And nearly all of them can be closed on paper — before a single piece of wood is cut.
Seven gaps, in the order they usually bite
Fix it on paper: state which operations are CNC, which are manual, and require the production pilot to be run on the production line — not in the sample room. A "production sample" pulled from the actual first 50 pieces is worth more than any handmade prototype.
Fix it on paper: species, grade and moisture content at time of machining, with a verification method (pin meter, five readings per batch) and a requirement for the kiln or conditioning record. Also specify the target equilibrium range for the destination climate — a product shipped to a dry northern winter needs a tighter spec than one shipped to a humid market.
Fix it on paper: commercial name and botanical name, origin, grade, solid vs. finger-jointed vs. plywood core, and an explicit limit on defects — max knot size, sapwood allowed or not, acceptable colour variation. Then seal a physical reference panel and sign it. Nobody argues with a signed board.
Fix it on paper: put tolerances on the mating dimensions specifically — not a blanket ±0.5 mm over the whole drawing. A sliding fit on an 8 mm joint wants something in the region of +0.05/−0.00 on the slot and −0.05/−0.10 on the tenon, plus a stated requirement that fit-critical dimensions are checked at first article and then at a defined interval through the run (every 500 pieces is a workable starting point for a long order). Also ask the factory how they manage cutter life. A workshop that logs tool changes will tell you immediately; a workshop that doesn't will tell you it's never been a problem.
Fix it on paper: coating brand and product code, colour reference (Pantone or RAL) with an agreed tolerance — ΔE ≤ 1.5 is achievable on a solid colour, more realistic to allow ΔE ≤ 2.5 on stained open-grain wood — plus gloss level measured at 60°, dry film thickness, number of coats, and an adhesion test method (cross-hatch to ASTM D3359 is standard). Then seal two signed colour panels, one held by you and one by the factory.
Fix it on paper: for every non-wood component, specify material, surface finish, and a testable requirement — salt spray hours for metal parts, water-resistance class for adhesives (D3 for interior, D4 where there's any chance of wetting), pull-out or shear strength for joints, and coating compliance for anything a child can touch.
Fix it on paper: a process card — the full operation sequence with the parameters and dwell time for each step — attached to the order and referenced in the inspection protocol. Inspectors can then check a process record rather than guess at it. It also gives you something concrete to point at when the schedule does slip, which it will.
Symptom → cause → the line to put in your drawing
| What you see | Most likely cause | What to specify |
|---|---|---|
| Joints looser than sample; panels rock | Moisture content above spec at machining; tenon cut undersize | Moisture 8–10% verified at machining; fit tolerances on mating dimensions |
| Parts must be forced together; cracking at shoulders | Cutter wear producing interference; species substitution with different movement | Tolerance pair on slot and tenon; tool-change log; first-article dimension check |
| Colour differs from approved sample | New pigment batch; different spray operator; film thickness | Coating code, Pantone/RAL reference, ΔE tolerance, gloss at 60°, signed panels |
| Surface feels rougher | Sanding passes reduced; worn abrasive; skipped grit sequence | Grit sequence stated (e.g. 180 → 240 → 320); sealed feel reference |
| Cracks or open glue lines weeks after arrival | Drying or conditioning time cut; adhesive downgraded | Process card with dwell times; adhesive class D3/D4 named |
| Odour or mould inside sealed cartons | Product packed before full cure; moisture trapped | Minimum cure time before packing; carton moisture barrier; container desiccant |
| Engraving depth inconsistent | Laser power drift; uncalibrated fixture; different operator | Depth range in mm; contrast reference; fixture-based setup, not hand positioning |
The numbers behind those specification lines are their own subject — see how to read a Chinese wooden product quotation for where the material and finishing costs actually come from, and the failure case library for what happens when they get skipped.
Three documents that close most of this gap
You don't need a hundred-page quality manual. In practice, three artefacts prevent the large majority of the problems above — and all three are cheap.
A signed golden sample, held in two places
One approved sample signed by both sides, with the deviations noted in writing. Keep one copy with you and one at the factory, and photograph it under consistent lighting. The point isn't bureaucracy — it's that "matches the sample" becomes a checkable statement rather than an opinion.
A process card attached to the order
The operation sequence with parameters: moisture target, machining method, grit sequence, coating system, number of coats, drying time between coats, cure time before packing, assembly method, packing method. Most factories have this internally; they simply don't hand it over. Ask for it. A supplier who won't share a process card is telling you something.
First-article approval with measured dimensions
Before the run starts, the factory produces a small number of parts on the production line. Those parts get measured against the drawing — not looked at, measured — and approved or corrected. This single step catches most tooling, material and finish problems while they still cost a few hundred dollars instead of a container.
- Moisture content readings for the actual batch being machined
- Material certificate or invoice showing species and grade purchased
- First-article parts measured against the fit-critical dimensions on the drawing
- Coating batch number plus a sprayed panel compared to the sealed reference
- Process card with dwell times, signed off by the production supervisor
- Photographs of the line setup, not just of finished parts
None of this is unreasonable. Factories that work with serious export brands do all of it already — they just do it for clients who ask.
The uncomfortable part
Here's the thing most buyers don't want to hear: on many of these projects, the drift isn't caused by a bad factory. It's caused by a specification that never existed. A drawing with no tolerances, no material grade, no coating code and no process requirement is not a specification — it's a picture. And when the specification is a picture, the factory fills in every blank with whatever is cheapest and fastest that day.
That is not dishonesty. That is literally what you asked them to do.
The good news is that closing these gaps is mostly clerical work done early, not inspection work done late. It costs a few hours on paper. It saves a container.
Frequently asked questions
Common questions about why an approved sample and a production shipment drift apart.
Why does my wooden product sample look different from the production shipment?
Because a sample is made under favourable conditions — often by hand, by a skilled worker, with sharp tooling and no schedule pressure — while production runs on a line, at a rate, with a different operator and worn cutters. The seven most common causes are a different build method, moisture content above spec, material substitution, tool wear changing fit dimensions, a new coating batch, swapped hardware or adhesive, and a compressed process schedule. Almost all of them can be closed by specifying them in the drawing before production starts.
What moisture content should wooden products have for export?
The working range for export wooden products is typically 8–12%. For European and North American indoor markets the safer target is 8–10%, because wood keeps losing moisture after assembly and joints loosen as it does. Specify the moisture content at the time of machining rather than at delivery, and ask for kiln or conditioning records plus pin-meter readings for the actual batch.
How do I stop a factory from changing my design after sampling?
Not with inspection — only with specification. Put tolerances on fit-critical dimensions, name the species and grade, specify the coating product code with a colour reference and tolerance, define the grit sequence and number of coats, and attach a process card with dwell times to the order. When every blank in the drawing is filled in, there is nothing left for the factory to decide for you.
What tolerance should I put on a wooden joint?
A blanket ±0.5 mm across the whole drawing is not enough for mating parts. On an 8 mm sliding joint, workable values are roughly +0.05/−0.00 on the slot and −0.05/−0.10 on the tenon. The important point is that the two dimensions are specified as a pair: cutter wear makes external features larger and internal features smaller, so the two errors add up in the same direction.
What is a golden sample and why does it matter?
A golden sample is a production-built part approved by both sides, signed, and held in two places — one with you, one at the factory. It turns 'matches the sample' from an opinion into a checkable statement. Combined with a process card and a first-article dimension check, it prevents most of the drift between sample approval and the shipped container.
What causes wooden parts to crack weeks after arrival?
Most often moisture movement: the wood was above target moisture content when machined and went on shrinking after assembly. The second most common cause is a drying or conditioning schedule that was compressed because the shipment was late. Neither shows up in a final inspection two days after production, which is exactly why they have to be caught on paper rather than at the dock.
Can colour variation between batches be eliminated?
Not entirely — wood is a natural material and pigment lots vary — but it can be controlled to a level nobody notices. Specify the coating product code, a Pantone or RAL reference, an agreed ΔE tolerance (≤1.5 on a solid colour, ≤2.5 on stained open grain), gloss measured at 60°, and dry film thickness. Then seal two signed colour panels, one held by each side.
Send me the drawing you're about to quote
I'll tell you which of these seven gaps are open in it, and what to add before it goes to a factory. No charge, no obligation — and if I think it's already fine, I'll say that too.