The second container is where appearance grading goes wrong. The first one matched the sample, everyone was happy, and the purchase order said "rustic grade, natural colour variation". Six months later the reorder lands and the knots are bigger, the sapwood is heavier and three pallets read noticeably darker. Nothing was breached, because nothing was defined. Setting hardwood flooring grading appearance acceptance criteria with a China supplier means replacing adjectives with a defect schedule: numeric limits, a stated measurement method, and physical boards that both parties hold.
This is not paperwork for its own sake. Appearance is the only major flooring attribute with no natural unit. Thickness has millimetres, moisture has percent, species has a botanical name. "Character" has nothing until you give it something.
Why the adjective fails, and why exotic species have no rulebook
Buyers often assume a recognised grading standard is doing the work in the background. Usually it is not, for three reasons.
First, the standard most often quoted does not apply. NHLA rules grade lumber, not flooring, and they grade for the yield of clear cuttings obtainable from the poorer face of a board. That is a sawmill and furniture-trade yield standard. Writing "FAS grade" on a flooring purchase order asks the mill to apply a rule that was not built to describe a finished floor surface, and a reasonable mill will quietly substitute its own interpretation.
Second, the standards that are built for flooring are regional and product-specific. The European product standards for parquet and wood flooring are split by construction — solid parquet elements, lamparquet, multi-layer parquet, solid wide boards each have their own standard, with CE marking handled under a separate harmonised standard. US appearance grading rules from the flooring trade bodies are written mainly around domestic species. Chinese mills more often work internally to GB/T standards for solid and engineered wood flooring. All of these exist; none of them is automatically in your contract.
Third — and this is the one that decides the matter for a custom parquet order — the species are exotic. Burma teak, black walnut, rosewood, Indonesian rosewood (Sonokeling) and cedar do not all sit inside a domestic North American or European grading rule. Sonokeling in particular has a colour spread running from light purple-brown to near-black streaking within a single log, which is a range question rather than a defect question. Once you are outside the species a published rule covers, you are writing the rule yourself whether you intend to or not.
So the first question to the mill is simply: which standard do you grade to, and can you send me your written grade definitions? If the answer is an in-house grade — which is common and not a problem — then the definitions have to be written down and attached to the order. An in-house grade with a documented defect schedule is more useful than a national standard nobody applies.
The defect schedule: adjectives converted into numbers
Build one table and attach it to every purchase order for that product. The limits below are a template showing the shape of the clause and the units to use — the actual numbers come from your project and from what the mill can hold, which is why the right sequence is to draft the table, send it, and let the mill mark up what it can and cannot meet.
| Attribute | Unit | Tighter grade, example limit | Character grade, example limit | How it is measured |
|---|---|---|---|---|
| Sound (tight) knot | mm, largest face dimension | ≤ 10 mm, max 1 per board | ≤ 40 mm, max 3 per linear metre | Steel rule across widest axis, after finishing |
| Unsound or loose knot, knot hole | mm | Not permitted | ≤ 15 mm, filled, filler colour per retained board | Measured after filling |
| Sapwood on the wear face | % of face area, per board | ≤ 5% | ≤ 25%, and ≤ __% of boards per pack | Visual against retained boards |
| Colour position | — | Within retained range set | Within retained range set (wider) | Three-board range set, defined light |
| Slope of grain | ratio over 300 mm | ≤ 1:10 | ≤ 1:6 | Straightedge and rule |
| Surface check / shake | mm | Not permitted | Length ≤ 50 mm, width ≤ 0.3 mm, surface only | Rule and feeler gauge |
| Pin and worm holes | mm and count | Not permitted | ≤ 2 mm dia., ≤ 3 per board, filled | Rule |
| Mineral or colour streak | % of face length | ≤ 10% | Permitted, no limit | Visual |
| Filler on the wear face | % of face area | ≤ 0.5% | ≤ 2%, colour matched | Visual |
| Machining defects (torn grain, chip-out, snipe) | mm | Not visible under stated viewing conditions | ≤ __ mm, not on the wear face | Stated viewing conditions |
| End split | mm from board end | Not permitted | ≤ 20 mm, hairline | Rule |
| Joint gap, dry-assembled | mm | ≤ 0.2 mm | ≤ 0.3 mm | Feeler gauge, 5-board dry lay |
| Overwood / lippage at joint | mm | ≤ 0.2 mm | ≤ 0.3 mm | Feeler gauge on straightedge |
Two structural points about this table matter more than any single number in it.
Grade after finishing, not before. A hand-rubbed wood wax oil penetrates the grain rather than filming over it, and penetration pulls tonal differences forward that were invisible on a sanded blank. Colour and streak that looked uniform in the white can separate visibly once oiled. If the acceptance grading happens on unfinished boards, you are grading a different product from the one that arrives. Write the clause so that grading occurs on finished boards, and so that any deliberate surface texture — hand-scraping, relief carving, brushing — is graded as texture, not as a defect. The texture options you can actually specify belong in the same document, described separately from the defect limits, or an inspector will log the intended scrape marks as machining faults.
Limits per board are not enough. Every board in a shipment can sit inside the schedule and the floor can still look wrong, because all the heavily figured boards landed in two packs. Add a distribution clause: no more than a stated percentage of boards in any single pack at the outer limit of the colour or defect range, and random mixing across packs at packing. This is the single cheapest clause in the document and the one most often missing.
Colour is settled with boards, not numbers
Colour is where numeric limits break down. Instruments can measure it, but a matte oiled surface reads differently by viewing angle and light source, and a ΔE figure agreed in a lab is a poor tool for an argument on a job site. The workable mechanism is a range set.
A proper range set is at least three boards: the light limit, a mid reference, and the dark limit. For a species with strong streaking, add a fourth showing the maximum acceptable streak density. Each board is signed and dated across the face and onto the edge, so a swapped board is visible. Produce three copies — one held by the factory, one by you, one issued to the inspection body — and record the set number on the purchase order.
Then handle the part almost everybody forgets: retained boards keep changing. Oiled teak darkens appreciably under light in its first months. Black walnut tends the other way, mellowing and lightening with exposure. A range set cut a year ago and stored in a bright office is no longer the reference it was on the day it was signed. Store the set in the dark, date it, state in the contract that comparison is made against a set of comparable age or with an agreed allowance for ageing, and re-issue the set on a defined interval for a repeat programme.
Specify the viewing conditions for the comparison too: daylight or a D65-equivalent source, diffuse rather than raking, boards laid flat and viewed from standing height. Colour disputes are frequently disputes about the lamp.
This is precisely the work the sample stage is for, and it is worth doing properly before production rather than after — the sample approval process is where a range set gets built, and the sample boards worth requesting for a custom floor are chosen to bracket a range rather than to show one pretty face.
The clause that ends arguments: viewing conditions and sampling
Two mills can apply the same defect schedule and reach different conclusions because they inspected differently. Fix the inspection, not just the limits.
Viewing conditions. State that a board is assessed from a standing position at roughly 1.5–1.8 metres above the floor, under diffuse illumination of roughly 500 lux, viewed at approximately 90° to the surface. Defects visible only under low-angle raking light or at close range are excluded from the count. This mirrors how a floor is actually seen and it removes the inspector-on-hands-and-knees problem, which is responsible for a large share of wood flooring claims.
Sampling and acceptance. Appearance is a variable attribute across many boards, so use a sampling plan rather than 100% inspection. State the plan explicitly: ISO 2859-1, general inspection level II, single sampling, with named AQLs for major and minor appearance defects — 2.5 major and 4.0 minor is a common starting pair for finished wood products — applied to boards, not to cartons. Say which defects count as major and which as minor; without that mapping the AQL is decorative.
The below-grade allowance. Wood flooring trade practice allows a small percentage of a mill run to fall below the nominal grade, commonly cited at around 5%. Whatever number applies, decide yours deliberately and write it, rather than inheriting the mill's default silently. State whether it is measured per pack or per lot — per pack is much tighter, and much more expensive.
Where inspection happens. Grading at the mill before packing is worth more than grading at destination, because the remedy is still cheap. Fold the appearance schedule into the wider pre-shipment quality control scope rather than treating it as a separate visit, and use the same checkpoint discipline that a 12-point furniture inspection applies to construction and finish.
Dimensional tolerances belong in the same document
An appearance grade with no dimensional table produces floors that look right and lay badly. Add these lines, and add the condition that makes them meaningful:
| Dimension | Suggested unit | Note |
|---|---|---|
| Thickness | ± mm, within pack and across lot | State both; within-pack is what the installer feels |
| Width | ± mm | Tight tolerance here controls joint gaps |
| Length | ± mm | Looser is usually acceptable |
| Squareness | mm out of square across width | Critical for herringbone and chevron |
| Straightness — bow, crook, twist | mm per metre | State the measured length |
| Assembled joint gap | mm, dry-laid 5 boards | The number the client sees |
| Tongue and groove fit | go / no-go or ± mm | Specify at the stated moisture content |
The condition: tolerances are only meaningful at a stated moisture content and temperature. Measure at, say, 20 ± 2 °C and at the contract moisture content, and say so in the clause — a width tolerance verified on boards two points wetter than the contract band is not the tolerance you bought. That link runs both ways, which is why the moisture content specification and the grading spec should be written as one document rather than two.
Common questions
Can I just approve a sample and skip all this? A sample settles one board. It does not settle the range around that board, which is what a shipment consists of. An approved sample plus a range set plus a defect schedule settles a shipment. Approving a single sample is how the second-container problem starts.
What if the factory grades to its own in-house standard? That is common and workable. Ask for the written definitions, convert them into the numeric table above, and have the mill mark up anything it cannot hold. A mill that pushes back on specific lines is giving you real information; one that agrees to every limit unchanged has probably not read them.
Does a wide-plank floor need tighter or looser limits? Looser per board, tighter on distribution. Wide planks and precious hardwoods mean fewer boards per square metre and more visible character on each one, so a single heavily marked board occupies a much larger share of the finished floor. Character limits usually need to widen; the pack-distribution clause needs to tighten.
How do I keep a reorder consistent two years later? Three things: the retained range set with its date and ageing clause, the lot number of the original timber batch recorded on the order, and attic stock bought at the time. Grain and colour matching across floor, stairs and trim works because the components come from the same timber lot — that logic applies to reorders as well, and the honest answer is that a reorder from a new lot is a new colour approval, not a repeat.
What to ask the supplier next
- Which grading standard do you work to, and can you send the written grade definitions for the grade you are quoting?
- Can you mark up this defect schedule with the limits you can actually hold for my species and pattern?
- Will you issue a signed, dated three-board colour range set in triplicate, and re-issue it on a stated interval?
- Is grading carried out on finished boards, after the oil, or on sanded blanks?
- What is your standard below-grade allowance, and is it per pack or per lot?
- Can you commit to a pack-level distribution rule for colour and character, with random mixing at packing?
- What is your dimensional tolerance table, and at what moisture content and temperature is it verified?
- Will you record the timber lot number on the packing list so a reorder or attic stock can be traced to it?
A concrete mill makes these questions easier to test. Anrantabu Flooring in Huiyang, Huizhou runs cutting, parquet assembly, hand-sanding, quality inspection and multi-coat hand-rubbed oil finishing inside one 20,000 m² plant with over 300 handcraft workers, and works in Burma teak, black walnut, rosewood, Sonokeling and cedar — species spread wide enough that the range-set question is unavoidable. Send the defect schedule as a marked-up table rather than as a paragraph, and read the mark-up carefully: what a mill declines to promise tells you more about the floor you will receive than what it agrees to.
