{"id":1949,"date":"2026-10-24T22:49:39","date_gmt":"2026-10-25T01:49:39","guid":{"rendered":"http:\/\/topbambu.com\/bamboo-batch-consistency-tolerances\/"},"modified":"2026-10-24T22:49:39","modified_gmt":"2026-10-25T01:49:39","slug":"bamboo-batch-consistency-tolerances","status":"publish","type":"post","link":"https:\/\/topbambu.com\/pt\/bamboo-batch-consistency-tolerances\/","title":{"rendered":"Batch consistency for bamboo: writing tolerances that hold up"},"content":{"rendered":"
\n

The complaint that arrives three months into a bamboo programme is almost never “the product is bad”. It is “the second order does not match the first”. Colour is a shade darker, the board is a millimetre thinner, the cup weighs less, and the retailer’s shelf now holds two populations of the same SKU. None of this is a defect in the sense an AQL table measures, which is why it survives inspection and lands on the buyer. Batch consistency is a specification problem, and it is solved before the first purchase order rather than at the third.<\/p>\n

Why bamboo drifts more than plastic or metal<\/h2>\n

Bamboo is a natural material with a short growth cycle, and the variables that change its appearance are seasonal and agricultural rather than industrial. Culm age at harvest changes density and colour. Position along the culm changes the fibre. Carbonisation \u2014 the heat treatment used to darken material \u2014 changes colour with time and temperature, and a factory running the same recipe in a wet season and a dry season will produce different results. Moisture content at the time of machining changes final dimensions after the goods have travelled. None of these is a process failure; they are the material behaving like a material.<\/p>\n

The consequence is that a specification written as “natural colour, 18 mm thick” is not a specification. It is a description, and it will be satisfied by any two batches that a reasonable person would call natural and about 18 mm. The drift is inside the wording.<\/p>\n

Five characteristics, and how to write each one<\/h2>\n

The table below is the structure we use when a buyer asks for a consistency clause. It is deliberately about method as much as about limits, because a limit without a method is unenforceable.<\/p>\n\n\n\n\n\n\n\n
Characteristic<\/th>\nHow to measure<\/th>\nWhat the PO should say<\/th>\nHandling the natural variation<\/th>\n<\/tr>\n
Thickness<\/td>\nCaliper at defined points \u2014 for a board, four corners and the centre; state the points in the drawing<\/td>\nNominal plus a symmetric tolerance, with the measurement points and the instrument type named<\/td>\nSpecify whether the tolerance applies before or after conditioning; otherwise the argument is about the weather<\/td>\n<\/tr>\n
Length and width<\/td>\nTape or calliper on the finished, conditioned piece<\/td>\nNominal plus tolerance, and which faces are reference faces<\/td>\nAllow for shrinkage in the direction across the grain; bamboo moves less than solid timber but it moves<\/td>\n<\/tr>\n
Unit weight<\/td>\nScale on 10 pieces from the lot, recorded individually rather than as an average<\/td>\nA range, plus a maximum spread within the lot<\/td>\nWeight is the cheapest proxy for density and moisture; a wide spread inside one lot is an early warning<\/td>\n<\/tr>\n
Colour<\/td>\nVisual comparison against a retained golden sample under defined lighting, or a spectrophotometer if the buyer has one<\/td>\n“Matches retained sample” plus a stated lighting condition and a defined colour-difference limit<\/td>\nAgree whether the reference is the sample or the previous shipment, and keep both<\/td>\n<\/tr>\n
Surface and edges<\/td>\nVisual and tactile, with a written defect list<\/td>\nNamed defect categories rather than the word “good”<\/td>\nSeparate cosmetic variation from defects; a colour shift is not a defect unless the spec says so<\/td>\n<\/tr>\n<\/table>\n

Two of those rows do most of the work. Naming the measurement points turns thickness from an opinion into a reading. Defining the lighting condition turns colour from an argument into a comparison. Neither costs anything to write.<\/p>\n

Where AQL fits, and where it does not<\/h2>\n

Acceptable quality limits, as set out in ISO 2859-1, are the standard machinery for deciding how many pieces to inspect and how many defects to accept. They are well suited to countable attributes: a crack, a burr, a missing hole, a print that is off-register. They are poorly suited to gradual drift, because a lot can be uniformly a shade too dark and still pass every attribute check.<\/p>\n

That is the gap buyers fall into. They write an AQL plan, inspect 80 pieces, find two minor defects, accept the lot, and then discover the colour is wrong across all 5,000. Attribute sampling answers “how many are bad”; it does not answer “is this the same product as last time”. The second question needs a retained sample and a measurement, not a sample size.<\/p>\n

Three practical adjustments make AQL work on bamboo. First, define critical defects as a separate class with zero acceptance \u2014 a splinter that can cut, a loose handle fixing, an edge that fails a sharpness check. Second, keep cosmetic variation out of the defect classes and put it in the colour clause where it belongs. Third, record the measurement data from the inspection even when no defect is found, so that the next lot has something to be compared with. The numbers you did not write down are the ones you will need.<\/p>\n

The golden sample, done properly<\/h2>\n

A golden sample is only useful if it is a controlled object rather than a nice piece somebody kept. Four things make the difference. It should be signed and dated by both parties<\/b>, so there is no dispute about which sample it is. It should be from a production run<\/b>, not from a tooling trial, because trial pieces are made slowly and by hand. It should be retained in two places<\/b> \u2014 one with the factory, one with you \u2014 because a sample that exists only at the factory is a sample the factory can replace. And it should be photographed under a defined light<\/b> and the photographs stored with the PO, so that a warehouse in another country can check against something.<\/p>\n

Then decide what happens when the sample and the shipment disagree. The clause that works is not “must match the sample” but “must match within the stated tolerance, measured by the stated method; where it does not, the parties will compare against the retained sample at [named laboratory], and the lot will be reworked or credited”. That gives the disagreement a procedure instead of an argument.<\/p>\n

If the product is a board or tray where the grain and colour are the visible selling point, the serving board<\/a> range is a good place to see how construction affects the surface. For a broader view of how the inspection fits into the shipment, our guide to importing bamboo cutting boards into the US<\/a> sets out the sequence. Buyers who want to write the tolerance into a first order can send the drawing and the target market<\/a> and we will reply with the measurement points we would propose.<\/p>\n

Where this guidance stops.<\/b> We do not print AQL acceptance numbers, thickness tolerances or moisture figures in this article. The acceptance numbers depend on lot size, inspection level and the AQL you choose, and they must be read off the standard’s tables rather than copied from a blog. The dimensional and moisture limits are a commercial decision between you and your factory, and they depend on the product and the market. What we will say is that a number you chose deliberately, with a method attached, beats a tighter number somebody else chose for you.<\/div>\n

A specification checklist<\/h2>\n
    \n
  1. Drawing with measurement points marked, and the instrument named.<\/li>\n
  2. Nominal plus symmetric tolerance for each dimension, with the conditioning state stated.<\/li>\n
  3. Colour reference: golden sample, lighting condition, colour-difference limit.<\/li>\n
  4. Defect list split into critical, major and minor, with a photograph of each.<\/li>\n
  5. Sampling plan referencing the standard, not a remembered rule of thumb.<\/li>\n
  6. Retained sample procedure: two locations, signed, dated, photographed.<\/li>\n
  7. Remedy clause with a named laboratory and a cost allocation.<\/li>\n<\/ol>\n

    Seven lines. Most bamboo disputes that reach the point of an email argument would have been settled by any three of them.<\/p>\n

    Frequently asked<\/h2>\n
    \nWhy does the same SKU arrive in a different colour?<\/summary>\n

    Bamboo colour depends on culm age, harvest season and the carbonisation or steaming process. A factory running the same recipe across seasons will produce variation. The remedy is a retained sample, a defined lighting condition and a colour-difference limit, not a stronger word in the specification.<\/p>\n<\/details>\n

    \nIs AQL enough to control batch consistency?<\/summary>\n

    No. AQL sampling counts defective pieces; it does not detect a lot that is uniformly outside the specification. Use it for attributes and use measurement against a retained sample for drift.<\/p>\n<\/details>\n

    \nShould I specify moisture content?<\/summary>\n

    Yes, and specify the method and the point of measurement along with the value. Moisture at the time of packing determines how the piece behaves after a month in a container, so a specification without it leaves the outcome to the voyage.<\/p>\n<\/details>\n

    \nWho keeps the golden sample?<\/summary>\n

    Both parties, signed and dated, with photographs in the shipment file. A sample held only by the factory is not a control, because it can be replaced without your knowledge.<\/p>\n<\/details>\n<\/div>\n

    Sources<\/h2>\n
      \n
    1. ISO 2859-1 \u2014 Sampling procedures for inspection by attributes<\/a> \u2014 ISO<\/li>\n
    2. Instituto Nacional de Metrologia, Qualidade e Tecnologia<\/a> \u2014 INMETRO<\/li>\n
    3. Associa\u00e7\u00e3o Brasileira de Normas T\u00e9cnicas<\/a> \u2014 ABNT<\/li>\n<\/ol>\n

      Every figure above is traceable to these sources; where they disagree, both positions are stated. Values specific to our own products (carton size, 20GP loading, NCM, certificates) remain pending factory confirmation and are not presented as measured data. Regulatory summaries are a buyer’s working reference, not legal advice \u2014 confirm with your broker before you file.<\/p>\n