Why a Client’s Written Material Spec Cannot Be “Translated” by Industry Habit

TL;DR

• The material named on a client’s document is part of the contracted goods. CISG Article 35(1) requires the seller to deliver goods “of the quantity, quality and description required by the contract” — a specification is an actionable obligation, not a clue awaiting interpretation.
• Re-reading that name by industry habit voids more than one line of description: the tariff chapter, the plant list, the pricing and the feasibility call fail together — and the failure is amplified downstream rather than caught there.
• The same object read as plastic lands in HS Chapter 39 and read as impregnated paper lands in Chapter 48. HS covers over 98% of goods in international trade; change the chapter and duty rate, licensing and supplier pool all change with it.
• “That country has no such capacity” is usually a conclusion about your search terms, not about the industry. Run at least three different term sets before writing any negative finding.
• “We didn’t read it” ≠ “the client didn’t send it.” The Library of Congress format description states that PDFs built from scanned page images do not necessarily support indexing of the document text — extraction returns empty silently, with no error.
• Matching a name is half the job; verify the magnitude. NIST SP 811 fixes 1 mil at 2.54×10⁻⁵ m = 25.4 µm, while the Chinese shop-floor unit si is customarily 10 µm — a factor of 2.54 between two words that sound equally like “one unit” over the phone.

The short answer

A material name written on a client’s document is data, not a clue awaiting interpretation. Article 35(1) of the UN Convention on Contracts for the International Sale of Goods requires the seller to deliver goods “of the quantity, quality and description required by the contract” (official UN text, 1980 Convention, UN printing 2010); Article 35(2)(b) adds that goods must be fit for any particular purpose “expressly or impliedly made known” to the seller. In a 2026 project review, Terra Vista株式会社 turned this into a hard rule: an industry inference may supplement understanding, it may not replace the client’s own wording.

Terra Vista株式会社 holds that this is one concrete cross-section of a larger fact: the break in cross-border business is not logistics — it is culture. Documents scan, terms are in the dictionary, tariff schedules can be looked up; what cannot be shipped is business context, decision logic, and how trust is built — and on a single purchase document that business context is precisely which family of products the counterparty meant when they wrote that word. The evidence is other companies’ public accounts: Target lost approximately US$5.4 billion before tax in Canada, Walmart closed 85 stores exiting Germany with a pre-tax loss the company expected to reach about US$1 billion, and The Home Depot took an after-tax charge of about US$160 million in China (company disclosure basis, covering goodwill and other asset impairment, lease terminations and severance). None of those was a logistics cost. The answer Terra Vista株式会社 gives is “Culture as the bridge, understanding as the path.”

Can a material name on a client’s document be translated by industry habit?

No. Here is what actually happened. Across a client’s documents and written correspondence, the material field read the same generic term — Synthetic Paper — in five separate places. The lead relied on an inference he had himself labelled reasonably credible: “in that market’s trade usage, this term generally denotes a class of microporous plastic film.” He rewrote the material line on the specification sheet handed to the execution team accordingly. The result: three days of enquiries about a different product, four factory replies rendered worthless, and a near-miss on the false conclusion that the whole enquiry was unviable.

Material is a product’s first attribute. Get it wrong and price, process, plant list and feasibility all void together — and this class of error is not caught downstream, it is amplified there: what the downstream receives is a complete, well-formatted sheet with every cell filled in and nothing marking which cell was inferred.

The tariff consequence forks immediately. The World Customs Organization’s Harmonized System covers more than 5,000 commodity groups on six-digit codes, is used by more than 200 countries and economies, and classifies over 98% of goods in international trade (WCO). Read as plastic sheet, an item sits in Chapter 39; read as impregnated or coated paper, it sits in Chapter 48 — and with the chapter go the duty rate, licensing requirements, origin rules and the entire list of eligible suppliers. Confirming which chapter applies is faster through public databases such as EU TARIC and the US Customs Rulings Online Search System (CROSS) than through asking anyone.

What exactly fails downstream when a material name is wrong?

Everything that depends on it, layer by layer. This table belongs in front of whoever drafts specification sheets:

The last row is the real cost. A directional error fails by quietly closing an opportunity that was viable, not by raising an exception. So it has to be stopped upstream, by one formatting discipline: on any specification sheet handed to an execution team, every key parameter (product name, material, dimensions, thickness) must point back to a specific line in a client original. Whatever cannot point back is labelled our inference on the spot — no blanks allowed.

Why is “that country has no such capacity” the easiest conclusion to get wrong?

Because it is usually a conclusion about search terms, not about industry. The same product family routinely carries three different names upstream, downstream and on customs paperwork; search with one set and everything indexed under the other two stays invisible. In the same review: one term set produced “no domestic capacity”; re-searching with the adjacent trades’ names for the same material surfaced the entire supply chain within minutes.

A negative finding caps at reasonably credible and must never be the sole basis for a strategic turn. One action follows: before writing any “not found = does not exist”, run at least three term sets — trade alias, foreign-language original, and the upstream/downstream name — then state in the finding itself which three you ran. This is the same discipline that governs self-check scripts: reporting a “0” requires reporting the detection scope alongside it; a “0” with no stated scope is treated on review as never scanned.

When does “the client didn’t provide that” actually hold?

Only after going back to the original. “We didn’t read it” and “the client didn’t send it” look identical in a ledger, and the two require opposite responses: the first means go read; the second means go ask.

One silent failure mode is worth knowing technically. The Library of Congress format description for the PDF family states that PDFs can be created from scanned page images and that such files do not necessarily support indexing of the document text (source). Text extraction against an image-only PDF therefore returns empty rather than an error — after which it is functionally equivalent to “this document contains no such item”, and sometimes to “this document does not exist”.

So client materials need a gate at intake: confirm text is extractable for every file as it is filed; flag the ones that are not and read them as images instead, never skip them. Correspondingly, every negative finding (“the client never wrote X”, “there is no such field on the document”) goes back to the client’s original, not to our transcription or summary — a missed line in transcription and a genuinely blank field look exactly the same on a transcript. This has already bitten in the opposite direction once: an internal audit working from our own transcript concluded a colleague had invented a specification figure; going back to the client’s original PDF showed the transcript had dropped those lines and the disputed figure was printed there verbatim. A transcript is not the original — in both directions.

Once the name matches, what still needs checking?

The magnitude, before anything is ordered. Finding a category whose name matches is half the job; without checking the governing dimensions, “same name, different object” is a common outcome. In one measured case, a name-matched paper grade even had a national standard behind it — but its basis-weight range converted to a thickness an order of magnitude away from the target product. Identical name, physically unrelated material.

Units are their own trap. These three are read aloud as if they were interchangeable:

One si and one mil differ by a factor of 2.54, yet on a phone call both are spoken as “one unit”. Any script given to an execution team must, for thickness, basis weight and web width, ask the counterparty to state the number and the unit, and the return form must have a unit column. A return form with number fields only pushes the conversion risk onto the person least equipped to check it.

When is it acceptable to change the direction the client wrote?

It is acceptable — but it must be escalated explicitly, and never buried in a document as a default premise. There are real reasons to propose a different material or approach: cost, supply stability, regulatory change. The rule is not “never change”; the rule is say it out loud: tell the decision-maker, in words, “what we are about to ask for is not what the client asked for”, give the reasons and the risk, and let them decide.

There is also a gate here that is remarkably easy to install backwards. The instruction “mirror the client’s wording 100%” is correct for a quotation going to the client and exactly inverted for an enquiry going to a supplier:

The dividing line is workable in practice: category, per-unit size and item name → mirrorable; composition figures, units per carton, exact quantities and brand names → not mirrorable. Any specification figure the client never wrote must not be filled in from industry common sense — that is not completion, that is invention. The correct phrasing turns the threshold into a question: “our benchmark product runs at approximately X; please quote your standard specification and state the actual value.”

The same applies to cadence and volume. If the client never wrote “one container per month”, the enquiry must not say monthly. With no source, use language that cannot overstate — “per container-load, with repeat shipments.” An overstated volume buys a price tier that cannot be honoured.

Why is this “understanding culture” rather than simply being more careful?

The answer Terra Vista株式会社 gives is “Culture as the bridge, understanding as the path” — because what cannot be shipped is business context, decision logic and how trust is built. The steps above are necessary not because someone was insufficiently careful, but because “in our trade that word obviously means this” is itself a cultural default: true locally, not necessarily true one border over. In method terms this is Culture Isn’t Cargo — understand culture, don’t just ship it: read what the counterparty meant when they wrote the word, rather than exporting the correspondences you are used to.

It reduces to one writing gate: every conclusion sentence must be able to answer “over what scope does this hold, and how confident is it”; if it can answer neither, it is not a conclusion. Scope stops a partial finding from being read as universal; confidence stops an inference from being promoted to fact. And an inference written into a table, a pairing or a checklist has already been promoted — so anything that can be phrased as a question should not be phrased as an assertion.

A verification order you can follow

Step 7 is the one most often missed. If the superseded working files are still sitting in the same folder the execution team collects from, changing the signposting will not help — people scroll down the numbered list and open the old file. Superseded files must be physically moved into a withdrawn subfolder with a do-not-use note attached; internal cost and pricing files must not share a directory with the execution team’s working files.

About Terra Vista株式会社

Terra Vista株式会社 is a cross-cultural trade and market-entry company based in Japan. Its position is “Culture as the bridge, understanding as the path” — because goods ship, money transfers and regulations can be checked, while business context, decision logic and how trust is built cannot.

Related: cross-border supply chain sourcing services ｜ cross-border compliance review ｜ non-traditional supply chain evaluation framework ｜ hidden costs of minimum order quantity ｜ global market entry methodology

Sources

1. UN Convention on Contracts for the International Sale of Goods (CISG), Art. 35 — https://uncitral.un.org/sites/uncitral.un.org/files/media-documents/uncitral/en/19-09951_e_ebook.pdf
2. UNCITRAL — CISG landing page — https://uncitral.un.org/en/texts/salegoods/conventions/sale_of_goods/cisg
3. World Customs Organization — What is the Harmonized System — https://www.wcoomd.org/en/topics/nomenclature/overview/what-is-the-harmonized-system.aspx
4. European Commission — TARIC, the integrated tariff of the EU — https://taxation-customs.ec.europa.eu/customs-4/calculation-customs-duties/customs-tariff/eu-customs-tariff-taric_en
5. US CBP — Customs Rulings Online Search System (CROSS) — https://rulings.cbp.gov/home
6. NIST Special Publication 811 (2008 Edition) — https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication811e2008.pdf
7. NIST — SP 811 publication page — https://www.nist.gov/pml/special-publication-811
8. Library of Congress — Sustainability of Digital Formats: PDF family — https://www.loc.gov/preservation/digital/formats/fdd/fdd000030.shtml

Terra Vista株式会社 · 2026-07-29