Who Should Write Your Japanese Product Copy: EC Copywriter, Translation Vendor or Machine Translation

Short answer: buy a translation vendor when the source text is factual and you need it rendered accurately; buy a Japanese e-commerce copywriter when the page has to rank and convert in a Japanese marketplace; use machine translation as a first pass and never as the last step on a published claim. Most foreign brands end up buying all three, tiered across the catalogue — and the expensive mistake is buying one quality level for every SKU.

The reason this is a procurement decision rather than a craft decision is that the three suppliers are accountable for different outcomes. A translation vendor is accountable for the Japanese matching your English. A copywriter is accountable for the page working. A machine is accountable for nothing at all, which becomes your problem the moment a regulated claim goes live under your store name.

The Three Suppliers, Defined Before They Are Compared

A translation vendor takes your English source and returns Japanese of an agreed quality level, usually with a second linguist checking the first. Scope is defined by your source text. Pricing is per source English word. The deliverable is a faithful rendering, and the vendor's professional obligation ends at fidelity.

A Japanese e-commerce copywriter writes Japanese product copy from your product facts, your images and your category — not necessarily from your English sentences. Scope is defined by the page and the platform. Pricing is per Japanese output, per page, or as part of a retainer. The deliverable is copy that fits Rakuten's or Amazon's fields, uses the vocabulary Japanese buyers in that category search with, and carries the reassurance cues Japanese product pages are expected to carry.

Machine translation, with or without human post-editing, converts your source at near-zero marginal cost. Raw output is free to nearly free; post-edited output — the industry calls it MTPE — is a human pass over machine drafting, and Japanese vendors commonly quote it at around 60% of their standard translation rate, with reports of unit prices well under half at the freelance end. Amazon's own guidance for sellers expanding internationally notes that its Build International Listings tool attempts to translate and create detail pages in a target store, while separately advising sellers to seek out qualified translators or agencies and warning that poorly translated listings risk not connecting with the customers you are trying to keep.

What Each One Actually Costs Per SKU

The single most useful thing to understand about this market is that translation and copywriting are priced on different sides of the same text. Translation is billed on the length of your English. Copywriting is billed on the length of the Japanese it produces. That means the two quotes are not directly comparable, and it means a long English source is expensive to translate even when the resulting Japanese page should be short.

Working from published Japanese reference rates, and taking a 400-word English product description as the unit:

Two honest caveats about that table. First, the per-SKU column is our own arithmetic on published reference rates, not a quote and not a benchmark — real quotes move with volume, format, turnaround and how much of your source is repeated boilerplate a translation memory can discount. Second, the copywriting row and the translation row are drawn from different tiers of the market: a ¥400 crowdsourced description is a marketplace-freelancer deliverable and an agency translation is a two-linguist process. Comparing the bottom of one band with the top of another is how brands talk themselves into a bad purchase.

What survives those caveats is the structural point: if your English source is verbose, translating it is often more expensive than commissioning Japanese copy of the length the page actually needs. Japanese e-commerce pages are not shorter than Western ones — the Rakuten smartphone description field alone accepts thousands of characters — but they are structured differently, and paying by the English word to produce a Japanese page you then restructure is paying twice.

Option 1 — A Translation Vendor

Japan has a large, professionalised translation industry with published rate guidance, and for the right deliverable it is the correct purchase and the cheapest route to reliable accuracy.

Buy it for: specification tables, materials and ingredients lists, sizing charts, care instructions, manuals, warranty text, safety notices, and anything where the value is in being exactly right rather than in being persuasive. These are also the places where a copywriter has no advantage — there is nothing to reinterpret in a dimensions table.

The limits are scope limits, not competence limits. A translation vendor did not quote for keyword research, and you should not expect it. The Japanese words buyers type are frequently not the dictionary equivalents of your English — categories run on their own vocabulary, often mixing katakana loanwords, kanji compounds and brand-specific shorthand, and choosing between them is a search decision, not a linguistic one. Nor did the vendor quote for fitting the output into a platform's fields; that constraint is covered below and it is where translated copy most visibly falls apart.

The contractual point that matters. A translation contract warrants that the Japanese renders the source. It does not warrant that the source was legal to say in Japan. If your English says a serum reverses signs of ageing, a competent vendor will render that faithfully — and faithfully is exactly the problem, as the compliance section explains.

Option 2 — A Japanese EC Copywriter

This is the supplier most foreign brands under-buy, usually because the quote looks like an expensive version of translation when it is a different service.

What you are actually buying is four decisions. Which Japanese search terms the page targets. What goes in the first thirty characters, because that is what a Japanese shopper sees before anything else. Which reassurance the page has to supply for an unfamiliar foreign brand — shipping origin, returns, who to contact, why this is safe to buy. And what to leave out, because the English source contains claims, comparisons or superlatives that either do not land in Japan or should not be made there at all.

Platform constraints make this a writing job, not an editing job. Rakuten's item name field runs to 127 full-width characters, but only the first thirty to fifty are reliably visible in listings; the catchcopy field is 87 full-width characters and is truncated far shorter in mobile category views. Those numbers turn copy into a compression problem with a ranking consequence — we worked through the field-by-field version of this in our Rakuten product page localization guide, which covers how the page is written. This article is about who writes it, and the two are deliberately kept apart.

Where to find one. Three realistic sources: a Japanese freelance writer through a domestic crowdsourcing platform, a specialist e-commerce content studio, or the writing capacity inside an operations agency already running your store. The first is cheapest and requires you to supply the category judgement and the quality control. The third is the only one where the writer can see the analytics, the review complaints and the search-term report that should be feeding the next revision — the difference between copy that is written once and copy that is maintained.

Building the capability internally is the same calculation we ran in agency versus in-house team, and at a single-market, low-SKU stage the answer is usually the same: you cannot fill a full-time role with copywriting alone.

Option 3 — Machine Translation, With and Without a Human Pass

Machine translation has genuinely changed this decision, and the right question is no longer whether the output is good — it is which parts of the job the machine is not doing.

What it does well. Volume, speed, cost, and consistency of terminology when it is fed a glossary. For internal comprehension, supplier documents, incoming enquiries and first drafts of long-tail listings, it is straightforwardly the correct tool, and the same logic we applied to enquiry triage in Japanese customer support outsourcing applies here: machine for reading, human for anything published.

What it structurally cannot do. It translates the text you gave it. It cannot decide that your English source was the wrong starting point, cannot pick the Japanese keyword with the search volume, cannot cut a 400-word paragraph into an 87-character catchcopy, and cannot tell you that a phrase which is unremarkable in English is a regulated claim in Japanese. Those are not quality gaps that improve with a better model — they are jobs outside the task.

Post-editing is the middle option, and it needs to be specified. "Light" post-editing fixes errors and leaves the machine's phrasing; "full" post-editing brings the text to human-translation quality. The price difference between the two is real and the quality difference is larger than most buyers expect, so a quote for post-editing that does not say which one you are getting is not a quote. Note also that the discount is measured against the translation rate, not against a copywriting rate — post-edited machine output is cheap translation, not cheap copywriting, and it inherits every scope limit in the section above.

Who Is Liable for What the Copy Says

This is the part of the decision that is almost never in the vendor comparison, and it is the part that can cost more than the entire localization budget.

Responsibility for a representation sits with the business making it. Under Japan's Act against Unjustifiable Premiums and Misleading Representations, cease-and-desist orders and surcharge orders are directed at the advertiser. The Cabinet Office guideline on the management measures businesses must take — issued under the Act and in force since 2014 — describes the arrangements a business is expected to have in place, including confirming the basis for the representations it makes. Outsourcing the typing does not outsource that duty.

The numbers attached to getting it wrong. The surcharge for misleading representations as to superiority or advantage is calculated at 3% of sales of the relevant goods, over a period capped at three years. The 2023 amendment, in force from 1 October 2024, added a commitment procedure allowing a business whose representations are under suspicion to apply for approval of a corrective plan and avoid an order, introduced a direct penal provision for misleading superiority and advantage representations, and created a 1.5× uplift on the surcharge for a business that has received one within the previous ten years.

The machine-translation-specific hazard. Restricted claim language is where faithful translation becomes a liability. Japan regulates what may be said about the effects of cosmetics, quasi-drugs, supplements and foods, and the boundary sits in different places than it does in most Western markets — we mapped the beauty-category version in our Japanese pharmaceutical and cosmetics advertising compliance guide. A machine will render your English efficacy claim into fluent Japanese without pausing, and a translation vendor will render it accurately because that is what accuracy means. Only a supplier briefed to flag claims — or your own review step — catches it.

One more rule that catches agencies and influencer copy. Since 1 October 2023, representations that consumers cannot identify as coming from the business have been designated unfair representations in their own right. If part of your Japanese copy programme involves writers producing content that reads as independent, that designation applies to you as the advertiser.

None of this is legal advice and we are not a law firm; the practical takeaway is procedural rather than legal. Whichever supplier writes the copy, name the person who approves it before it publishes, and make sure that person reads Japanese. A sign-off step costs minutes per SKU. It is the cheapest line in this entire budget.

A Decision Rule That Survives Contact With Your Catalogue

Buying one quality level for the whole catalogue is what makes this expensive. Tier it instead:

• Hero SKUs — purpose-written Japanese copy. The products that carry your category ranking, your advertising spend and the first impression of the brand. Written by a Japanese EC copywriter, to the platform's fields, with search terms chosen deliberately. This is typically ten to thirty SKUs, not the whole range.
• Mid-tier — machine translation with full human post-editing. Products that need to be correct, findable and not embarrassing, but will not justify bespoke copy. Specify full post-editing, and have the item name and catchcopy written rather than translated even here, because those two fields do the ranking work.
• Long tail — machine output plus a narrow human check. Restrict the human pass to regulated phrasing, product facts, units and specifications. Accept that the prose is functional.
• Never tier down: the compliance check in a regulated category, and the named sign-off. Both are cheap per SKU and neither scales with catalogue size the way copy cost does.

The arithmetic test for each tier is simple: if a SKU will not produce materially more gross margin in its first Japanese year than the cost of localizing it properly, it does not get the expensive tier. At the illustrative ¥7,200–11,200 per SKU for agency translation, a 500-SKU catalogue is a ¥3.6–5.6 million decision before a single unit sells — which is the moment most brands discover they were about to buy a quality level their category could not repay.

Two adjustments. In a regulated category — cosmetics, supplements, food with health-adjacent messaging, electricals with Japanese safety marking — move the whole catalogue up one tier. And on Shopify rather than a marketplace, the copy job expands beyond product pages to navigation, checkout, policy and trust pages: the scope covered in the Shopify Japan localization checklist and our Shopify Japan localization service.

Where Each Model Breaks in Practice

• Paying by the English word for a page you then rewrite. Brands send a 900-word Western product story to a translation agency, receive faithful Japanese, and then cut it to fit a marketplace template. The cutting was the actual work, and it was paid for twice.
• Translating the item name. The item name is a ranking decision with a character budget, not a label. It is the single field most reliably damaged by treating it as translation.
• Buying post-editing without specifying the level. Light and full post-editing are different products at different prices. Quotes that do not distinguish them tend to deliver light and be defended as compliant.
• No glossary, no terminology consistency. Three suppliers over two years produce three names for the same product attribute. Fixing it later means re-editing the whole catalogue.
• Nobody reads Japanese on your side. The failure mode is silent: the pages load, the traffic arrives, conversion is simply lower than it should be, and there is no error message pointing at the copy.
• Compliance treated as a language question. It is a claims question. A perfect translation of a claim you may not make in Japan is a perfectly executed problem.

Where LAUNOVA Fits, and Where It Does Not

We produce Japanese e-commerce copy as part of running the store, not as a standalone translation desk. That means item names, catchcopy, descriptions and campaign copy written in native Japanese by the same team watching the search-term reports and the reviews, so a page that is under-converting gets rewritten rather than re-translated. It comes as part of a full mandate through our Japan EC operations service, or scoped to a specific channel — the Shopify-side version is Shopify Japan localization, and marketplace launches usually start from a Rakuten setup review.

What we are not: a translation agency, a certified translation provider for legal or patent documents, a law firm, or a regulatory consultancy. If you need a manual, a contract or a technical specification translated, a JTF-member translation company is the right supplier and will do it better and cheaper than we would. If you need a lawyer's opinion on a specific claim in a regulated category, that goes to a Japanese lawyer or a specialist consultant, and we will say so. Our pricing is scoped to the work rather than published as a rate card — tell us your catalogue size, your category and your platform and we will tell you which tier each part of the range should actually be on, including where machine translation is genuinely enough.

The Short Version

Translation is priced on your English and accountable for fidelity. Copywriting is priced on the Japanese and accountable for the page working. Machine translation is nearly free and accountable for nothing, which is fine until it renders a claim you are legally responsible for. Buy translation for facts, copywriting for the SKUs that carry the business, machine output for the tail — and keep two things off the discount list: a compliance check in regulated categories, and one named person who reads Japanese and signs off before anything publishes.