A retail brand opening a new language market usually starts by translating the product pages, and then discovers that the product pages were the easy part. The navigation still reads in the source language’s logic. The size chart assumes one market’s conventions. The returns page, the delivery promise and the care instructions were written by three different teams and none of them was briefed. The campaign that drives traffic to all of it was transcreated; the shop it lands in was not.
E-commerce localisation is the work of making an online store read as though it had been built for the market it is selling into, rather than translated towards it. That is a wider brief than a word count suggests, and most of the cost and most of the risk sit outside the product descriptions.
This piece sets out what the work actually covers, which parts need transcreating and which need rendering exactly, and how to brief it so the store does not go live half-localised.
Key Takeaways
- An online store is not one content type. Product copy, interface strings, operational pages and campaign copy each need a different treatment, and briefing them as one job is the most common reason a localised store reads unevenly.
- Specification content must be translated exactly and marketing content must be transcreated. Mixing the two treatments is what produces either an unreadable product page or an inaccurate one.
- The economics of a catalogue are repeat-driven. A dedicated Translation Memory is what makes the second season, the second market and the restock cost less than the first.
What e-commerce localisation covers
The store is four distinct content types, and they do not share a treatment, a reviewer or a release cycle.
Briefed as a single pile of words, a store comes back with campaign lines rendered literally and dimensions rewritten creatively — precisely backwards. The first act of a sound brief is to separate the four.
| Content type | Examples | Treatment | Where it goes wrong |
|---|---|---|---|
| Product marketing copy | Product names, descriptions, collection and edit copy, lookbook and editorial text | Transcreate | Rendered word for word, so the sensory and aspirational effect disappears and the page reads as a parts list |
| Product specification data | Dimensions, materials, composition, weights, care instructions, compatibility, ingredient lists | Translate exactly | Rewritten for flow, so the figures or the material names drift; or left in the source language entirely |
| Interface and navigation | Category and filter names, sort labels, buttons, form fields, error messages, search prompts | Translate, then test in place | Translated as a spreadsheet of strings with no screen context, so labels no longer fit, no longer match the category they filter, or read as instructions rather than labels |
| Operational and trust content | Delivery, returns, exchange, sizing guidance, payment, warranty, contact, FAQ | Translate exactly, then localise the substance | Translated faithfully while still describing the source market’s delivery windows, return rights and payment methods |
| Campaign and CRM | Homepage and banner copy, launch lines, email and eDM, abandoned-cart and post-purchase messages, social | Transcreate | Produced on a different cycle by a different team, so it uses different words for the same products than the store does |
Two of those rows are the ones brands tend not to budget for. Interface strings and operational pages carry very few words and a disproportionate share of the damage, because a shopper meets them at the point of deciding whether to hand over money.
A shopper will forgive a product description they skimmed. They will not complete a checkout they do not trust.
Translate exactly, or transcreate: where the line falls
The single most useful decision in an e-commerce brief is drawing the line between content that must not change and content that must.
IPPWORLD defines transcreation as “creative translation”, a process that adapts the emotional expressions and contextual relevance of the source message that goes beyond literal meaning of words. That is the right treatment for a collection name or a product story. It is the wrong treatment for a fibre composition.
A workable rule, applied per field rather than per page:
- Anything a shopper could hold you to, translate exactly. Measurements, materials, composition, ingredients, care symbols, compatibility, warranty terms, delivery windows, return rights.
- Anything whose job is to make the product wanted, transcreate. Names, descriptions, collection copy, editorial, campaign lines.
- Anything a shopper clicks, translate and then see it in place. Interface strings cannot be judged in a spreadsheet; a label that is correct in isolation can be wrong above the thing it labels.
- Anything describing how you operate in that market, check the substance before the words. A perfectly translated returns page that describes the wrong jurisdiction’s rights is worse than no page.
IPPWORLD’s four elements — Language, Emotion, Culture and your brand tone of voice — describe what the transcreated half of that list is working on. The exactly-translated half is working on accuracy alone, and the two should not be reviewed by the same person.
The catalogue problem: volume, repetition and memory
A catalogue is the one content type where the second time costs less than the first, and a brand that does not set itself up for that pays full price every season.
Product content repeats heavily. The same fabric, the same fit description, the same care line, the same warranty paragraph recur across hundreds of items and return with every restock and every season. Handled as fresh copy each time, a brand pays again for words it has already paid for, and gets a different rendering of them.
IPPWORLD builds a dedicated Translation Memory for each client and reuses it across projects. For a retail catalogue that matters in three ways: repeat and updated content costs less over time, the same attribute reads the same way everywhere it appears, and a new market inherits the decisions already made rather than starting from the source language again. IPPWORLD’s own argument for value is cost-effectiveness across the total cost of marketing and translation, not a low price per word, and a catalogue is where that argument is easiest to see.
Volume at pace is a solved problem when the memory and the process are in place rather than improvised. For Frasers Hospitality, IPPWORLD delivered a website relaunch in 15 languages — including Chinese, Japanese, Arabic and German — handling over half a million words in eight weeks, integrated into the client’s Adobe Experience Manager environment alongside CRM content, brochures and fact sheets. That is not a retail catalogue, and the relevant transferable facts are the throughput, the language count and the fact that it ran inside the client’s own content system rather than beside it.
Integration: the store, not the document
Localised content that arrives as files has not been delivered; it has been handed over.
The difference between a language programme that holds and one that decays is almost always whether the content goes back into the system the store is actually run from. Where translated copy is pasted in by hand, three things follow: the store and the source diverge at the first update, nobody can say which version is live, and the next season starts with a reconciliation exercise instead of a brief.
IPPWORLD’s published differentiators include Direct Integration, and its lead-tier services cover website localisation including content management system and booking-engine integration. The relevant precedents are digital products rather than brochures. For FWD, the Omne app was localised into seven Asian languages — Simplified Chinese, Traditional Chinese for Hong Kong, Japanese, Vietnamese, Khmer, Thai and Bahasa Indonesia — with content management system integration and linguistic quality assurance that allowed a simultaneous launch across its Asia-Pacific markets. For Resorts World Sentosa, IPPWORLD is the appointed language service provider to the Digital and Regional Marketing teams, working in Sitecore with a Translation Memory maintained per language.
The questions worth settling before a word is written are therefore operational: which system holds the master content, who publishes, how a product update propagates to every language, and whether the Translation Memory is being written to as part of that flow or alongside it.
How to brief it
A brief that names the surfaces and the treatment per surface will cost less and come back better than one that names a word count and a deadline.
IPPWORLD’s process runs in six steps: Understand the Project Brief, Pre-Transcreation Study, Transcreation, Client Review, Update and Approval, Final Delivery. The Pre-Transcreation Study is where market and brand research happens before any copy is produced, and it is the step a thin brief starves.
What to put in front of a language partner:
- A surface inventory, not a word count. Product copy, specification fields, interface strings, operational pages, campaign and CRM — with the treatment you expect for each.
- The markets and the language variants. Not “Chinese” but which Chinese: IPPWORLD works in Simplified Chinese for mainland China, Singapore and Malaysia and Traditional Chinese for Taiwan and Hong Kong, and they are not interchangeable in retail copy.
- The fixed terms. Product and collection names, size and shade naming, loyalty tiers, delivery tiers — decided once, per language, before the catalogue runs.
- The system and the release cycle. Which platform, who publishes, how often, and what happens to a product update.
- Who reviews, in market. Separate the accuracy reviewer from the brand reviewer, and give the brand reviewer the tone-of-voice guide rather than the source text.
- The sequence. Settle interface and operational content first, then the catalogue, then the campaign — so the campaign lands in a finished shop.
At IPPWORLD, we are an award-winning transcreation agency that specialises in creative translation and marketing localisation for the hospitality, travel, lifestyle and retail sectors, named Best Hospitality Creative Translation & Localisation Agency 2019 at the LUXlife Hospitality Awards. We work in 30 languages: 18 Asian, 10 European and two Middle Eastern.
Conclusion
E-commerce localisation goes wrong in predictable places, and almost none of them is the product description. It goes wrong in the filter labels, the returns page, the size guidance and the gap between a transcreated campaign and an untranscreated shop.
Separate the store into its four content types and assign a treatment to each. Translate specification and operational content exactly; transcreate the copy whose job is to sell. Test interface strings in place rather than in a spreadsheet. Put a dedicated Translation Memory behind the catalogue so the second season and the second market cost less than the first. And settle the integration question before the writing starts, because content that cannot get back into the system will not survive the first update.
IPPWORLD’s clients choose it for quality, turnaround time and service, in that order. On a catalogue, those three are the same decision.
Frequently Asked Questions
What does e-commerce localisation include beyond translating product pages?
Which parts of a product page should be transcreated and which translated exactly?
Why does a Translation Memory matter more for retail than for other content?
Can localised content be loaded straight into our platform?
We sell into several Chinese-speaking markets. Is that one language?
In what order should an online store be localised?
How many languages can a retail programme run in at once?
Send your surface inventory, your markets and your platform and we will mark what needs transcreating, what needs rendering exactly, and where the memory and the integration belong. More on the work is on our website localisation and marketing translation pages, and the sector work in our retail and lifestyle practice.