How to Choose an E-commerce Translation Agency

A retail brand looking for an e-commerce translation agency will get three very different kinds of proposal, and the differences will not be obvious from the documents. One quotes a low rate per word against a large volume. One quotes a day rate and a named writer. One asks what system the store runs on before it quotes anything.

All three can do a creditable job on a product description. They behave very differently when the catalogue restocks, when a fourth language is added, when the campaign and the shop have to agree, and when something needs fixing in the system rather than in a document.

This piece sets out what to compare, in what order, and the evidence to ask for rather than take on assertion.

Key Takeaways

Compare suppliers on the second order, not the first. Almost any provider can deliver a one-off catalogue; the difference shows at the restock, the new market and the system update.
Three questions separate the field faster than anything else: who owns the Translation Memory, can content go back into your platform, and who reviews brand voice as distinct from accuracy.
The lowest rate per word is frequently the highest total cost, because literal output on brand-facing copy gets redone. Price the programme over two seasons, not the first export.

The three kinds of supplier

The market divides by what each type is built to optimise, and none of the three is wrong — they are answers to different questions.

	Volume-priced generalist providers	Individual freelance translators	Creative-translation and localisation partners
Built to optimise	Throughput at the lowest unit cost	Craft on a bounded brief	Brand consistency across surfaces and markets over time
Strongest on	Large one-off exports; many languages at once	A single language, a single voice, a close working relationship	Multi-surface programmes; repeat and updated content; integration
Weakest on	Register and brand voice; the creative half of product copy	Coverage, continuity and volume; cover during absence	Not the lowest unit rate on a one-off export
Translation Memory	Often held by the provider; portability varies	Usually none, or personal	Dedicated per client and reused across projects
Platform integration	Varies; frequently file delivery	Rarely	Part of the brief
Behaviour at restock	Re-quotes the volume	Re-quotes the time	Prices the increment against the memory
Behaviour at a new market	Adds a language pair	New supplier needed	Inherits the decisions already made

Most retail brands end up with a hybrid, and that is a reasonable destination: a partner holding the brand-facing surfaces, the memory and the terminology, with capacity bought where volume is genuinely undifferentiated. What does not work is the reverse — buying brand-facing copy on unit price and then paying a partner to repair it.

The question is not who can translate your catalogue. It is who will still be making the same decisions about it in two years.

Five criteria, in order

Order matters, because the first two are difficult to change later and the last three can be fixed in a quarter.

Sector and register fit. Can they show marketing copy, in your product category, in your target languages, that reads as though it was written in them? Ask for the target-language samples, not the English source and not a case-study summary. A provider who can only show document translation will produce document translation.
Translation Memory ownership and portability. A memory accumulates your terminology decisions and is the asset that makes later work cheaper. Establish who owns it, in what format it can be exported, and whether it is per client or pooled. A memory you cannot take with you is a switching cost disguised as a benefit. IPPWORLD builds a dedicated Translation Memory for each client and reuses it across projects.
Integration with your platform. Whether localised content can go back into the system the store runs from, or arrives as files to be pasted. This decides whether the programme holds together after the first product update. IPPWORLD’s lead-tier services cover website localisation including content management system and booking-engine integration, and Direct Integration is one of its six published differentiators.
Review model. Who checks accuracy, who checks brand voice, and whether those are different people. A single reviewer doing both will optimise for whichever they are more comfortable judging, and in practice that is accuracy.
Continuity. Named linguists per language, cover arrangements, and whether the same people return next season. Brand voice is held by people who remember the last decision.

The evidence pack to ask for

Ask for artefacts rather than claims, because every supplier claims quality and the artefacts differ enormously.

Target-language marketing samples in your category, with the source alongside, so an in-market colleague can judge register rather than accuracy.
A sample glossary and style guide from a real engagement, redacted. This shows whether terminology is managed or improvised.
A Translation Memory export, or a written statement of the export format and ownership terms.
The integration path for your specific platform, described concretely — not “we integrate with all major systems”.
A named reference in your sector or an adjacent one, and how the relationship started: appointed provider, competitive pitch or public tender.
The process, in steps, with the research step named. IPPWORLD’s runs Understand the Project Brief, Pre-Transcreation Study, Transcreation, Client Review, Update and Approval, Final Delivery — the second step is where market and brand research happens before copy is produced, and a process with no equivalent step is a process that starts writing immediately.
A paid pilot on a real subset. One category, two languages, your actual product export. It costs a fraction of the programme and settles more than any proposal.

What the engagements actually look like

Language relationships in this category tend to form in one of three ways, and knowing which you are running changes what you should ask for.

IPPWORLD’s published engagements show all three. For Resorts World Sentosa, and separately for Far East Hospitality and Millennium Hotels & Resorts, IPPWORLD is the appointed language service provider — a standing arrangement across websites, collateral and campaigns with a Translation Memory maintained per language. For Gardens by the Bay, the work came through a public tender with trials and became a two-year preferred-vendor contract covering brochures in 10 languages. For FWD, a contracted digital-product engagement localised the Omne app into seven Asian languages with content management system integration and linguistic quality assurance supporting a simultaneous Asia-Pacific launch.

IPPWORLD was also named Preferred Asian Language Localisation vendor by Starwood Hotels & Resorts in 2008, and won a Singapore Tourism Board public tender in 2010 for the translation and multilingual production of Singapore’s Heritage Site Markers. Those are useful to a retail buyer only as evidence of how the relationships are formed and held — appointed status and tender wins are both earned on trials and kept on consistency.

For a retail brand the practical read is: if you are buying a single catalogue export, run a pitch. If you are buying a multi-season, multi-market programme, you are appointing a provider, and the criteria that matter are the ones in the list above rather than the unit rate.

Price: compare the programme, not the export

A rate per word is a price for the first time you buy a sentence, and a catalogue makes you buy most sentences more than once.

Three costs sit outside the rate and routinely exceed it:

Rework. Literal output on brand-facing copy gets rewritten in market, usually by someone whose time is not in the translation budget. IPPWORLD’s clients come to it having experienced badly translated content, which is the same cost described from the other side.
Repetition. Without a memory, every restock and every season re-buys the recurring attribute, care and fit language at full price.
Reconciliation. Where content cannot go back into the platform, someone maintains the mapping between source and languages by hand, permanently.

IPPWORLD’s position is premium quality at competitive rates, argued as cost-effectiveness across the total cost of marketing and translation rather than a low price per word. Whether that holds for any given brand is a question to settle with a two-season model rather than a rate card — but the model is the right instrument, and a supplier who will only discuss the rate is answering a different question.

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

Choose on the second order. Any competent supplier will deliver a readable catalogue once; the differences appear at the restock, the fourth language, the platform update and the first time the campaign and the shop have to use the same word for the same product.

Judge sector and register fit from target-language samples, settle Translation Memory ownership and portability in writing, establish the integration path for your actual platform, insist that accuracy review and brand review are separate, and ask for named continuity. Then run a paid pilot on one category in two languages and let the output decide.

IPPWORLD’s clients choose it for quality, turnaround time and service, in that order. That ordering is worth putting to any supplier you are considering, and worth noticing if they reverse it.

Frequently Asked Questions

What should I compare e-commerce translation suppliers on?

Five things in order: sector and register fit shown through target-language marketing samples; Translation Memory ownership and portability; whether content can go back into your platform; whether accuracy review and brand review are separate; and continuity of named linguists. The first two are hard to change later, which is why they come first.

Why does Translation Memory ownership matter when choosing a supplier?

Because the memory accumulates your terminology decisions and is what makes later work cheaper. If it is pooled rather than dedicated, or cannot be exported, it becomes a switching cost rather than an asset. Establish ownership, format and exportability in writing before signing. IPPWORLD builds a dedicated Translation Memory for each client and reuses it across projects.

Is a low rate per word a good deal for a retail catalogue?

Not reliably. Three costs sit outside the rate and often exceed it: rework when literal output on brand-facing copy gets rewritten in market, repetition when there is no memory and every restock re-buys the same recurring language, and reconciliation when content cannot go back into the platform. Model the programme over two seasons rather than pricing the first export.

Should we use an agency, a freelancer or both?

Most retail brands end up with a hybrid, and it is a reasonable destination: a partner holding the brand-facing surfaces, the Translation Memory and the terminology, with capacity bought where volume is genuinely undifferentiated. The arrangement that does not work is the reverse — buying brand-facing copy on unit price and then paying to repair it.

What evidence should I ask for before appointing?

Target-language marketing samples in your category with the source alongside; a redacted glossary and style guide from a real engagement; a Translation Memory export or a written statement of its format and ownership; the integration path for your specific platform described concretely; a named sector reference and how that relationship started; the process with its research step named; and a paid pilot on one category in two languages using your real product export.

How do these relationships usually start?

Three ways, all visible in IPPWORLD’s published work. As an appointed language service provider across surfaces, as with Resorts World Sentosa and Far East Hospitality. Through a public tender with trials that becomes a term contract, as with Gardens by the Bay’s two-year preferred-vendor contract covering 10 languages. Or as a contracted project, as with the FWD Omne app in seven Asian languages.

How many languages should we start with?

Fewer than the eventual target, and chosen so the second market tests the model rather than repeats the first. IPPWORLD works in 30 languages — 18 Asian, 10 European and two Middle Eastern — so the constraint is rarely coverage; it is whether your terminology, review model and integration hold up before you multiply them.

Send your platform, your markets and a sample product export and we will come back with the treatment per field, the integration path and a pilot scope. More on the work is on our marketing translation page, the method in how we work, and the sector work in our retail and lifestyle practice.