Masstranslate
Making a whole product catalogue exist in every market at once.
Translating a catalogue before the window closes
A retailer puts a season on sale on one date, in every market at once. Thousands of stores, seventy-odd markets. The clothes are ready. The photography is ready. What is not ready is the words: every title, every description, every piece of search metadata, in every language, by that date.
That is the job Masstranslate was built for. We built it alongside retailers who had exactly that problem, and then made it something any merchant can use.
It writes, then it edits
The pipeline runs in two stages. The first produces the translation. The second goes back over what the first produced and edits it. Splitting the two is what separates this from putting a catalogue through a translation API and hoping.
Then it goes to you. Nothing publishes until a merchant approves it, so the volume is ours and the last call on quality is yours.
Some of the words are yours
A catalogue carries language nobody gets to reinterpret: the line names, the fabric you invented a word for, the phrasing legal signed off on. Masstranslate picks those out as it goes and keeps a glossary of them, so the twelfth market says what the first one did. On top of that you can write rules of your own, per market, for the places where a term has to land differently in Lisbon than it does in Lyon.
Search wants different things in different places
Every market has its own search engine habits, and the same product needs different metadata to be found in each. What we handle is the mechanical half: a URL slug that reads properly in the local language, and hreflang set correctly so search engines know which version belongs to which market.
That is technical correctness, and it is worth being precise about what it buys you. Getting it right means you are findable. It does not by itself mean you rank.
Getting it into the store
There are native integrations for Shopify, WooCommerce, Wix, Webflow and Magento, and an API for anything built in-house. Connecting a store takes hours rather than a project.
Headless and composable stacks go through an edge proxy we wrote for them, which is the part that took the longest. A catalogue on Next.js or Nuxt with the content coming from somewhere else entirely has no single place to put a translation, so the proxy serves the right language at the edge instead of asking the storefront to know about any of this.
What a catalogue costs to translate by hand
Take one product going into twenty markets. Title, description and search metadata come to around 200 words, so twenty markets is 4,000 words of translation for that single product. At the $0.12 to $0.18 a word agencies charge for common European pairs, carrying that one product costs between $480 and $720, and it occupies a translator for two working days.
Most products do not cost that much to make.
Scale it and the arithmetic gets worse. A 5,000 product catalogue into twenty markets is 20 million words: somewhere between $2.4 million and $3.6 million, and 45 translator-years. Post-editing machine output roughly halves the days, and somebody still has to read every line.
The wall you hit first is the calendar. Three hundred new products going live in twenty markets on one date is 1.2 million words, which is 600 translator-days. To land that inside a five day window you need 120 translators working at once, every one of them holding the same glossary in their head and spelling your product line the same way.
One seasonal drop
A five day window needs 120 translators at once
Three hundred new products going live in twenty markets on the same date. The work does not get smaller as the date gets closer; it only has to happen in parallel.
300
new products
20
markets
1.2M
words to translate
600
translator-days of work
Translators working at the same time
So there are two problems stacked on each other: a bill most catalogues cannot carry, and a schedule no team can staff. The second is the one that decides whether you ship.
The reason to do any of it is well established. CSA Research surveyed 8,709 shoppers across 29 countries and found 76% prefer to buy where the information is in their own language, and 40% will not buy from a site in another one. A market you have not translated is a market where the majority of shoppers would rather be somewhere else.
What it costs
A flat rate rather than a price per word. That matters most in the week it would hurt: a per-word bill scales with exactly the seasonal drop that made you need the tool, and arrives after you have committed to the launch.
What we can claim, and what we cannot
More than a hundred thousand product translations have gone through it across live markets. That number says the machinery holds up under a real catalogue. It does not say how many were published unchanged, and that is the figure that would actually tell you something.
We also have not measured what any of it did commercially. The argument for translating a catalogue well rather than adequately is that it sells more, and we have not put a number on that.
What we would measure next
How often a merchant edits before publishing, and which categories they edit most, because that points straight at where the second stage is weakest. Whether a market's revenue moves after a catalogue is redone properly. And how the glossary behaves over a year, as a brand's own vocabulary drifts and the rules written in January stop matching what the copy team means in November.