+40% machine translation accuracy
Internal linguistic teams working together
15 million words per year
How Internetstores brought their internal translation teams together and adapted to remote work
Workflow speed, Machine translation
Internetstores is Europe's leading online retail group for outdoor equipment. Specializing in adventure brands, they own a selection of nature & biking brands and focus on e-commerce. The Bikester, Fahrrad, and AddNature marketplaces put Internetstores on the top spot when it comes to online shopping for nature hardware. They employ more than 800 people in Stuttgart, Berlin, Lyon, and Stockholm.
The project scope
E-commerce marketplaces like Bikester, Fahrrad, or AddNature combine no less than 10 languages as Swedish, Norwegian, French, Italian, or Polish. All products are multilingual and placed in a group Product Information Management (PIM) system.
Internetstores rely on internal translation teams to making sure all products are available in those languages at any time.
This means the 25 linguists from different brands, located in different countries, need to synchronize to deliver translations at a group level.
Sharing translation projects & translation memories is key to deliver content with efficiency.
That is why Internetstores' Localization manager needed to enhance workflows and processes for a quicker translation turnaround. They were searching for a tool to stop worrying about sharing individual translation memories. There obviously was a need for machine translation to provide a better basis for linguists to work on. They were using Google and Deepl MT to translate to Nordic languages like Swedish. But using these engines did not provide the quality they hoped...
Finally, translators were losing time when trying to display the initial product pages, which is important to create translations that match the context of the actual products. They had to search for each product in the PIM, then search on the website... They would either lose time or translation quality.
With Wezen, all personal translation memories merged into 1 common memory, live between all linguists. This now ensures people working on similar products can re-use existing translations, and corrections are always integrated for future texts.
Internetstores did not even have to wait for Wezen to integrate with their PIM to benefit from the solution. Working with PIM exports, the localization manager is able to create translation tasks for everyone in a few clicks, and everything happens in Wezen afterwards 'as opposed to using multiple software like JIRA, Confluence, CAT Tools,...). This means that product catalogs are translated and integrated more seamlessly than ever.
As for access to the initial product pages, the links are now provided within each translation task. This means no more searching in the PIM and websites, translators can display the pages in a few clicks!
As a localization manager, it is definitely a relief to have someone from Wezen always there to support a question about import automation, handling new content types, adding new MT engines, and helping with change management through training and feedback sessions.
A great part of the success of this story must be credited to the positive relationship between Internetstores teams and the Wezen Customer Success manager.
Configuring a custom engine for translation requires a good understanding of the content. As a result of this close collaboration, the accuracy of the engines went up to 80%, as old generic engines would sit around 50% accuracy.
Ultimately, the engines are actually helping translators be more efficient, with the correct tone of voice specific to Internetstores.
Meet our team member