eBay is a global success story, with 62 million active users and over 800 million items for sale on the site at any given time. As a global market leader, the opportunity for future growth lies with cross-border, multi-language listings. Their goal is to allow users to list products and instantly access the global market, creating an experience where every customer can search for product information in their local language, irrespective of the seller’s language.
Obviously, delivering this isn’t easy, and eBay have one of the most advanced and talented localisation teams in the industry, focused on solving these and a myriad of other challenges on a daily basis.
Building on eBay’s expertise
Lingo24 started working with eBay in 2015 and, ever since, we have worked together on a whole range of language challenges, delivering millions of words.
As experts themselves in localisation, eBay are in many ways a dream customer, with clear briefs, glossaries and project goals. However, they also have high expectations on quality levels and deadlines. By creating an expectation for excellence and helping us create an environment to achieve it, month-by-month, we’ve managed to strengthen our partnership and ensure we consistently meet eBay’s exacting standards.
The projects are as diverse as the language combinations. An example of a recurring application is post editing Machine Translation (MT) for product
descriptions, reviews and product titles. This human-checked content is then used to retrain eBay’s MT engines, as part of the continual improvement efforts.
Using professional translators with specific expertise
One of the key areas where we added value in these projects was to separate the product descriptions by category. It seems obvious, but we made a significant improvement in quality by ensuring the team working on translating dishwasher listings was different to those working on canoes or education! We introduced the use of translators with specific sector experience and created a global team of experts, all heavily engaged, to ensure high quality, localised content.
Adopting and adapting to eBay’s evolving tech stack
eBay are focused on using the best tech for the job-at-hand and, throughout our time working with them, we have adopted a range of proprietary, open-source and in-house tools, which have improved workflows and quality of translated content.
We’ve worked closely with eBay to test new platforms and even provided our own platform, Coach, whenever other options were not the best fit for project requirements. We’ve also been able to bring to the fore some of our tech experiences when using and on-boarding new technologies.
Specific outcomes have included working with the eBay team to generate metrics on the level of productivity improvements from using MT as a performance tool. This requires data on the level of post-editing needed, including time per segment, edit distance, as well as usage of automated QA checks, and delivering real time performance feedback on the work produced.
Categorising listings for local markets
Other areas we have worked on are diverse and include localising categories to ensure they are relevant and optimised for different geographies.
This is demanding work, as the level of research required of our team and the support administered by our project management team have been both been significant. Context is critical, and success can only be achieved through exhaustive research and a collaborative approach between translator, PM and client. These often cross over several months of activity and multiple resources, which adds inherent complexity, but again we have managed to take the load and deliver consistently against eBay’s expectations.
Optimising content for search
“Browse nodes sanity checks” is eBay terminology for our most recent addition to the project portfolio. Here, we are making sure that the localised search titles are consistent with best practice and, where required, adapting and improving relevance, removing redundancy and checking key terminology is accurate.
These activities allow us to build teams around language challenges, and flex and scale quickly – strengths that we can then carry across to other applications, and, indeed, other clients. They also give us an inherent understanding of the search, optimisation and specific challenges of working with User Generated Content across multiple locals.
In parallel, we also become experts on some interesting sub-categories – who knew there was such diversity of available Lego on eBay? Well, at the end of the last project, quite a few translators and all of our related Project Management team asked for a short break from any further Lego-related projects (it seems you CAN have too much of a good thing).