How Udacity Saved 15% on High-Quality Course Translations
This case study illustrates how Udacity managed to cut translation costs by 15% while improving the quality of their Korean course translations, through strategic collaboration with Hansem Global, leveraging our expert quality assurance processes and translation capabilities.
Project Summary
Udacity is talent transformation platform specialized in technical skill training and career development. They work with more than 1,400 experts and have a vast library of exclusive content that has served 21 million learners in 195 countries. Being a content-driven platform serving global learners, high-quality translations are a critical part of their growth strategy. And that’s where they got stuck.
Challenges
- Udacity had major issues with their Korean translation quality which was impacting customer experience and causing problems with learning outcomes. Problems included inconsistencies, incorrect technical terminology, and awkward literal translations that sounded machine-translated.
- On top of that, Udacity’s content volumes were massive, with a large number of courses per quarter. Lastly, they dealt with complicated file types including YAML, JSON, SRT, and PDF.
- They needed to partner with a translation agency that had domain expertise in AI and data science and the capacity to handle their large volume of work.
Our Solutions
When Udacity came to Hansem requiring high-quality Korean translations, the first thing we did was run an in-depth audit. Upon analyzing the previous work, we found inconsistencies, wrong use of terminology, as well as clumsy literal translations – unsuitable for Udacity’s highly educated target audience.
Our solution to Udacity’s problems was trifold:
Quality, SME-driven human translations.
We employed professional bilingual linguists – no machine translation. For the duration of the project, we handpicked subject matter expert linguists to translate the machine learning, data science, Java, and SQL-related topics Udacity needed.
We employed automation.
While artificial intelligence was not used for translations, we did implement automation to avoid manual terminology work and to leverage from translation memories (TMs). This meant we were able to automatically reuse previously translated content, which saves money, time, and helps maintain consistency across the board. To ensure Udacity got maximum benefit, we updated the TM in real time (even when multiple, concurring projects were handled at the same time.) As a bonus, we also built the glossary as the translation project was in progress. The more we translated, the more we added to the glossary.
Quality Assurance automation.
We implemented a thorough Quality Assurance process that included automated quality reports shared among all the project stakeholders (in-house reviewers, Quality Assurance Team, project managers, etc.)
Additionally, we also created an automated Edit Distance report to monitor and ensure translations were edited to the standard, in a tone of voice cohesive with other Udacity materials.
Over the course of the project we:
- Worked with multiple complicated file types (YAML, JSON, SRT, and PDF)
- Translated more than 10 million words over three years.
- Handled 5 highly technical nano-degree courses per quarter with SME linguists.
Outcome
By the second quarter of working together, Udacity was saving 15% of translation costs. The more we continued translating for them, the more their savings increased, and the more their Translation Memory (TM) leverage was enriched. Because Udacity had a lot of repetitive content, the cost-savings were directly proportional to the large number of words we translated for them.
We also exceeded expectations in terms of quality and turnaround time:
- There were no linguistic issues found or negative feedback during client/SME review.
- All projects were delivered one week before the due date.
- Ad hoc requests were expedited and consistently met the challenging TAT.
With the improved quality of translation, the user experience improved, course comprehension increased, and translated courses received higher ratings.
Conclusion
When Udacity came to Hansem Global, they’d been disappointed by translations and localization providers before. Their Korean translations were in poor shape, and we had to roll up our sleeves and get to work. We turned the difficulties posed by the large word count and complicated subject matter into a winning translation program that now efficiently connects Udacity with the Korean market. What’s more, we helped them save 15% over the past translation efforts.