Key Takeaways
- Big Data Scale: Translating training materials with a volume of several million words in just over a year.
- Innovation 2012: Early adoption of cloud Translation Memory (Memsource), enabling real-time team collaboration.
- Process Engineering: Application of custom scripts to convert raw CSV files into HTML, securing the code against corruption.
- Efficiency: Executing a gigantic project with a team of fewer than 10 people.
- Corporate Consistency: Co-creating a unified management methodology for the entire SABMiller brewery group in Poland.
Context: One System to Rule Them All
Implementing the global SAP standard at Kompania Piwowarska (part of the then-giant SABMiller) was like performing surgery on the living organism of the enterprise. To ensure Polish employees could effectively use the new environment, it was necessary to translate a gigantic amount of technical documentation, training materials, and user interfaces. We are talking about a volume in the range of several million words.
The challenge lay not only in the quantity but also in the format. The client provided data in raw CSV files generated directly from the database. The structure was rigid: the translation had to land in a precisely defined spot (exactly after the seventh comma of a multilingual document). Every error – deleting a comma or breaking quotation marks – threatened a critical error when importing data back into SAP.
Developers at the Service of Translation: Solving the CSV Problem
Working on such files in Excel was impossible – the risk of human error across thousands of rows was too high. We needed a solution that separated the technical layer from the linguistic one. Instead of translating manually in spreadsheets, we bet on Localization Engineering.
Together with a programmer, we developed a proprietary script that automated the process:
- Step 1 (Extraction): The script extracted the content to be translated (embedded in HTML code) from the CSV file and saved it in HTML format.
- Step 2 (Fooling the Tool): We imported the prepared files into the Memsource system. The CAT tool parser recognized the HTML tags and... hid them from the translator. The translator saw clean text, while the document structure remained safe "under the hood."
- Step 3 (Reconstruction): After translation, a second script decoded the files and inserted the Polish text perfectly into the required spot in the CSV structure.
Thanks to this "hacker" approach, we reduced the risk of code corruption to almost zero.
The Cloud in 2012: Real-Time Collaboration
Today, working in the cloud is standard, but in 2012, the translation industry still relied heavily on files sent via email. We decided to take a risk and implement the Memsource (now Phrase) system in its server version.
It was a revolution in efficiency. A team of just 10 translators and 2 proofreaders worked on a single, shared Translation Memory (TM). When one translator approved a term like "correction invoice," the rest of the team saw this translation as a suggestion in their segments split seconds later. This ensured ironclad terminological consistency, crucial for SAP documentation.
The Result: Millions of Words, One Standard
The project ended in full success. In a relatively short time and with the involvement of a small but highly specialized team, we delivered complete, localized documentation. The use of local SAP Term dictionaries and rigorous Quality Assurance (QA) ensured that the system implementation at the Poznań brewery proceeded without linguistic obstacles.
This experience taught me that in large translation projects, technology is just as important as linguistics. A well-written script can save hundreds of hours of manual labor.
Verified references
