B2B SaaS Platform (NDA, Europe)

The client is a B2B SaaS platform headquartered in Europe, serving customers across 25 markets. Their product ships continuously, two-week sprints, sometimes hot-fixes mid-sprint, and every sprint introduces around 3,000 new or modified strings across UI, in-app help, email templates, and onboarding flows.When they came to us, the localization setup was a patchwork. One agency handled European languages, another covered APAC, and a third did MENA. Each vendor had its own project manager, its own quote format, its own quality issues, and its own delivery schedule. The average turnaround from a developer committing a new string to that string appearing in all 25 languages was five days. For a team trying to ship weekly, that was a release blocker.The bigger problem was sprint adherence. Roughly 1 in 6 sprints either shipped with missing translations (English fallback visible in production), or got delayed because the localization deliverables hadn't arrived. Their Head of Globalization was spending most of her time chasing vendor status emails instead of working on strategy.They had tried building an internal localization team. It worked for about a year and then scaling stalled. Hiring senior linguists in 25 languages out of one European office wasn't realistic, and the contractor pool they were assembling was expensive and inconsistent.

The first conversation wasn't about translation. It was about their dev workflow. We spent the first two weeks just understanding how they shipped product. Which TMS they used (Phrase), how their CI/CD pipeline was structured, where strings lived in their repo, how their PMs flagged content for translation, and what their definition of "done" was for a localized release.The core decision was to stop treating localization as a separate workflow. We integrated directly into their Phrase instance through API, set up webhooks for new content, and built a routing layer that automatically assigned incoming strings to the right language pod based on content type and priority tag.We restructured the linguist side too. Twenty-five language pods, each with a senior translator and a reviewer. Each pod was on-call during their regional business hours, which meant strings committed by the dev team in Berlin at 10 AM were typically being worked on within an hour, with Asian languages picked up overnight and European languages handled same-day.The QA layer was the part that took longest to get right. SaaS strings have specific risks that document translation doesn't (character limits in UI elements, ICU MessageFormat placeholders, plural rules, gendered grammar in form labels). We built an automated QA layer that ran every translated string through validation checks before it went back to the client's TMS, catching things like broken placeholders or strings that would overflow buttons in German or Russian. Anything that failed automated checks went to the reviewer with a flag.We also rebuilt their translation memory. The three previous vendors each had their own TM, with overlapping entries, conflicting terminology, and roughly 35% leverage on average. After three months of consolidation and cleanup, TM leverage was up to 62%, which translated directly into lower cost per word and faster turnaround on familiar content.The integration engineering work, two of our engineers worked with the client's platform team for the first six weeks, was probably the highest-leverage investment. Once the pipeline was running, manual intervention dropped to almost zero. Strings flow in, get translated, get QA-checked, and flow back into the right branch without anyone emailing anyone.By month three, the 5-day TAT was 18 hours. By month six, sprint adherence was at 99.7%. The Head of Globalization had stopped chasing status emails.