Localisation is a structure decision, not a translation task.
Multilingual budgets disappear when translation is treated as something applied to finished modules. Designing for it from the start costs a fraction as much.
Why does localisation get so expensive?
Because it is usually treated as a final step. Modules get built and approved in the source language, then translated, and everything that assumed English text length breaks. German runs roughly a third longer, layouts overflow, voiceover timings no longer match.
Every one of those is cheap to prevent and expensive to fix.
What does designing for localisation involve?
Text expansion allowance built into the templates, no text baked into images, voiceover scripts written to be re-recorded rather than dubbed over fixed animation, and a component structure where a market can republish without rebuilding.
The Roche Diabetes Care catalogue was built this way from the beginning, which is why global affiliates could localise and republish rather than commissioning new versions.
How is regulated content handled across markets?
Carefully, because a linguistically accurate translation can be regulatorily wrong. The Neuro Profiler financial literacy programme was localised across English, French, Spanish and German with the European regulatory context checked per market rather than assumed from the source.
Common questions.
How many languages can be handled at once?
The number matters less than whether the structure was designed for it. Four is straightforward when the templates allow for it and painful when they do not.
Should we use machine translation?
As a first pass for volume, with human review, on content where accuracy is not regulated. Not on medical or financial content, where a plausible-sounding error is the whole risk.
Where this applies.
Complex and regulated financial concepts made understandable for audiences from young learners to adults making investment decisions, localised across English, French, Spanish and German.
Read the case study →Training for rapidly evolving HCP-facing software was going out of date faster than the agency model could keep up with. Julia Mattey designed a global blueprint and production process in Articulate Rise, catalogued in Cornerstone.
Read the case study →Other guides.
Tell me what is not working.
If your training is expensive to maintain, slow to update, or not landing with the people it is for, I can usually tell you fairly quickly whether that is a content problem, a platform problem or a process problem.