In Sao Paulo, the fitness app told pregnant users to “push through the warning.”
The English phrase had meant tolerate mild discomfort. The Portuguese translation sounded like an instruction to ignore a medical alert. Lucía Ferrer learned this from a Brazilian physiotherapist who posted a furious video.
Her Barcelona company had expanded into nine languages using machine translation reviewed by freelancers.
“Remove the pregnancy program in Brazil and continue,” said the growth chief. “One bad sentence shouldn’t derail a country.”
Lucía asked how many other sentences were bad. No one knew.
Professional review of all languages would cost more than the next marketing campaign and delay Japan by months. Freelancers had already signed off, though each had reviewed isolated strings without seeing the exercises.
Lucia suspended the app’s higher-risk programs outside English and Spanish. Subscribers demanded refunds. App-store rankings fell. The Brazilian partner canceled a promotional deal, saying safety should have been designed before sales.
The company hired local clinicians to review complete user journeys, not word lists. They found smaller errors everywhere: jokes that sounded insulting, units left unconverted, instructions that assumed users could kneel. Fixing them produced fewer dramatic features and many unglamorous tickets.
The board removed Lucía as chief executive after two weak quarters. She stayed as head of product, reporting to the growth chief who had argued against the pause.
Japan launched a year late with fewer programs and a local medical advisory group. Signups missed the revised forecast. Complaints were ordinary: price, music, difficulty.
At a team meeting, the new chief executive praised the safe launch as a foundation for scale. Lucía listened without correcting the history.
Later, the Brazilian physiotherapist emailed to say the revised pregnancy program was responsible, if cautious. Lucía preferred the second adjective. It sounded earned.
