Localization at Lyft
Terena Bell is an independent journalist writing for The Atlantic, Washington Post, Fast Company and others. She is former CEO of In Every Language and was on the GALA and ALC boards.
When it comes to crowdsourcing, ride-share apps are the commensurate model. Instead of hailing a traditional cab, customers use an app to catch a ride in an amateur driver’s private vehicle. But interestingly enough, Lyft, a popular US-based app, doesn’t crowdsource its translation. Instead, localization program manager Brian McConnell works with three language service providers and a translation management software (TMS) vendor.
Welcome to Client Talk, a bimonthly column where we chat with the people who buy translation. How do they make their purchasing decisions? By connecting away from the sales environment, we hope to discover some larger logic behind client decision making. Collectively, what do these interviews tell us about how buyers see the industry as a whole?
Lyft’s headquarters are in San Francisco, California, but you can catch a ride in both the United States and Canada. At the time of this interview, the company didn’t translate into French, though — just US Spanish. Its November 2017 Canadian expansion was into the predominantly English-speaking city of Toronto.
McConnell himself is far from new to the industry, having worked as head of localization for both the news sharing site Medium and sales software company Insightly, Inc. He was also director of API integration for translation provider Gengo and founded the open source translation platform Worldwide Lexicon.
Having been both vendor- and buy-side, McConnell says, “[Clients’] biggest mistake — and one virtually every company makes — is waiting too long to invest in localization engineering work. By then, they usually have extensive technology debt baked into their systems, which takes at least 6 to 12 months to clear out.” Note that Lyft opened June 2012, but didn’t launch localization until November 2018, 21 months after McConnell began work.
The need and budget
Currently, Lyft translates what McConnell calls “all user touch points — apps, websites, support, email, notifications, etc” — for Spanish-speaking drivers and customers: around 500,000 source words. At the time of interview, five more languages were planned.
Lyft doesn’t share financial numbers, but when it comes to what the company spends on localization, McConnell does say it’s important to put budget in perspective, noting for any buyer “direct translation costs are a fraction of the engineering and organizational costs associated with building a platform that can operate in multiple languages and countries.”
When are professional translators used?
Translation is performed by three different vendors: one for software, one for contracts and privacy policies, and a third for marketing and web content. The latter isn’t a translation provider so much as a transcreation company. One of these vendors does linguistic quality assurance on the others’ work and, according to McConnell, “in-house linguists” also review “high-value or high-visibility content for extra quality.”
At Lyft, software development is decentralized, which means each engineering team produces its own content. So the company uses a popular translation management platform to organize source material in a single place. This allows McConnell to deploy a continuous localization process, which is especially helpful for high-volume, time-sensitive information like coupons and event announcements. “An automated translation pipeline — continuous translation — was critical to supporting these use cases,” he explains.
So on a scale of 1-5, where does Lyft rank professional translation?
A 5, which is the same score he gives the importance of a TMS: “Without a TMS, we would not have been able to keep track of the constantly-changing source material awaiting translation or review. Having a centralized translation service enables us to provide these teams with a highly-automated translation pipeline, as well as enforce best practices with respect to content strategy, translator comments, etc.”
An emerging pattern
Client Talk may have started as a column about purchasing translation, but as we profile larger and larger clients, it’s slowly shifting focus toward the importance of translation technology. This isn’t intentional, as MultiLingual is vendor-neutral — which is why we’ve omitted the name of Lyft’s partner. But vendors aside, once word count rises above a certain threshold, buying tech and buying translation become synonymous.
McConnell says, “The single most important bit of advice I have for early-stage companies is to implement localization tooling as soon as you can and make it part of your engineering process. Even if you remain in your home language, you can use the same tooling to ‘translate’ from rough English to finely-edited English. Developers can write functional but not elegant English, which then gets translated into finished copy using the same workflow you would use to translate into another language.”