I led localization and internationalization at Lyft and Notion among other startups. The two experiences offer a contrast between a company getting international expansion right, and not so right. Both are great companies with great people, but operating internationally is tricky and it’s easy to get wrong.

Notion

Notion has been hugely successful internationally, with over 70% of its users hailing from outside the United States. The company offers a suite of knowledge and project management services as a SaaS offering, and has a rabid following. The company localized its product soon after it saw organic growth in Korea and Japan, so it largely avoided accruing tech debt compared to other companies I worked for.

Among the things Notion did that contributed to its international success:

• The mobile and web apps were localized soon after the company saw organic growth in non-US markets. Tech debt in the code base was largely cleaned up as part of this early effort.
• The engineering team invested in translation automation so translations were continually updated and checked in without engineering intervention.
• Most other user touch points were localized from 2021-2022 using integrations with our primary translation management system.
• The company hired power users to fine tune translations produced by vendors. This made the difference between a product that felt lazily translated versus one that felt like it was made for users in their language.
• The company opened field sales offices in Japan, Korea and the EU to boost enterprise sales, and also worked with distributors in key markets like Japan. Field offices were another source of feedback that further improved product quality in other languages.
• The product itself was well designed and appealed to users in international markets.
• The founder, Ivan Zhao, is from China, and intuitively understood the importance of language support and accessibility.

All of these things reinforced each other, and as a result non-US markets account for a large majority of the company’s revenue. While I can’t share specific numbers, revenue from non-US markets is roughly 100x what we spent on localizing the product and user touchpoints. If we hadn’t invested in adapting the product for international markets, we would not be seeing anywhere near the revenue we received from non US regions.

Lyft

Lyft, on the other hand, was a story of what might have been. I joined the company in 2017, initially to work on their expansion into Canada, and then to roll out support for additional languages. Where Notion experienced rapid growth internationally, Lyft struggled to operate outside of a few Canadian cities, despite the app being localized in a half dozen languages.

Lyft’s biggest mistake was to assume that it’s core product and business model would work outside of US cities. The problem is that ground transportation (taxi service) is heavily regulated in most places, and Lyft’s ride sharing business model was a non-starter in most cities outside the US. In retrospect, Lyft should have borrowed a page from Uber and acted as an agent for existing taxi services. For example, when you request an Uber in Berlin, you get a licensed taxi cab. This is super convenient, especially for travelers. It is also a scalable business model with good profit margins, similar to a travel agency.

Lyft had a number of other issues besides this “original sin” that held it back. Among them were:

• The company had been operating in the US and only in English for many years. It acquired a huge amount of tech debt that had to be cleaned up just to enable the app to function in other languages and regions. This took about a year and a half, and cost well over a million dollars to fix, and the opportunity cost of this delay was far larger.
• While the app appeared simple, it communicated with dozens of service end points, each of which were owned by different teams, each of which had different priorities (internationalization was usually not one of them). Getting all of these teams aligned on the project took way longer than it should have.
• The company was too “agile”. While I don’t want to return to waterfall project management, being too agile can get in the way of long term planning and execution. Localization was constantly stuck in team backlogs, which further slowed progress.
• Uber had already localized their apps into over a dozen languages. They killed us in the Spanish speaking market in the US (Spanish speakers were three times more likely to have the Uber app on their phone compared to Lyft).