In a blog post published in the beginning of March on Medium, behemoth streaming platform Netflix outlined its creative approach to testing its site’s localization. Designed by Netflix’s Internationalization team, Shakespeare — the name for the tool, not the Bard — has helped Netflix improve their copy testing, allowing them to test functions that weren’t available before, such as language optimization for emerging markets and word choice validation.
The localization efforts of Netflix — a streaming platform that surpassed 200 million users worldwide this past January — are much more complex than simply subtitling and dubbing their videos: they have to ensure that their website is user (and language) friendly so as to encourage visitors to use and sign up for their service. Thus, localizing for the markets that they serve is an important task for the streaming platform to optimize and perfect.
And it was out of this need that Netflix’s Internationalization team created Shakespeare. As outlined in the Medium post, written by Tim Brandall, Shawn Xu, Pu Chen, and Jen Schaefer, their localization efforts stumbled upon a number of areas of difficulties before the advent of Shakespeare: namely, there was no uniform way to test localized or transcreated copy across the platform in the different languages it serviced. Recognizing that they needed to improve this process significantly, Netflix assembled a team of engineers, content designers, data scientists, and language managers, to produce a system that would identify copy variants more efficiently.
To allow for this localization improvement, the Shakespeare system is connected to Netflix’s translation tool by way of a message repo. When users visit the website (or one of the Android or iOS apps), they encounter one of the copy variants that is being tested. They are prompted to respond to cues such as “see prices”; their responses are then compared by the team to determine which variant performed the best. According to the authors, Shakespeare has allowed Netflix to significantly reduce the time required to deploy A/B copy tests: while before, this type of copy testing required days or even weeks to produce reliable results, Shakespeare offers results in minutes.
Shakespeare is also much easier to configure, allowing the teams running it to make copy updates in real-time. The Medium post states that Shakespeare has conducted more than 50 copy tests successfully. Shakespeare is an important innovation in localization, as it allows localized copy to be tested without referencing or relying on its English source, meaning Netflix is now able to naturally cater to its user bases in countries that speak languages other than English.
