Optimize for international audiences
The TED iOS app back in 2018. This was the preview video I created for Apple Store.
Born and raised in China, I fell in love with TED early on and treated it as my go-to destination for idea inspiration and English learning. Being able to join the mobile team at TED after my grad school was a dream come true. Only this time, I shifted from being an end-user to the designer who was tasked to optimize the iOS and Android app for international audiences speaking over 100 languages. I couldn't be more excited about this opportunity.
This experience provided me the chance to conduct international user research, and learn about how languages and cultures are at play in product design.
Product designer, full time
Nov 2016 – Dec 2018
We had over 80% of our app users located outside of the US. However, other than localizing the app in 20 languages and providing subtitles in more than 100 languages, there was a lack of insights into international user behaviors or their unique pain points.
Thanks to the international nature of TED events, we were able to conduct in-person and remote interviews with TED users representing 13 languages in total.
a) Identify unique international user needs and behaviors.
b) Understand how English proficiency affects content preference.
c) Understand what kind of TED content is relevant for international users.
Findings and archetypes
As we conducted the research, I started to notice the pattern of how English proficiency influences their watching behaviors. Based on the findings, three archetypes emerged.
1. Users who require local subtitles
This group of users is the least proficient in English. They wouldn't be able to understand TED Talks without local subtitles. Based on app and subtitles settings preference, 70% of our app users are estimated to fall into this group.
2. English practicer
This group of users strives to learn and improve their English by watching TED Talks. Their unique behaviors include 1) switching between English and local subtitles, 2) changing playback speed, and 3) pausing the talk to look up particular words. Mapping these patterns to our in-app tracking data, we can estimate the size of the group to be 20% of total app users.
3. Fluent English speaker
This group is the most proficient in English. They usually grew up speaking English or immersed in a similar environment. They usually don't need subtitles while watching and have their app language set to English. It's estimated to be about 10% of the app users.
This exercise provided us concrete ideas of the unique needs and respective size of each category. It continued to serve as a great reference throughout the design process.
Product iterations and testings
Informed by the research findings, we identified certain areas that we could start to improve in the app. We learned that subtitles and the video player functionality in general play a large role in users' viewing experience. In addition, we should try to recommend talks that they are most comfortable to watch based on their English proficiency.
Feature 1 – dual subtitles
Knowing that English practicers are switching back and forth between English and their local subtitles, we decided to tackle the top requested feature from a few Asian regions (China, Japan, and Korea) – the ability to watch TED Talks with two subtitles at the same time.
I grew up watching dual subtitles in movie theaters so it wasn't a foreign concept to me at all. However, it's the subtitles availability that made this design challenging. TED Talks are subtitled by volunteers so there is a period of time when local subtitles are not available for us to enable dual subtitles. As a result, we came up with the "Fallback subtitles" concept to let users fallback to English when their preferred language isn't yet available.
Enable dual subtitles in Android.
Before launching the feature, I ran the prototype with 5 users on usertesting.com and received all green lights to go.
After launching the dual subtitles feature, the app received very positive feedback in the Play Store. In addition, a year after the feature was launched, the average time spent per session by non-English users with dual subtitles on has almost doubled, comparing to non-English users who don't have the dual subtitles on.
Feature 2 – personalized homepage
From our research, we learned that we can roughly tell our users' English proficiency apart by their app language settings. For example, users who have their app language set to Chinese but video subtitles set to English can be categorized as "English practier". It means that when populating home screen content, we can recommend latest talks that have not yet been translated into Chinese. If a user have both their app and subtitles languages set to Chinese, it's a signal that this user may not be able to view TED Talks without Chinese subtitles. In this case, we only recommend latest talks that have been translated on their homepage.
How the home screen content adapts to app language settings.
Unfortunately, the team was going through an organizational structure change, so this feature was designed but has not yet been launched. I look forward to the day that we can get closer to this vision as I strongly believe that the redesign can significantly improve our international audience's experience.
Qualitative and quantitative findings can go hand in hand.
After identifying three archetypes, we asked ourselves what would be the relative size of each group. From there, it inspired us to track users' in-app subtitles changing behavior, and it successfully helped us identify a rough look at the group size.
Find ways to simplify and invent.
Dual subtitles user flow can become rather complicated, and we have almost no other app with similar features to reference. We simplified the design by only enabling the dual subtitles feature for users who have their primary languages in Non-English. It helped save technical effort too.
An one-size-fits-all design solution for all cultures may not be possible.
I learned from this project that many cultural nuances can influence how the product design differs by county. For example, Japanese written language is similar to Chinese, so their transcripts can be sliced the same way for dual subtitles. However, Korean, on the other hand, is different. There is hardly a one-size-fits-all solution.