Personalization
A personalization system that learns what millions of viewers love — and continuously reshapes Tubi around their tastes.

Tubi
2024
Product design, Personalization, Growth
With more than 100,000 titles, Tubi's challenge wasn't having enough to watch — it was helping each viewer find the tiny fraction of that catalog they'd actually enjoy.
I designed a personalization system that began by learning viewers' tastes through genre onboarding, then continued learning from signals like likes, dislikes, watchlists, searches, ratings, watch behavior, and follows.
Through continuous experimentation, we evolved onboarding from simple genre selection to richer preference signals — including increasing selections from three to five, which drove a 66% lift. The resulting framework helped turn personalization from a one-time setup into an ongoing conversation between the viewer and Tubi.










