Spotify Home Feed Case Study: From Playlists to Predictive Moments

Design · 6 min read

Spotify reorganized its home experience to focus on moments — commute, workout, chill — using short surveys and passive signals like time of day and headphone type. Cards are optimized for quick actions: play, save, and queue, with voice-enable suggestions in supported markets.

The ranking model combines collaborative filtering with contextual bandits that test format effectiveness. A/B tests show immediate uplift from predictive cards but also highlight long-term discovery decay if bandit exploration is too limited.

Design choices minimize friction: each card reveals metadata about why it was recommended. The teardown finds that transparency increases acceptance for algorithmic picks, especially when paired with controls to reset preferences, offering designers a clear path to build trust into recommender UIs.