In July 2015, Spotify launched Discover Weekly, a playlist of new songs delivered to each listener every Monday and tailored to their taste. Spotify’s engineering team called it one of its most successful feature launches ever, writing that “every Monday Spotify delivers 75 million unique mixtapes to music lovers all over the world.” Each playlist was a personalized set of around 30 tracks the listener had not heard before but was predicted to enjoy.
The system combined several recommendation techniques. It looked at what each user played, saved, and skipped, then found patterns by comparing that behavior against the listening of millions of others and against the contents of human-curated and user-made playlists - a form of collaborative filtering - while also analyzing audio attributes of the songs themselves. The team noted that combining “clever algorithms with human curation” let them manage personalization at scale, refreshing tens of millions of playlists every week.
Discover Weekly became one of the clearest mainstream examples of recommendation AI shaping culture. Spotify reported that by its tenth anniversary, listeners had streamed more than 100 billion tracks from Discover Weekly playlists. For many users it was the first time an algorithm felt like a knowledgeable friend recommending music rather than an obvious upsell.
For business readers, Discover Weekly shows how a recommender system can become a flagship product feature and a retention engine in its own right. It is a consumer-facing cousin of the technology behind the Netflix Prize, turning behavioral data into a weekly habit that keeps people coming back.