How Does Spotify Recommend Songs? (Explained)

Spotify is considered the best music streaming platform mainly due to its algorithmic-based playlists and recommendations. But how does it work exactly? Let’s talk about it.

How does the Spotify recommendation algorithm work?

Basically, Spotify’s algorithm (or more accurately, algorithms) analyzes the listening habits of millions of users to predict what each individual listener would want to hear in the future.

Spotify recommendations are made up of several algorithms. If you want to learn more about what the recommendations on Spotify are based on, you need to understand each algorithm.

1. The BaRT system

Image courtesy of

Spotify’s recommendations begin with an AI system called Bandits for Recommendations as Treatments (BaRT). The BaRT model has two components: exploitation and exploration.

The exploitation model recommends songs based on your listening habits. This includes what songs you listen to and how you engage with them (like, skip, or add to a playlist).

If a song has not been played often and there’s not enough data to prove the relevancy of the song to the user, then the system may not recommend that song to that user.

The exploitation model has its weaknesses when recommending songs because it relies heavily on listening history. However, the BaRT system makes up for it with the exploration model.

The exploration model recommends songs without considering your listening history. It makes recommendations based on what’s popular around you and what’s popular globally.

Spotify needs to gauge your interest in a song. By recommending a song you haven’t listened to, Spotify understands whether you’ll be interested in listening to similar songs.

It is useful to recommend new songs uploaded to Spotify to new users.

When both the exploitation and exploration model are used together, Spotify can make accurate predictions on what songs you are likely to enjoy, making their recommendations spot-on.

2. The 30-second rule

Spotify algorithms look at how long you listen to a particular song. According to former product director Matthew Ogle, the most important time frame is the first 30 seconds of a song.

If you listen to a song for longer than 30 seconds, Spotify recognizes your behavior and labels the song something you like. The app will recommend similar songs in the future.

However, if you fast-forward within the first 30 seconds of a song, the algorithm labels that behavior as a “thumbs-down” for that particular song and artist.

3. Collaborative filtering

If you’ve shopped from Amazon before, you’ve likely seen their recommendations. The ones that say “customers who bought this item also bought…”

That is collaborative filtering. It is a method of making automatic predictions (filtering) about the interests of a user by collecting preferences from many users (collaborating).

In simpler terms, collaborative filtering allows Spotify to collect data on what people like and don’t like and then use that information to make recommendations.

It’s like if you have a friend who really likes a certain type of song, and you trust their opinion. So if they recommend another song to you that’s similar to the one they like, you might like it too. 

Collaborative filtering is like having lots of friends who can give you recommendations on things you might enjoy. And that is one of the ways Spotify recommends songs to you.

4. Natural Language Processing

NLP stands for Natural Language Processing. Natural language refers to the way we humans communicate with each other using language, like when we speak or write.

So, NLP is a branch of artificial intelligence (AI) that focuses on teaching Spotify how to understand, interpret, and generate human language.

For example, NLP can help Spotify analyze a review of a song by scouring through social media, blog posts, and articles to determine whether the review is positive or negative. 

It also helps that Spotify has acquired Echo Nest which is a music intelligence company.

Together, they are able to identify what people are talking about in each song and make recommendations based on songs that have similar terms through NLP.

Final words

Spotify recommends songs based on other people’s playlists and your own taste profile. The algorithms bring them all together. That’s why their recommendations hit the mark.

Every Monday morning, Spotify combines data from 2 billion playlists and your personal taste profile, then does a little magic filtering and delivers your Discover Weekly playlist.

Related articles:

Scroll to Top