How do Spotify recommendations work?
At Spotify, we aim to create great and unique experiences for each user. Our goal is to connect everyone with what they love and help them discover something new. No two listeners are the same, so everyone’s Spotify experience, and many of our recommendations, is personalised. When asked what they like about Spotify, most listeners cite our personalisation as their top feature. You might wonder how we generate these recommendations across the Home feed, playlists, search results or other parts of the service, and we want to help demystify how they work.
At Spotify, people and technology work together to deliver relevant recommendations. Some recommendations are based on editorial curation, like a pop playlist created by music editors. Other recommendations are tailored to each listener’s unique taste, like a personalised playlist powered by our expert-designed algorithms.
We believe recommendations shouldn’t just optimise for the next click, but evolve with your taste. We have dedicated teams in place making sure your recommendations enable real engagement and cultivate meaningful connections. We’re always working on improving our recommendation systems to make sure we’re showing you relevant and enjoyable content.
Editorial curation
Editors at Spotify use data insights, sharp ears and an understanding of cultural trends to place content where it’s likely to resonate most with fans around the world. They employ thoughtful curation to recommend content on Spotify, like in editorial playlists. Around the world, editors at Spotify have extensive knowledge about local music and culture, which allows them to make their programming decisions with the best listener experiences in mind.
Personalised recommendations
Spotify offers algorithmic recommendations that are relevant, unique and specific to each user. Our algorithms select and order content across each listener’s Spotify experience, including in Search, Home and in personalised playlists.
To make these recommendations, our algorithms rely on a number of inputs. The importance of these inputs may vary over time, depending on your individual use of Spotify. We believe that your taste profile is the most important input for creating the best overall user experience. Below, you’ll find more details on the most significant inputs and how they work.
Your “taste profile”
As you engage with Spotify, actions such as searching, listening, skipping or saving to Your Library influence our interpretation of your taste. We call this your “taste profile”, and it gives our algorithms an indication of what you’re interested in and how you like to listen.
- Example: If you listen to a certain artist, we might recommend more songs by that artist.
- Example: Our Release Radar playlist recommends the latest releases we think you’ll enjoy based on similar music you’ve listened to.
- Example: If you listen to a sports podcast, we might recommend other sports podcasts to you.
Trends
Our algorithms consider trends and the general tastes and behaviours of users to improve recommendations for everyone. When users take similar actions on certain content, our algorithms have more information about how people might want to interact with it.
- Example: If many users interact with a specific search result (e.g. a song or a podcast), we are likely to recommend that result to other users who search for similar things.
Information about the content
Our algorithms take into account the characteristics of the content itself, such as its genre, release date, podcast category, etc. This allows us to identify which content has similar characteristics and might be enjoyed by similar listeners.
- Example: If you listen to a lot of pop music, we might recommend other pop songs that are similar.
- Example: if you listen to a lot of crime audiobooks, we might recommend other crime audiobooks.
Listener safety
As a platform, we evaluate the impact that we have on creators, listeners and communities. Spotify works to ensure that appropriate safety measures and processes are in place, including measures to prevent exposure to harmful content. We take algorithmic responsibility seriously and collaborate between policy, product and research teams, as well as consult with external experts such as the Spotify Safety Advisory Council.
Spotify’s Platform Rules apply to all content on the platform, including recommended content. These rules were developed by internal teams with input from a wide range of outside experts. When we become aware of potentially violating content, that content is reviewed against our policies, and the appropriate action will be taken. These actions include, for example, restricting the violating content from being recommended.
How can you influence what your recommendations are based on?
Your recommendations are constantly influenced by your engagement with content on Spotify. The more you listen to content you like and the more you interact with the app, the more we think you’ll enjoy your recommendations.
We also offer ways for you to influence and give feedback on what shows up in your recommendations and see less of something specific. Some examples are listed below:
- Exclude from taste profile: When you exclude a playlist from your taste profile, that playlist will have less influence on your future recommendations.
- Giving feedback on recommendations: When you tap [not interested/thumbs down] for a recommendation on Spotify, you’ll be given fewer recommendations that are similar to it.
- Explicit content filter: When you turn off explicit content, anything with an explicit tag will be greyed out and you won’t be able to play it.
In some cases, you can also organise and filter your recommendations based on what you most want to see. For example, you can filter your Home page to only see podcasts, or only see music.
How do commercial considerations impact recommendations?
Spotify prioritises listener satisfaction when recommending content. In some cases, commercial considerations, such as the cost of content or whether we can monetise it, may influence our recommendations. For example, Discovery Mode gives artists and labels the opportunity to identify songs that are a priority for them, and our system will add that signal to the algorithms that determine the content of personalised listening sessions. When an artist or label turns on Discovery Mode for a song, Spotify charges a commission on streams of that song in areas of the platform where Discovery Mode is active (Discovery Mode is not active in our editorial playlists). This signal increases the likelihood of the selected songs being recommended, but does not guarantee it. We only recommend songs there is a high probability listeners will enjoy. As with all recommendations, we take note when a listener isn’t engaging with a song – including those in Discovery Mode – and factor this in when determining what to recommend in the future.
