Similar Songs Finder

How to Find Songs Similar to Your Favorites: A Data-Driven Guide

Last.fm has tracked over 100 billion individual scrobbles since its founding in 2002. Each scrobble is a timestamped listening event. When millions of people listen to a track, then immediately or regularly listen to another, those two tracks develop a statistical relationship. That relationship is what powers this tool.

What the Match Score Actually Measures

The Last.fm track.getSimilar endpoint returns a match value between 0 and 1 for each result. We display this as a percentage. A score of 0.85 means 85% match.

The score is not a direct measure of musical similarity. It does not compare keys, tempos, or chord progressions. Instead, it measures listener overlap: how often people who listen to track A also listen to track B, weighted by listening frequency and recency. This approach is called collaborative filtering. It catches non-obvious connections that purely audio-based systems miss.

A practical example: "Hotel California" by Eagles and "Stairway to Heaven" by Led Zeppelin share many listeners. Even though they differ in structure, they cluster together in listener behavior. Both sit in the "classic rock deep listen" category for a large segment of Last.fm users.

Why Obscure Tracks Return Fewer Results

Last.fm needs a minimum amount of listening data to calculate reliable similarity. A track with 200 total scrobbles does not have enough data points to build a meaningful graph. If you search for a niche artist and get zero results, this is normal. The algorithm requires enough co-listening events to reach statistical confidence.

For tracks with thin data, try the seed artist instead. Our Similar Artists Finder uses artist-level data, which aggregates scrobbles across all of an artist's tracks. Artist-level data is more robust for smaller acts.

Using the Match Score Filter

Setting the minimum match score to 0% returns all results, sorted from highest to lowest. Setting it to 50% restricts to genuinely similar tracks. The default of 10% is a practical middle ground. It removes near-zero matches (results that barely correlate) while still giving a wide selection.

For genre exploration, keep the threshold low. For building a focused playlist around a specific sound, push it to 40% or higher. High-scoring results tend to be tracks that co-appear in user-created playlists or are frequently listened to back-to-back.

Real-World Discovery Examples

Search for "Paranoid Android" by Radiohead and you will find Thom Yorke's other projects, Portishead, and Massive Attack in the results. These are not obvious genre matches, but they reflect how Radiohead fans actually listen. That is the kind of cross-genre discovery that genre tags alone cannot surface.

Search for "Running Up That Hill" by Kate Bush (particularly after its 2022 resurgence) and you will find a mix of 1980s synth-pop and modern artists who draw from that era. The algorithm picked up new listeners from younger demographics who then explored similar sounds.

If you want to go deeper into discovery by artist rather than by track, use the Similar Artists Finder to get a grid of related acts with listener counts. For exploring by genre tag, try our Genre Tag Explorer.

How Last.fm Builds Its Similarity Graph

Last.fm uses a combination of explicit user data and implicit listening signals. Explicit signals include user tags, love/hate ratings, and playlist creation. Implicit signals are the raw scrobble stream: what you play, how often, in what sequence.

The system builds a co-occurrence matrix: for every pair of tracks, how many users have scrobbled both? This matrix is enormous. Last.fm indexes data from over 300 million registered users and nearly 10 million unique artists. The resulting similarity graph covers a wide range of genres and time periods.

According to Last.fm's API documentation, the similarity scores are calculated from listening patterns and updated regularly. The API is free to use and returns up to 100 similar tracks per query (this tool caps at 30 for clarity).

Practical Use Cases

Building playlists: Start with a seed track and add the top 10 high-match results. Repeat with one of those as the new seed to chain discoveries.

A&R and sync research: Find tracks that occupy the same listener space as a reference track. This is useful when searching for sync placements with a similar emotional profile.

DJ set planning: High match scores indicate tracks that the same audience enjoys. They tend to work well in sequence. This is not a BPM or key-matching tool, but it surfaces tracks that feel contextually related.

Expanding your library: If you have listened to a track hundreds of times, find similar ones you have never heard. Sort by highest match and work down until you find an unfamiliar artist.