How the Spotify Algorithm Works (and How to Work With It)

The three ways Spotify surfaces music

Most “grow on Spotify” advice treats Spotify as one system. It’s three. Understanding which bucket you’re dealing with changes what you should do.

Editorial playlists are curated by Spotify’s human editors: New Music Friday, RapCaviar, genre and mood lists. These are the ones you pitch for. You submit an unreleased track through Spotify for Artists before release, an editor considers it, and it either gets added or it doesn’t. There is one door you can knock on, and this is it.

Algorithmic and personalized playlists and sessions are generated per listener from behaviour. Discover Weekly, Release Radar, Daily Mix, On Repeat, Radio, and Autoplay all fall here. You can’t pitch your way onto any of them. What you can do is earn them by getting real listeners to engage with the song in ways the system reads as genuine interest.

Listener playlists are made by users and other artists. A real person adds your song to their playlist, that playlist has listeners of its own, and some of those people discover you through it. This matters because a real user playlist add is one of the signals vendor research points to as meaningful for Discover Weekly eligibility. More on that below.

The signals the system reacts to

Spotify has said its recommendations are driven by how people engage with a track. The engagement signals it and its editors reference include intentional plays (someone searching for or clicking the song, as opposed to passive background radio), saves to library, adds to personal playlists, repeat listens, completion versus early skips, shares, and follows of the artist.

The 30-second mark matters in two ways. It’s when a stream counts for royalties, and a skip before 30 seconds is the clearest negative signal: the system reads it as that listener not wanting that song in that context. Which means a song that people skip quickly is actively working against itself in the algorithm, regardless of how many total streams it accumulates.

Save rate and skip rate in the first window after a track reaches new listeners are the levers that come up most in vendor research. Spotify doesn’t publish specific thresholds or weights for these signals, and any number you see cited (a “20% save rate target” for instance) is an external estimate, not a Spotify setting. Treat the direction as right and the specifics as a guess.

Release Radar: the follower-based pathway

Release Radar refreshes every Friday with a personalized playlist of new music for each listener, built primarily from artists they follow. A new track stays eligible for up to 4 weeks after release, and each listener gets at most one song per artist per week.

This is the algorithmic surface you have the most direct influence over, because followers are something you can build. Pitch your song at least 7 days before release through Spotify for Artists to control which track from your release lands in your followers’ Radar. If you don’t pitch, Spotify picks. New followers you gain inside the 4-week window can still receive the song while it’s eligible.

For the full mechanics, including what happens if you miss the pitch cutoff and how to build the follower base that makes Release Radar worth something, see how Release Radar works.

Discover Weekly: the taste-based playlist

Discover Weekly refreshes every Monday with roughly 30 tracks per listener, chosen through collaborative filtering: what you listen to combined with what listeners with similar taste listen to. It surfaces music the listener probably hasn’t heard before.

Following an artist does not put you in their Discover Weekly. That’s Release Radar’s job. Discover Weekly is about taste overlap, and the signals that vendor research points to most are real-user saves, playlist adds by consistent listeners, and completion. There is no submit button. The honest limit here is that Spotify doesn’t document a checklist for it. What works is getting real listeners to engage, and letting the system find the overlap from there.

The full mechanics, the sourcing distinctions between what Spotify confirms and what vendor analysis estimates, and what you can realistically do to improve your odds are in how to get on Discover Weekly.

How editorial and algorithmic connect

The two systems are separate, but they feed each other. Practitioners describe it this way: an editorial placement or a real promotional push exposes the track to new listeners. If those listeners save it, complete it, and come back to it, that engagement strengthens the behavioural signals that personalized systems read. After the editorial placement rotates off, that accumulated signal can carry the track into Discover Weekly, Radio, and Autoplay.

Spotify hasn’t documented this as a formal pipeline. It’s how practitioners describe momentum based on what they observe. The practical implication is worth taking seriously: a pitch that lands on an editorial playlist is valuable beyond the placement itself, because engagement from new listeners can prime the algorithmic systems. An editorial add that people skip does little. What converts matters more than the add alone.

build the pitch that gets your song in front of editors with the free Spotify pitch generator →

Discovery Mode: the paid-ish lever

Discovery Mode lets you flag specific tracks for priority placement in Radio and Autoplay sessions in exchange for a reduced royalty on the streams that come from those placements. No upfront fee. The cost is the reduced per-stream pay on the boosted streams, with a commission widely reported as about 30% (not a published Spotify number; confirm your exact terms with your distributor).

It doesn’t touch editorial playlists, Release Radar, or Discover Weekly. It’s a lean-back, radio-context tool. Whether it makes sense for an independent artist depends on your goals and your royalty math. For the full breakdown and the payola criticism, the Discovery Mode guide covers it.

Why streams drop after week one, and what sustains them

A spike in week one and a drop in week two is normal. It’s not a failure. What’s expiring: the 4-week Release Radar eligibility window eventually closes, editorial rotations cycle weekly, and the concentrated day-one push from your own audience is spent. What sustains streams past that window is the song converting new listeners into saves and follows so the personalized systems keep serving it, and a steady release cadence that feeds Release Radar with new material.

A scoped Spotify Fan Study reported that about 75% of a release’s first-year streams happen after the first month, from a sample of February 2024 tracks with at least 1,000 annual streams by artists with at least 1,000 monthly listeners. The sample caveat matters. But the direction is right: week one is not the verdict. For everything that drives the decay and what you can do about it, see why streams drop after release week.

If you’re planning your release around these windows, the per-stream economics are in how Spotify royalties work. And if you’re building toward a first release, the basics of Release Radar and follower strategy are in getting to your first 1,000 streams.

pitch the song that starts the engagement loop →