How the Spotify algorithm picks songs: radar, saves and signals

Spotify's algorithmic playlists are not a lottery and not a queue. They are a recommendation system that reads listener behaviour and repeats what works. Nobody outside Spotify has the full model, and the weights change, but the signals it reads are visible in your own data and in public platform documentation. You cannot pitch Discover Weekly. You can give it a reason to exist.

The one exception is Release Radar. That is the single algorithmic surface you can lock in with a button: pitch your song through Spotify for Artists at least 7 days before release and it goes into your followers' Release Radar (Spotify: Release Radar).

How the system decides

Spotify's personalized playlists are built from patterns of listening: who plays a song, whether they finish it, whether they skip it, whether they save or add it, and who else listens to similar music. The system compares your song's listeners with listeners of other songs and looks for overlap. The more a song earns strong reactions from people who already like your genre, the further it can travel.

That is why buying streams fails on the product level before it fails on the ethics level. Bot accounts produce plays with no listening behaviour attached: no saves, no repeat plays, no follows, no playlist adds. They feed the system noise and can push the song toward the wrong audience. Spotify removes artificial streams from public numbers and withholds the royalties (Spotify for Artists: artificial streaming).

The signals that matter

Save rate, not stream count, is the number artists talk about most, and the one public number with a documented link to editorial outcomes. MusicPulse's analysis cites a Chartmetric finding that artists with a save rate above 8% on a recent release were 2.7 times more likely to earn an editorial placement on their next submission (MusicPulse, 5 March 2026). Treat it as a correlation to test on your own releases, not a switch to flip.

Release Radar: the surface you control

Release Radar is a weekly personal playlist of new music by artists a listener follows. Two facts make it the most exploitable algorithmic playlist on Spotify:

• The 7-day rule. Pitch your chosen song at least 7 days before release and Spotify adds it to your followers' Release Radar that week. Miss the window and Spotify may choose a different song from the release, or none at all.
• It is not an editorial decision. No editor approves the add. This is delivery plumbing, and it runs on schedule.

So the smartest algorithmic move is also the simplest: send the pitch early, then tell your existing fans the song is coming. Our guide to the Release Radar pitch benefit covers what followers see and why the date is so unforgiving.

Discover Weekly and the mixes

Discover Weekly, Daily Mix, Radio and Autoplay are earned, not requested. There is no submission form, no distributor route, and no legitimate service that can place a song inside them. What you can do is create the conditions:

1. Deliver clean metadata. Correct genre, mood and language tags put the song in front of the right seed listeners in the first place.
2. Seed with real listeners. A small, engaged audience that finishes and saves the track is stronger evidence than a large passive one.
3. Keep the release alive. Algorithmic playlists refresh, and a song that performs can re-enter months later. Do not stop promoting on day 7.

Our playlist types guide explains where each surface sits, and Soundcharts is one way to watch playlist and chart movement over time if you want third-party data next to Spotify's own.

Why the first 72 hours matter

Editorial teams and the recommendation system both look at early performance. Artist Advisory's pitch guide tells artists to watch saves, skips and playlist adds across days 1 to 28, with the first 72 hours carrying the most weight for late adds and follow-on playlist decisions (15 April 2026).

Practically, that means release week is not a victory lap. It is the moment to ask your existing listeners to save the song, to send the live link to the curators who asked for a reminder, and to watch your source of streams to see whether the plays are coming from real people (Spotify for Artists).

A three-step test for every release

1. Lock Release Radar: pitch at day -7 or earlier, with every field filled in.
2. Seed behaviour: get saves and full listens from real fans in the first 72 hours through your own channels and honest curator placements.
3. Read the split: at day 7 and day 28, compare active and programmed streams. If programmed plays bring no saves, the playlist was the wrong audience, not a stepping stone.

Five mistakes that starve the algorithm

1. Chasing streams instead of saves. A hundred real saves beat a thousand bot plays.
2. Sending the pitch late. The Release Radar guarantee is lost, and editors have less time.
3. Sloppy genre tags. Wrong tags route the song to the wrong listeners and produce skips, which teach the system to stop.
4. Stopping after week one. Late algorithmic pickups are real; keep promoting for at least 28 days.
5. Buying "algorithmic placement". No service can sell a place in Discover Weekly. Offers that promise one are selling streams, and Spotify's terms forbid it.