Adding Transparency and Control to Spotify Discovery Playlists
A Discovery Trust Layer that explains why a song was recommended and lets listeners tune discovery
✳The problem they went after
Spotify's most musically curious listeners are the ones drifting away from discovery playlists. They cannot see why a song was recommended and have no way to influence the algorithm on purpose, so they skip heavily, rarely finish a playlist, and regress to familiar music. That erodes the personalisation advantage among exactly the users it was built for.
✳What they actually did
Samuel's Discovery Trust Layer has three parts that work together. A why this song explainability layer puts a tap-to-reveal modal beside each track showing two to three human-readable recommendation signals, with thumbs-up or thumbs-down feedback attached. A discovery mode toggle slides between safe and explore, adjusting how much novelty the next playlist refresh carries. A guided tuning flow lets listeners name genres they want more of, artists they are tired of, and their preferred discovery intensity.
✳What you can take from it
Worth stealing
- Explainability paired with a control for every signal
- Safe-to-explore toggle changes novelty weighting
- Tuning flow captures tired-of artists explicitly
- A/B on a 10 percent cohort with four-week follow-up
Read this if
PMs working on recommendation systems where users distrust the algorithm
The transferable bitExplainability is paired with control, which is what lifts it above a UI flourish. Every explanation comes with a way to act on it, and the feedback loops back in. The validation plan is equally concrete: a twenty percent lift in discovery playlist completion within three months, tested on a ten percent cohort over four weeks with a four-week follow-up.
✳The submission itself
Open the PDF →Reading on a phone? The embedded viewer is cramped — open the PDF in a new tab instead.
