Within Algorithms
Can recommendations stay relevant and diverse?
Spotify's own research shows the tension between immediate relevance and broader, more varied listening over time.
On this page
- What Spotify's diversity research found
- Why similarity can reduce listening variety
- Why lower popularity recommendations may be easier than bigger taste jumps
Page outline Jump by section
Introduction
Spotify’s recommendation research points to a central challenge in music discovery: the songs most likely to earn an immediate play are not always the songs that broaden a listener’s musical world. The platform’s own studies suggest that recommendation systems face a persistent tradeoff between relevance and diversity. Recommendations that closely match existing tastes often maximise short-term engagement, but they can also narrow listening patterns over time. Conversely, introducing more varied or unexpected music can expand a listener’s horizons, yet carries a greater risk that the recommendation will be ignored. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
This tension matters because recommendation systems have become one of the main ways people encounter new music. The question is no longer simply whether algorithms help discovery. It is whether they can help listeners discover music that is both appealing now and beneficial for long-term exploration.
What Spotify’s diversity research found
One of the most influential pieces of evidence comes from Spotify researchers’ large-scale study, Algorithmic Effects on the Diversity of Consumption on Spotify. The researchers examined millions of listening events and developed measures of musical diversity based on how varied a user’s listening behaviour was across songs and styles. They found that higher consumption diversity was strongly associated with desirable long-term outcomes, including user retention and conversion. In other words, listeners who explored a broader range of music tended to be healthier long-term users of the platform. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
However, the same study also found that listening driven by algorithmic recommendations was generally associated with lower diversity than listening driven by other forms of exploration. Users who relied more heavily on recommendations tended to consume music that was more similar to what they already knew, while users who expanded their diversity often did so through behaviours less dependent on recommendation surfaces. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th… [2U of T Computer Science]cs.toronto.edualg effects spotify www2020U of T Computer ScienceAlgorithmic Effects on the Diversity of Consumption on Spotifyby A Anderson · 2020 · Cited by 376 — To investigate…
The finding is important because it challenges a common assumption that recommendation automatically equals discovery. Spotify’s research suggests that recommendation can increase discovery in the sense of introducing unfamiliar tracks, while still reducing diversity if those tracks remain highly similar to a listener’s existing preferences. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
A related Spotify field experiment examining podcast recommendations revealed a similar pattern. Personalised recommendations increased consumption substantially, but they also reduced individual-level diversity. The researchers described this as an “engagement-diversity trade-off”: optimising recommendations solely for engagement can increase usage while simultaneously narrowing the range of content consumed. [arXiv]arxiv.orgarXivThe Engagement-Diversity Connection: Evidence from a Field Experiment on SpotifyMarch 17, 2020…
Why similarity can reduce listening variety
The relevance side of the tradeoff comes from how recommendation systems succeed. A recommendation is usually judged successful when a listener streams, saves or returns to the suggested content. The safest way to achieve that outcome is often to recommend something highly similar to what the listener already enjoys. [Spotify]spotify.comunderstanding recommendationsSpotifyUnderstanding recommendations on SpotifyMar 12, 2026 — At Spotify, people and technology work together to deliver relevant recomme…
Imagine a listener who frequently plays contemporary indie rock. A recommendation system seeking maximum immediate relevance is likely to suggest more contemporary indie rock, perhaps from a different artist but with comparable characteristics. Such recommendations often perform well because they fit established preferences. Yet repeated use of this strategy can create a gradual narrowing effect. The listener may encounter many new songs, but from within a relatively small musical neighbourhood. [HEC Digital Marketing]digital.hec.caHEC Digital MarketingHow Spotify's Algorithm Shapes Global Music Discovery…Nov 9, 2025 — Spotify's biased recommendation system is und…
Researchers often describe this as a filtering or reinforcement mechanism. The algorithm learns from past behaviour, uses that behaviour to predict future preferences, and then feeds those predictions back into the listening experience. Each successful recommendation provides additional evidence that the listener likes similar material, encouraging further recommendations along the same path. [GOV.UK]GOV.UKThe impact of algorithmically driven recommendation…by D Hesmondhalgh · Cited by 60 — The impact of streaming platforms on musical pro…
This does not necessarily mean listeners become trapped in rigid “filter bubbles”. The empirical evidence on that question is mixed. However, Spotify’s own findings indicate that recommendation-driven listening tends to be less diverse than other forms of exploration, suggesting that similarity-based optimisation has measurable effects on listening breadth. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
Why lower-popularity recommendations may be easier than bigger taste jumps
An important insight from recommendation research is that not all forms of diversity are equally difficult to introduce.
Recommending a less popular artist who sounds similar to music a listener already enjoys is often a relatively small step. The recommendation remains highly relevant because it sits close to existing preferences. From the system’s perspective, this kind of recommendation carries limited risk. The listener may still stream the track even if the artist is unfamiliar. [arXiv]arxiv.orgarXivThe Engagement-Diversity Connection: Evidence from a Field Experiment on SpotifyMarch 17, 2020…
A much larger challenge arises when diversity requires crossing stylistic, cultural or genre boundaries. Moving an indie-rock listener toward experimental electronic music, jazz, folk traditions from another country or an entirely different musical culture involves a larger leap in taste. Such recommendations may contribute more to genuine diversification, but they are also less predictable and more likely to be skipped. [Music Information Retrieval Transactions]transactions.ismir.netMusic Information Retrieval TransactionsDiversity by Design in Music Recommender Systemsby L Porcaro · 2021 · Cited by 32 — In this overv…
This distinction helps explain why popularity diversity and taste diversity are not the same thing:
- Popularity diversity introduces artists outside the mainstream while remaining close to existing tastes.
- Taste diversity introduces music that differs substantially from a listener’s established listening patterns.
- The second form generally carries greater engagement risk because it asks listeners to venture further from familiar territory. [arXiv]arxiv.orgarXivThe Engagement-Diversity Connection: Evidence from a Field Experiment on SpotifyMarch 17, 2020…
As a result, platforms may find it easier to diversify recommendations through lower-popularity content than through radical shifts in musical style. A recommendation can be novel in terms of artist exposure while remaining highly relevant in terms of sound.
What the tradeoff means for music discovery
Spotify’s diversity studies suggest that the central challenge is not choosing between relevance and diversity, but balancing them over time. Immediate engagement metrics naturally favour recommendations that closely match known preferences. Yet the same research indicates that broader listening diversity is associated with valuable long-term outcomes for users. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
The implication is that successful music discovery systems may need to optimise for more than the next click or stream. They may need to consider whether recommendations help listeners develop richer listening habits over weeks, months and years. Research beyond Spotify has found that exposure to appropriately diversified music recommendations can increase curiosity, openness and willingness to engage with unfamiliar music, suggesting that diversity itself can create long-term value when introduced carefully. [arXiv]arxiv.orgarXivThe Engagement-Diversity Connection: Evidence from a Field Experiment on SpotifyMarch 17, 2020…
Spotify’s work therefore highlights a nuanced reality. Recommendations that are maximally relevant are not always maximally diverse, and recommendations that maximise diversity are not always immediately relevant. The most effective discovery systems are likely to sit between those extremes, using familiar music as a bridge toward broader exploration rather than treating relevance and diversity as mutually exclusive goals. [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th… [Spotify Research]research.atspotify.comSpotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th…
Amazon book picks
Further Reading
Books and field guides related to Can recommendations stay relevant and diverse?. Use these as the next step if you want deeper reading beyond the article.
The Filter Bubble
First published 2011. Subjects: World Wide Web, Invisible Web, Information organization, Social aspects, Subject access.
eBay marketplace picks
Marketplace Samples
Example marketplace items related to this page. Use the search link to explore similar finds on eBay.
Endnotes
-
Source: arxiv.org
Link: https://arxiv.org/abs/2003.08203Source snippet
arXivThe Engagement-Diversity Connection: Evidence from a Field Experiment on SpotifyMarch 17, 2020...
Published: March 17, 2020
-
Source: spotify.com
Title: understanding recommendations
Link: https://www.spotify.com/safetyandprivacy/understanding-recommendationsSource snippet
SpotifyUnderstanding recommendations on SpotifyMar 12, 2026 — At Spotify, people and technology work together to deliver relevant recomme...
-
Source: digital.hec.ca
Link: https://digital.hec.ca/en/blog/how-spotifys-algorithm-shapes-global-music-discovery-and-cultural-diversity/Source snippet
HEC Digital MarketingHow Spotify's Algorithm Shapes Global Music Discovery...Nov 9, 2025 — Spotify's biased recommendation system is und...
-
Source: GOV.UK
Link: https://www.gov.uk/government/publications/research-into-the-impact-of-streaming-services-algorithms-on-music-consumption/the-impact-of-algorithmically-driven-recommendation-systems-on-music-consumption-and-production-a-literature-reviewSource snippet
The impact of algorithmically driven recommendation...by D Hesmondhalgh · Cited by 60 — The impact of streaming platforms on musical pro...
-
Source: arxiv.org
Link: https://arxiv.org/abs/2208.09517Source snippet
arXivExploring Popularity Bias in [Music Recommendation]({{ 'algorithms-a7ef46/' | relative_url }}) Models and Commercial Steaming ServicesAugust 19, 2022...
Published: August 19, 2022
-
Source: arxiv.org
Link: https://arxiv.org/abs/2212.00592 -
Source: arxiv.org
Link: https://arxiv.org/abs/1912.04696Source snippet
arXivThe Unfairness of Popularity Bias in Music Recommendation: A Reproducibility StudyDecember 10, 2019...
Published: December 10, 2019
-
Source: GOV.UK
Title: the impact of recommendation algorithms on the uks music industry
Link: https://www.gov.uk/government/publications/research-into-the-impact-of-streaming-services-algorithms-on-music-consumption/the-impact-of-recommendation-algorithms-on-the-uks-music-industrySource snippet
impact of recommendation algorithms on the UK's...9 Feb 2023 — Recommendation 18 called for “research into the impact of streaming servi...
-
Source: youtube.com
Title: Spotify ML Question
Link: http://www.youtube.com/watch?v=vyZMYlGBSBMSource snippet
The Math Behind Recommender Systems...
-
Source: youtube.com
Link: http://www.youtube.com/watch?v=3lxo5hQ_3JYSource snippet
Spotify recommendation systems diversity tradeoff Spotify ML Question - Design a Recommendation System (Full mock interview) Exponent...
-
Source: research.atspotify.com
Link: https://research.atspotify.com/algorithmic-effects-on-the-diversity-of-consumption-on-spotifySource snippet
Spotify ResearchAlgorithmic Effects on the Diversity of Consumption on SpotifyDec 3, 2020 — In this work, we analyze our users through th...
-
Source: research.atspotify.com
Title: algorithmic effects on the diversity of consumption on spotify
Link: https://research.atspotify.com/publications/algorithmic-effects-on-the-diversity-of-consumption-on-spotifySource snippet
Spotify ResearchAlgorithmic Effects on the Diversity of Consumption on Spotify1 Apr 2020 — However, we also find that algorithmically-dri...
-
Source: cs.toronto.edu
Title: alg effects spotify www2020
Link: https://www.cs.toronto.edu/~ashton/pubs/alg-effects-spotify-www2020.pdfSource snippet
U of T Computer ScienceAlgorithmic Effects on the Diversity of Consumption on Spotifyby A Anderson · 2020 · Cited by 376 — To investigate...
-
Source: research.atspotify.com
Link: https://research.atspotify.com/publications/the-engagement-diversity-connection-evidence-from-a-field-experiment-on-spotifySource snippet
Spotify ResearchThe Engagement-Diversity Connection: Evidence from a Field...We present results from a randomized field experiment on Sp...
-
Source: transactions.ismir.net
Link: https://transactions.ismir.net/articles/10.5334/tismir.106Source snippet
Music Information Retrieval TransactionsDiversity by Design in Music Recommender Systemsby L Porcaro · 2021 · Cited by 32 — In this overv...
-
Source: Wikipedia
Link: https://en.wikipedia.org/wiki/MusicSource snippet
MusicMusic is the arrangement of sound to create some combination of form, [harmony]({{ 'harmony/' | relative_url }}), [melody]({{ 'melody/' | relative_url }}), rhythm, or otherwise expressive content...
-
Source: research.atspotify.com
Link: https://research.atspotify.com/publications?category=evaluationSource snippet
| Spotify ResearchThe Engagement-Diversity Connection: Evidence from a Field Experiment on Spotify. David Holtz, Benjamin Carterette, Pra...
-
Source: ide.mit.edu
Title: SSRN id3555927
Link: https://ide.mit.edu/wp-content/uploads/2020/03/SSRN-id3555927.pdfSource snippet
We present results from a randomized...
-
Source: upcommons.upc.edu
Link: https://upcommons.upc.edu/server/api/core/bitstreams/75b3d25c-1740-4a43-8651-d2525849e19b/contentSource snippet
BIAS IN MUSIC RECOMMENDER...by DG GONZALEZ — This thesis aims to further explore and expand upon the existing knowledge in biases presen...
Additional References
-
Source: researchgate.net
Link: https://www.researchgate.net/publication/362860235_Exploring_Popularity_Bias_in_Music_Recommendation_Models_and_Commercial_Steaming_ServicesSource snippet
(PDF) Exploring Popularity Bias in Music Recommendation...19 Aug 2022 — In this paper, we attempt to measure popularity bias in three st...
-
Source: studocu.com
Link: https://www.studocu.com/in/document/srm-valliammai-engineering-college/computer-science-and-engineering/33664233380281-asdfjgakshdjfoiaj/87436891Source snippet
Algorithmic Impact on Music Consumption DiversityIn this work, we study the user ex- perience on Spotify, a popular music streaming servi...
-
Source: dl.acm.org
Link: https://dl.acm.org/doi/10.1145/3608487Source snippet
ACM Digital LibraryAssessing the Impact of Music Recommendation Diversity...We show that exposure to specific levels of music recommenda...
-
Source: researchgate.net
Title: 388827947 Effects of algorithmic curation in users’ music taste on Spotify
Link: https://www.researchgate.net/publication/388827947_Effects_of_algorithmic_curation_in_users%27_music_taste_on_SpotifySource snippet
Effects of algorithmic curation in users' music taste on Spotify4 May 2026 — A comprehensive review of the literature reveals that the pr...
Published: May 2026
-
Source: amt-lab.org
Title: recommendation algorithms are pervasive now they need to diversify
Link: https://amt-lab.org/blog/2022/4/recommendation-algorithms-are-pervasive-now-they-need-to-diversifySource snippet
The Necessary Diversification of Pervasive...5 May 2022 — This article covers the recommendation algorithms built to support content del...
Published: May 2022
-
Source: beckyxzhang.github.io
Link: https://beckyxzhang.github.io/paperSource snippet
Assessing the diversity-accuracy dilemma of Spotify's...May 13, 2021 — In this paper, we explore how different algorithms impact accurac...
Published: May 13, 2021
-
Source: illumin.usc.edu
Title: algorithmic symphonies how spotify strikes the right chord
Link: https://illumin.usc.edu/algorithmic-symphonies-how-spotify-strikes-the-right-chord/Source snippet
Symphonies: How Spotify Strikes the Right Chord21 Jan 2024 — This article explores Spotify's recommendation algorithm, including how it h...
-
Source: music-tomorrow.com
Title: fairness and diversity in music recommendation algorithms
Link: https://www.music-tomorrow.com/blog/fairness-and-diversity-in-music-recommendation-algorithmsSource snippet
Are music recommendation algorithms fair to emerging...Sep 21, 2021 — In today's piece, we'll go through the main elements of this probl...
-
Source: music-tomorrow.com
Title: how spotify recommendation system works complete guide
Link: https://www.music-tomorrow.com/blog/how-spotify-recommendation-system-works-complete-guideSource snippet
Inside Spotify's Recommendation System: A Complete...1 Sept 2025 — As we move ahead into the 2020s, an ever‑increasing share of music co...
-
Source: apps.apple.com
Link: https://apps.apple.com/us/app/spotify-music-and-podcasts/id324684580Source snippet
apple.comSpotify: Music and Podcasts - App Store - AppleWith the Spotify app, you can explore an extensive library of music and podcasts...
Topic Tree