White Paper 3 of the Beyond the Playlist Series
Abstract
Spotify is the dominant streaming platform by subscriber count and cultural visibility, but it is not the only architecture through which listeners engage with the recorded music catalog in the streaming era. Apple Music, Tidal, Amazon Music, and YouTube Music each represent distinct institutional approaches to the organization, curation, and recommendation of music, shaped by the parent company’s broader business logic, its relationship to the music industry, its technical infrastructure, and its assumptions about who its users are and what they want from a music service. This paper examines each platform’s approach to music exploration and discovery, with particular attention to how their architectures differ from Spotify’s and from each other, where they succeed and fail as discovery infrastructure relative to the standard established in Papers 1 and 2, and what the comparative analysis reveals about the relationship between institutional logic and discovery design. The central finding is that no major streaming platform has developed a genuinely superior exploration architecture to Spotify’s, but each platform’s specific limitations and occasional strengths illuminate different aspects of the larger structural problem: that streaming platforms are designed and monetized in ways that make genuine musical exploration structurally subordinate to retention, comfort, and commercial promotion.
1. Introduction: Institutional Logic and Discovery Design
The comparison of streaming platforms is usually conducted on the terms the platforms themselves prefer: audio quality specifications, catalog size, pricing tiers, exclusive content, and interface design. These are real and sometimes meaningful differences, but they are not the most analytically useful frame for understanding how different platforms approach the discovery problem. A more revealing frame is institutional logic — the set of assumptions, priorities, business relationships, and structural incentives that shape a platform’s design from the inside out.
Each major streaming platform is not a free-standing music service but an extension of a larger corporate entity with its own institutional character. Spotify is a standalone streaming business whose survival depends on subscriber growth and retention. Apple Music is a feature within an ecosystem designed to sell hardware and retain users within Apple’s product environment. Tidal is a platform originally designed as a musician-owned alternative to the dominant model, subsequently acquired and repositioned, with a niche positioning around high-fidelity audio and artist compensation. Amazon Music is an extension of the world’s largest e-commerce and cloud infrastructure company, integrated with voice-assistant technology and subscription bundling. YouTube Music is the music streaming layer built on top of the world’s largest video platform, inheriting its recommendation engine and its creator economy logic.
These different institutional origins are not incidental to how each platform handles exploration. They are constitutive of it. The way a platform conceives of music — as a retained subscriber’s benefit, a hardware ecosystem feature, a fidelity statement, a bundled convenience, or a form of video-adjacent content — shapes what kinds of discovery tools it builds, what data it prioritizes, and what it considers a successful listening session. Examining these differences seriously requires setting aside the consumer-review frame of feature comparison and engaging with each platform’s discovery architecture as an expression of its institutional identity.
2. Apple Music: The Editorial Model and the Limits of Human Curation
2.1 Institutional Logic
Apple Music was launched in 2015 as a response to Spotify’s dominance and as an integration point within Apple’s broader ecosystem strategy. Its defining institutional characteristic is Apple’s relationship to creative industries as a self-conceived patron and curator rather than a neutral platform. Apple has consistently positioned itself as a company that understands and respects creative work — a positioning that dates to the original iPod era and the construction of iTunes as a legitimate alternative to music piracy. Apple Music inherits this positioning and expresses it primarily through its emphasis on human editorial curation as the platform’s differentiating feature.
Where Spotify’s discovery architecture is predominantly algorithmic, Apple Music’s is predominantly editorial. The platform employs a substantial team of music editors — many of them former music journalists, radio programmers, and industry professionals — who construct and maintain a large library of editorially curated playlists organized by genre, mood, activity, and cultural moment. These playlists are Apple Music’s primary discovery surface, and they are distinguished from Spotify’s editorial playlists by the genuine curatorial intent that shapes them, the depth of expertise some of them reflect, and the contextual annotation that accompanies them.
2.2 Editorial Depth and Its Limits
The editorial model has genuine strengths as a discovery approach that the algorithmic model, for all its scale, cannot match. A human editor with deep genre knowledge constructs a playlist according to musical understanding rather than behavioral inference — they know which artist influenced which, which record represents a turning point in a tradition, which contemporary artists are seriously engaged with their genre’s history and which are superficially appropriating its sonic surface. This knowledge produces playlists that have internal logic beyond sonic similarity and popularity proximity.
Apple Music’s genre-specific editorial playlists, particularly in jazz, classical, and folk traditions, frequently reflect this depth. They are more likely than their Spotify counterparts to include catalog material alongside recent releases, to situate contemporary artists in relation to historical traditions, and to function as implicit education in a genre rather than simply as a listening session soundtrack. For a listener who wants to develop genuine knowledge of a tradition rather than simply hear music that resembles music they already know, Apple Music’s editorial work is a meaningfully better starting point in these genre spaces.
The limitations of the editorial model, however, are equally real. Human curation does not scale the way algorithmic recommendation scales. A curatorial team of any realistic size can maintain genuine expertise across a limited range of genre territories; the catalog’s full depth and breadth — the regional scenes, the microgenres, the historical traditions that are not represented in any major music press — exceeds what a professional editorial operation can cover with genuine depth. Apple Music’s editorial strength is concentrated in the genre spaces where Western music journalism has traditionally focused: rock and its derivatives, jazz in its mainstream-accessible forms, hip-hop in its commercially visible iterations, and classical music at its most canonical. In the vast territories beyond this coverage — global music traditions, extreme genre niches, genuinely experimental work — Apple Music’s editorial playlists become thinner and less differentiated.
2.3 Apple Music Radio and Algorithmic Functions
Beyond its editorial playlists, Apple Music provides radio and recommendation functions that operate on broadly similar algorithmic principles to Spotify’s, drawing on listener behavioral data and audio feature analysis. Apple’s radio stations — including the flagship Beats 1 (now Apple Music 1) and its genre-specific stations — blend algorithmic generation with programmed content and live DJ-hosted shows that partially recover the radio model’s curatorial voice. This blended approach is one of Apple Music’s more interesting structural choices: it acknowledges that the live DJ format provides something that pure algorithmic generation cannot, while integrating that format within a streaming subscription context.
Apple Music 1 in particular represents an attempt to maintain something resembling a broadcast radio curatorial voice — a single channel with global reach, live programming, artist interviews, and editorial decisions made by human programmers — within the otherwise on-demand architecture of streaming. For discovery purposes, this creates a qualitatively different experience than Spotify’s radio functions: the listener who engages with Apple Music 1 is encountering someone else’s curation in real time, without the ability to skip or customize, and this loss of control is the source of its discovery value. The same accidental encounter dynamic that Paper 4 will examine in the context of broadcast radio is partially reproduced in the live streaming radio format.
2.4 The Ecosystem Constraint
Apple Music’s most significant structural limitation as a discovery platform is the ecosystem constraint that its institutional logic imposes. Because Apple Music is designed primarily as a feature that retains users within Apple’s product environment, its design priorities are shaped by the needs of the broader Apple experience rather than by the specific requirements of musical exploration. Interface decisions, feature development priorities, and the overall architecture of the platform are constrained by their integration with iOS, macOS, HomePod, and CarPlay in ways that sometimes work against exploratory listening.
The most practically significant expression of this constraint is the interface’s emphasis on the library — the collection of music the user has explicitly added to their Apple Music library — as the primary organizational surface. Unlike Spotify, which encourages a streaming relationship with the catalog in which the listener does not need to “own” music in any organizational sense, Apple Music’s interface more strongly implies a collection-building model inherited from iTunes. This has the cultural benefit of encouraging listeners to think in terms of albums and artists as objects of sustained engagement rather than simply as sources of tracks, but it also creates a friction between the organized library and the exploratory radio that can make extended discovery sessions feel structurally awkward in ways that Spotify’s more seamlessly streaming architecture avoids.
3. Tidal: Fidelity, Artist Positioning, and the Serious Listener
3.1 Institutional Logic
Tidal occupies the most clearly defined niche position of the major streaming platforms. Founded in 2014 and launched publicly with considerable fanfare as an artist-owned alternative to Spotify’s dominant model, Tidal has undergone substantial changes in ownership — most notably Jay-Z’s acquisition and subsequent sale of a majority stake to Jack Dorsey’s Square (now Block) — while maintaining a consistent positioning around two defining features: high-fidelity audio quality and a stated commitment to artist compensation and creative integrity.
The institutional logic shaping Tidal’s discovery architecture flows from this positioning. A platform that defines itself around fidelity and artistic seriousness is implicitly addressing a listener who cares about the quality of musical experience, not merely its ambient presence — a listener who listens attentively rather than as background, who values the album as a listening unit, and who is concerned with the relationship between creator and platform. This is, in principle, the listener profile most aligned with genuine exploratory listening.
3.2 High Fidelity and Its Relationship to Exploration
Tidal’s high-fidelity audio offering — which includes both lossless CD-quality streaming and, in its HiFi Plus tier, Dolby Atmos spatial audio and MQA-encoded material — is not merely a technical specification. It implies and partially produces a different mode of listening engagement. High-fidelity streaming is wasted on background listening; it is most meaningful for attentive, focused listening sessions in which sonic detail matters. A listener who has chosen Tidal specifically for its fidelity is signaling something about their relationship to music that distinguishes them, on average, from the broader streaming population.
This listener profile alignment has implications for discovery that are not fully reflected in Tidal’s explicit feature set. A platform whose user base skews toward attentive, quality-conscious listeners generates behavioral data that is systematically different from Spotify’s much larger and more heterogeneous user base. Collaborative filtering on Tidal’s data should, in principle, produce recommendations that reflect more developed taste profiles, more complete album listening (rather than track-level sampling), and greater tolerance for challenging or demanding material. Whether Tidal’s recommendation algorithms fully exploit this data advantage is a separate question, but the raw material for better discovery recommendations is arguably present in the user base composition in ways it is not for more mass-market platforms.
3.3 Editorial and Discovery Features
Tidal’s editorial curation has historically been strongest in hip-hop, R&B, and adjacent genres — reflecting the musical background of its original founding partners and the listener communities most culturally aligned with its artist-centered positioning. Its genre coverage is less comprehensive than Apple Music’s across jazz and classical traditions, and its editorial playlist output is substantially smaller than either Apple Music or Spotify’s.
Tidal’s most distinctive discovery feature is its emphasis on videos alongside audio — a natural extension of its high-quality media positioning — and its integration of live performance recordings, documentaries, and artist-created content into the listening experience. These materials function as contextual supplements to musical discovery: a live recording that shows an artist’s improvisational practice, a documentary that contextualizes an album’s creation, or an interview that illuminates an artist’s influences all provide the kind of background knowledge that transforms sonic encounter into genuine musical understanding. This is the dimension of discovery that purely audio-based platforms, including Spotify, cannot provide, and it represents Tidal’s most meaningful structural contribution to the exploration problem.
Tidal’s radio and algorithmic recommendation functions are generally considered less sophisticated than Spotify’s, reflecting the platform’s smaller user base (which thins the collaborative filtering data) and its historically smaller investment in recommendation infrastructure. The genre gravity well and novelty decay effects documented in Paper 2 are present in Tidal’s radio functions but operate differently given the different data profile of the user base. The thinner data may in some genre spaces paradoxically produce more surprising recommendations, because the algorithm has less confident data to anchor to and must draw on audio feature matching more heavily — producing sonic adjacencies that a data-rich system would not generate.
3.4 Structural Limitations
Tidal’s structural limitations as a discovery platform are directly related to its institutional positioning. A platform that emphasizes quality, artistic seriousness, and a relatively premium listener demographic necessarily has a smaller user base than mass-market competitors, and this data thinness constrains the sophistication of its algorithmic features. The tension between Tidal’s niche positioning and the data scale required for sophisticated recommendation is not resolvable without either compromising the niche position or accepting the algorithmic limitations that come with a smaller audience.
Tidal’s ownership history — the multiple changes of control, the persistent uncertainty about its commercial viability, the failed ambitions of the artist-ownership model — has also produced a platform that feels, in its interface and feature development, less consistently developed than Spotify or Apple Music. Discovery features that were announced have sometimes not materialized, and the overall user experience reflects the priorities of a platform that has been trying to establish commercial stability rather than one that has been able to invest consistently in exploratory listening infrastructure.
4. Amazon Music: The Convenience Model and the Ambient Listening Assumption
4.1 Institutional Logic
Amazon Music’s institutional logic is the most clearly distinct from the other platforms examined here, and understanding it requires understanding Amazon’s relationship to music as a product category within a much larger commercial ecosystem. Amazon does not primarily conceive of itself as a music company in the way that Spotify does, or as a creative industries patron in the way that Apple positions itself. Amazon is a logistics and convenience company that sells music streaming as part of a broader subscription bundle — Prime membership — and as a voice-assistant use case for Alexa-enabled devices.
This institutional positioning has profound implications for discovery design. A platform optimized for convenience and voice-assistant interaction is implicitly assuming a listener who wants music delivered with minimal friction in response to simple commands: “Alexa, play jazz,” “Alexa, play something relaxing,” “Alexa, play more like this.” The discovery architecture is designed around this use case, which means it is designed for listeners who are engaging with music as an ambient or background presence and who are interacting with the platform through a voice interface rather than a visual one.
4.2 The Voice Interface and Its Consequences for Discovery
The voice interface constraint is architecturally decisive for Amazon Music’s discovery capabilities in ways that have received insufficient analytical attention. Music discovery, as argued throughout this series, is a fundamentally browsing activity — it requires the ability to navigate a space of options, to encounter unexpected adjacencies, to follow recommendation chains across multiple steps, and to engage with contextual information that illuminates what one is encountering. All of these activities require a visual interface: the ability to see what is being recommended, to navigate to related artists or albums, to read editorial notes, and to make deliberate choices about the direction of exploration.
A voice interface collapses all of this navigational possibility into a linear conversational model. The listener can request music by genre, mood, artist, or activity; Alexa will play a queue; the listener can say “play more like this” or “I don’t like this”; and the session continues. The discovery space available within this interaction model is narrow and shallow — it is the discovery space of a very simple radio request rather than the navigational space of a visual catalog interface. Amazon Music’s investment in improving Alexa’s music intelligence has not resolved this fundamental constraint; it has merely made the linear model more responsive and more accurate within its inherent limitations.
Amazon Music’s visual interface — available on mobile and desktop — is more capable than the voice interface and supports conventional playlist and radio browsing, but it has historically received less development investment and attention than the voice experience, reflecting the platform’s strategic bet on Alexa as its differentiating feature. The result is a visual interface that is functional but less refined than Spotify’s or Apple Music’s, with discovery features that are adequate for casual use but underdeveloped for serious exploratory listening.
4.3 Algorithmic Character
Amazon’s algorithmic recommendation capabilities are, in principle, formidable. Amazon’s core competency is recommendation — the product recommendation engine that drives its e-commerce business is among the most sophisticated and commercially successful in the technology industry. However, the e-commerce recommendation model does not translate directly to music recommendation, for reasons that illuminate something important about the different nature of music as a preference object.
E-commerce recommendation is fundamentally a filtering problem: among millions of products, identify the specific items this customer is most likely to purchase given their purchase history, browsing behavior, and demographic profile. The recommendation is evaluated against a clear behavioral signal — purchase — and optimized accordingly. Music recommendation is a different kind of problem, because the relevant behavioral signals are more ambiguous, the preference structure is more complex, and the relationship between hearing something once and genuinely appreciating it is substantially weaker than the relationship between browsing a product page and buying it.
Amazon Music’s algorithmic recommendation reflects this translation difficulty. Its recommendations tend toward the obvious and the commercially familiar even more consistently than Spotify’s, because its model draws heavily on the kind of preference signals it knows how to process — explicit ratings, purchase history of physical music products through Amazon’s retail arm, and listening completion rates — rather than the more subtle behavioral signals that Spotify’s music-specific algorithms have been optimized to interpret. For discovery purposes, this produces a recommendation environment that feels commercially oriented in a more blunt and less musically nuanced way than Spotify’s genre gravity well effect, because there is less musical intelligence shaping which commercially adjacent options are presented.
4.4 The Bundling Effect
Amazon Music’s most significant structural feature from a discovery standpoint is not its algorithmic sophistication or its editorial depth but its bundling with Prime membership. Because a large proportion of Amazon Music’s users have not chosen it as their primary music service but have simply discovered it as an included feature of their Prime subscription, its user base includes a substantial population of casual or occasional listeners who would not necessarily subscribe to a standalone streaming service. This bundled population produces behavioral data that is systematically different from the self-selected populations of Spotify and Apple Music — it skews more heavily toward ambient and background listening, toward familiar and mainstream music, and toward lower engagement depth.
The bundling effect creates a discovery platform whose collaborative filtering data is dominated by the least exploratory listening behaviors in the streaming population. Recommendations derived from this data tend toward the most broadly accessible and least challenging music in any genre space, because the user base generating the data is not selecting music for reasons that reward marginal or demanding recommendations. This represents a structural ceiling on Amazon Music’s discovery potential that cannot be resolved by algorithmic sophistication alone — it is embedded in the composition of the user base that the business model produces.
5. YouTube Music: Inheriting the Algorithm
5.1 Institutional Logic
YouTube Music occupies a structurally unique position among the major streaming platforms because it is not a music service that has been built on top of a technology infrastructure but a technology infrastructure — YouTube’s recommendation engine — that has been partially repurposed as a music service. Understanding YouTube Music requires understanding YouTube itself, because the discovery architecture YouTube Music deploys is fundamentally inherited from the parent platform rather than purpose-built for music streaming.
YouTube is the world’s largest video platform and one of the most analytically consequential recommendation systems ever built. Its recommendation algorithm — responsible for determining which videos appear in the sidebar, the home feed, and the autoplay queue — has been extensively studied, reported on, and criticized for its tendency to drive viewers toward progressively more extreme or sensational content in the service of maximizing watch time. This tendency, which YouTube has made various attempts to moderate, reflects a fundamental design principle: the algorithm is optimized for engagement, measured as watch time, and it will route content toward whatever produces that engagement most efficiently, regardless of whether the routed content is what the viewer would have chosen, what the viewer would consider beneficial, or what the platform might prefer to be associated with from a reputational standpoint.
YouTube Music inherits this recommendation engine, modified for the music context. The implications are significant and not always appreciated in comparisons of streaming platforms’ discovery features.
5.2 The Watch Time Model and Music Discovery
YouTube’s watch time optimization model, when applied to music, produces a recommendation system with a distinctive character. Audio content on YouTube exists in a video wrapper, and the behavioral signals that drive the algorithm — view counts, watch completion rates, comments, likes, shares — are not identical to the listening signals that shape Spotify’s recommendations. A music video that is frequently watched to completion, commented on, and shared generates very different signal data than a deep album cut that is listened to passively. YouTube Music’s algorithm therefore tends to surface music that is engaging as a video experience, not merely as an audio experience — which systematically advantages music with strong visual components, strong cultural discussion value, and strong social sharing patterns.
For discovery, this has both benefits and costs. The benefit is that YouTube Music’s recommendation pathways are capable of traversing the full range of content on the platform, including unofficial uploads, live recordings, rare performances, and archival material that is not available on other streaming services. YouTube’s catalog is, in a functional sense, larger than any other streaming platform’s because it includes the vast unofficial archive of music that has been uploaded without formal licensing — bootleg concerts, rare television performances, regional music that has never been formally distributed through streaming channels, and historical recordings from before the streaming era. A listener who follows YouTube Music’s recommendation pathways into this unofficial territory encounters discovery possibilities that simply do not exist on any licensed streaming service.
The cost is that the watch time model systematically favors viral, socially discussed, and visually compelling music over quiet, demanding, or aesthetically introverted work. Music that generates strong social response — whether through its sonic character, its cultural associations, its artist’s celebrity, or simply its algorithmic promotion to a large initial audience — is amplified by YouTube’s recommendation system in ways that have nothing to do with musical quality. The TikTok-to-YouTube pipeline, in which tracks that go viral on TikTok are subsequently promoted by YouTube’s algorithm because their social signal data is extraordinarily strong, represents an extreme expression of this logic — a discovery ecosystem in which cultural virality and genuine musical significance have been so thoroughly conflated that distinguishing them requires deliberate critical effort.
5.3 Breadth Versus Depth in the YouTube Model
YouTube Music’s most significant structural advantage over other streaming platforms is breadth — the sheer range of content accessible through the platform, including material that exists nowhere else in streaming form. For a listener interested in exploring a historical tradition with limited formal streaming representation, YouTube’s unofficial archive is often the only digital resource available, and YouTube Music’s integration of this archive with a music streaming interface represents a genuinely meaningful extension of the catalog’s effective scope.
The structural disadvantage is depth. YouTube’s recommendation system is designed to maximize engagement across its entire content ecosystem, which means it is permanently prone to routing music listeners away from sustained engagement with a particular tradition, artist, or body of work and toward whatever adjacent content generates stronger engagement signals. The exploratory pathway that leads a listener from a foundational album into a tradition’s historical depth is precisely the pathway that YouTube’s algorithm is least likely to provide, because deep catalog exploration within a tradition does not generate the strong social signal data that drives the algorithm’s recommendations. A YouTube Music session that begins with serious exploratory intent is more likely to be routed toward recently viral material, toward the tradition’s most socially discussed rather than most musically significant figures, and toward adjacent content that preserves engagement rather than deepening it.
5.4 The Creator Economy Dimension
YouTube Music is also shaped by YouTube’s creator economy in ways that have no parallel in other streaming platforms. YouTube’s recommendation system is designed not only to serve viewers but to serve creators — to distribute content in ways that sustain the creator-advertising economy that generates YouTube’s revenue. This means that music discovery on YouTube Music takes place within an ecosystem that includes content specifically designed to game the recommendation algorithm — music-adjacent content that rides algorithmic recommendation pathways without providing genuine musical value, channels optimized for recommended views rather than musical depth, and promotional activity that exploits YouTube’s social signal infrastructure to surface commercially motivated content in ostensibly organic recommendation flows.
This dimension of the YouTube Music discovery environment does not have a direct equivalent on Spotify or Apple Music, and it adds a layer of noise to the discovery experience that listeners used to those platforms may not immediately recognize or know how to filter. The skill of navigating YouTube Music as an exploratory tool is therefore, in part, the skill of distinguishing between algorithmically amplified content and genuinely musically valuable content — a distinction that requires critical judgment that the platform’s interface does not assist and the algorithm does not support.
6. Cross-Platform Analysis: Structural Patterns and Divergences
Having examined each platform individually, it is possible to draw several comparative observations that illuminate the broader structural landscape of streaming discovery.
6.1 The Curation-Algorithm Spectrum
The platforms examined here can be roughly arranged on a spectrum from human-curation-dominant to algorithm-dominant discovery approaches. Apple Music sits closest to the human-curation end, with its editorial emphasis, its investment in professional curators, and its live radio programming. Tidal sits in a similar position within its more limited curatorial range. Spotify occupies the center of the spectrum, combining substantial algorithmic infrastructure with a significant editorial operation. Amazon Music and YouTube Music sit closest to the algorithm-dominant end, though for different reasons — Amazon because its editorial investment is limited, YouTube because its inherited recommendation engine reflects a fundamentally different institutional logic from music-specific curation.
Neither end of this spectrum is straightforwardly superior as a discovery approach. Human curation provides musical understanding and contextual knowledge that algorithms cannot supply; algorithmic curation provides scale, personalization, and adaptability that editorial teams cannot match. The platforms that provide the most potentially valuable discovery experiences are those that most effectively combine both approaches — and by this standard, Apple Music and Spotify, for different reasons and in different genre territories, are currently the strongest discovery platforms among the major services.
6.2 Catalog Depth and Data Asymmetry
All platforms face a version of the data asymmetry problem identified in Paper 2’s discussion of Spotify’s deep cut seed behavior: algorithmic recommendation performs better for well-streamed material than for catalog depth, and this asymmetry systematically disadvantages precisely the music most rewarding for exploratory listeners. The asymmetry is most acute on Amazon Music, whose bundled user base generates the least exploratory behavioral data; somewhat less acute on Spotify, whose larger user base includes enough dedicated exploratory listeners to generate usable data for moderately marginal material; and most partially mitigated on YouTube Music, where the unofficial archive extends the discoverable catalog beyond what licensing constraints allow on other platforms, even though the recommendation system does not reliably route toward that depth.
6.3 The Mobile Interface Constraint
All platforms have made mobile the primary interface through which most listeners interact with them, and this common constraint shapes discovery architecture across the spectrum. Mobile interfaces impose physical constraints on browsing depth — small screens limit the visual information density that supports exploration, touch navigation is less precise than mouse navigation for following recommendation chains, and the predominantly on-the-go contexts of mobile listening systematically favor comfort over challenge. The platforms that invest most heavily in desktop and web interfaces — Apple Music and Tidal, which both have more fully developed desktop applications than the others — partially mitigate this constraint for the listeners who use those interfaces, but the majority of streaming engagement across all platforms happens on mobile, and the discovery architecture reflects this reality.
6.4 Incentive Alignment and Discovery
Perhaps the most fundamental comparative observation is that no major streaming platform has fully resolved the incentive misalignment between subscriber retention and genuine discovery. Genuine discovery, as Paper 1 established, involves encountering unfamiliar music that may initially be uncomfortable or confusing and requires sustained engagement to appreciate. This is precisely the experience that retention-optimized platforms are structurally discouraged from providing, because it risks listener disengagement in the short term even when it produces listener enrichment in the long term. All five platforms examined here are subject to this structural misalignment, and none has developed a discovery architecture that consistently prioritizes genuine exploration over comfortable curation. The differences among them are differences of degree and emphasis, not differences of fundamental structural orientation.
7. What a Genuinely Exploratory Platform Would Require
The comparative analysis suggests several features that a platform genuinely optimized for musical exploration — rather than comfortable retention — would require, and which no current major platform provides in fully developed form.
It would require a recommendation system that distinguishes between different exploration modes — the listener who wants to find contemporary music adjacent to their existing taste, the listener who wants to understand a historical tradition in depth, the listener who wants to encounter a genuinely unfamiliar musical culture — and routes differently depending on which mode the listener has elected. Current platforms treat all exploratory listening as a single category and optimize for the most common variant, which is the least demanding and shallowest form.
It would require contextual information integrated into the exploration experience — the kind of historical and critical context that allows a listener to understand what they are encountering rather than merely hear it. This information is what music journalism, liner notes, and knowledgeable curatorial voices have historically provided, and its absence from streaming platforms’ discovery interfaces is a structural gap that none has adequately addressed.
It would require a treatment of catalog depth that is not systematically subordinated to recency — a recommendation architecture that can route listeners toward historically significant material with the same confidence it routes them toward recent releases, regardless of the streaming-era data profile of that historical material.
And it would require a conception of discovery success that is not reducible to skip rates and session length — a longer-term view of listener development in which the gradual acquisition of musical knowledge and the expansion of taste over time are recognized as genuine platform value rather than as incidental outcomes of a retention-focused service.
8. Conclusion: Different Shapes of the Same Ceiling
The comparative analysis of Apple Music, Tidal, Amazon Music, and YouTube Music against the discovery standard established in Papers 1 and 2 reveals that each platform has a distinctive character, shaped by genuine differences in institutional logic, technical infrastructure, and audience positioning. Apple Music’s editorial depth in certain genre territories, Tidal’s fidelity-appropriate attentive listener profile, YouTube Music’s unofficial archive breadth, and Amazon Music’s voice-interface convenience all represent real features that distinguish their discovery environments from Spotify’s.
But these differences do not add up to a platform that has genuinely solved the discovery problem. Each platform’s distinctive strengths are real but bounded, and each shares with Spotify the fundamental structural constraint that makes the playlist the ceiling rather than the floor of streaming discovery: the incentive to optimize for retention, comfort, and commercial promotion at the expense of genuine exploration. The ceiling is differently shaped on different platforms, but it is a ceiling nonetheless.
Paper 4 turns from the internal analysis of streaming platforms to an examination of what preceded them — specifically, what broadcast radio got right about discovery that streaming has not, and what it got wrong that streaming has correctly moved away from. The radio model’s genuine achievements in creating discovery occasions, providing curatorial voices, and sustaining shared listening culture represent a set of solutions to problems that streaming has not solved. Understanding those solutions, and the structural conditions that made them possible, is a necessary step toward imagining what a genuinely exploratory streaming architecture might look like.
This white paper is the third in the Beyond the Playlist series. Paper 4, “The Radio Analogy: What Broadcast Radio Got Right (and Wrong) About Discovery,” examines broadcast radio as a discovery mechanism — its sociology, its curatorial logic, and what its structural achievements reveal about the limits of algorithmic curation.
