Payment in exchange for ongoing conditional access to a controlled environment appears across physical spaces, institutional systems, marketplace access, and digital services. A gym membership grants access to facilities that remain under the operator's control and can be revoked. A university charges tuition for access to an educational environment that determines what credentials a student can earn and under what conditions. A seller platform charges fees for access to a marketplace whose rules, ranking systems, and customer relationships belong to the platform rather than to the seller operating within it. This structure predates the commercial internet and digital platforms aren't an exception to it; they're its most developed form.
The subscriber who pays for a streaming service, a productivity suite, or a professional network understands the subscription transaction clearly enough. The bill arrives, the payment processes, and the account remains active. What the subscriber doesn't see is the second transaction running alongside the first. Payment doesn't terminate extraction and sale of personal digital information; it leaves the behavioral record intact. The presence of a price creates the appearance that the transaction is complete but it does not.
Note from the Author: during this series I will be using some of the jargon that these industries apply. Two terms appear throughout and are worth defining at the outset. 'Participation' describes the act of engaging with a digital domain: the attention, behavior, time, and data that people contribute when they use a platform or service. 'Surface' describes the point of contact, the place where a person touches a platform, domain, or industry and where participation begins.
The Access Layer
Before any of the platforms examined in this paper, there's the layer through which the subscriber reaches them. An internet service provider can observe a subscriber's traffic across every service they use, across every domain documented in this series, and before any platform-specific filtering occurs. No platform examined in this paper has that position. The ISP is upstream of all of them.
AT&T's Internet Preferences program, reported by Ars Technica in 2015 and confirmed in AT&T product materials, offered GigaPower fiber service at $70 per month to subscribers who agreed to have their web browsing data collected and used for targeted advertising. Subscribers who declined the targeted advertising paid $99 per month. AT&T stated explicitly that advertisers pay for the opportunity to deliver targeted advertising to data-sharing subscribers, and that this payment is what allows the lower price to be offered. The $29 per month differential, or $348 per year, is a confirmed instance of an ISP assigning an explicit dollar value to a subscriber's browsing data in its own commercial materials. It's historical, limited in scope, and not representative of current industry-wide practice. It's included because it is the only disclosed, platform-confirmed pricing of ISP subscriber behavioral data available from a primary source.
The FCC adopted broadband privacy rules in October 2016 that would have required ISPs to obtain opt-in consent before selling or sharing subscriber web browsing data with advertisers and third parties. Congress repealed those rules in April 2017 under the Congressional Review Act, signed by President Trump, which limits the FCC's ability to issue substantially similar rules in the future under the Congressional Review Act. The repeal established on the legislative record that ISPs have the legal capacity to monetize subscriber browsing data without consent. The full value of that capacity isn't disclosed at the per-subscriber level the Personal Data Royalty Formula (PDR) formula requires. The internet access layer is upstream of everything in this series. Its behavioral transaction is real, confirmed in one disclosed historical instance, and legally established. The paper now moves to the layers where that value is measurable.
When a person pays for a streaming service, a music platform, or a productivity suite, they're entering a transaction whose terms appear straightforward: money in exchange for access. The platform receives revenue, the subscriber receives a service, the price is disclosed, the billing is regular, and the relationship is legible in a way that free platforms never are.
The subscription fee finances only part of what the platform does. It covers content acquisition, infrastructure, and operating costs. What it doesn't cover, and what it doesn't describe, is the second commercial activity running in parallel: the collection, organization, and commercial application of the behavioral record that each subscriber generates through their use of the service. This record isn't a byproduct of the service, but an input to systems whose commercial value is entirely separate from the subscription revenue the platform reports. These aren't the same transaction: one is recorded, the other is not. The subscription fee flows from the subscriber to the platform in exchange for access, but the behavioral record flows from the subscriber to the platform in exchange for nothing, because the subscriber isn't recognized as having contributed anything beyond the fee they already paid. This structure recurs across every paid platform examined in this paper. It's not incidental to how these platforms operate, it is the architecture.
The structure reaches its furthest extent in platforms whose behavioral record isn't a record of entertainment choices or work habits but a consolidated inventory of a subscriber's complete digital existence. A password manager holds every service a person uses, every account they maintain, and the frequency and pattern of their access to each. The subscriber paying for the security that a password manager provides is generating, in the same act, a highly consolidated behavioral inventory of their digital life. That inventory is the second transaction at its most complete: a subscriber who paid for one thing and contributed another, with no recognition that the contribution occurred.
Netflix, Inc. reported total revenues of $39.0 billion for the fiscal year ended December 31, 2024, a 16% increase over the prior year. The United States and Canada region generated $17.4 billion of that total, against 84.1 million average paying memberships in the region. Dividing United States and Canada revenue by the average paying membership base produces a confirmed per-member annual revenue figure of approximately $207. That figure represents what Netflix realizes from the average North American subscriber in a given year from subscription fees alone. It is derived directly from the company's Annual Report on Form 10-K and requires no estimation.
The $207 figure is the disclosed transaction. What the subscriber also contributes, and what doesn't appear in that figure, is the viewing record.
Netflix has indicated that a substantial majority of content streamed on its platform comes from its recommendation system rather than from a subscriber's own search. That figure describes the scale of the viewing record's commercial function. The platform isn't delivering a catalog that subscribers navigate independently. Instead, it's delivering a continuously personalized sequence of content whose construction depends entirely on what each subscriber has watched, abandoned, rewatched, paused, and searched for across their entire history on the platform. The recommendation system is the product and the viewing record is what builds it.
That record does more than generate recommendations. Netflix uses viewing behavior to make content acquisition and production decisions. When Netflix commissions a series, cancels one, or pays a specific amount for a specific license, those decisions are informed by the behavioral data assembled from hundreds of millions of subscribers. The subscriber whose viewing history contributed to a content decision worth hundreds of millions of dollars wasn't a consultant on that decision, they were the data from which it was derived. What the user received was their next recommendation. The platform received a commercial intelligence system that no individual subscriber could've produced and that all subscribers collectively funded through their behavioral contributions without any of them being recognized as having contributed.
Netflix also uses viewing data for subscriber retention modeling. The platform tracks which behavioral signals predict cancellation and builds intervention systems around them. A subscriber whose viewing patterns indicate disengagement may receive different recommendations, different notifications, or different content prominence than a subscriber showing strong engagement signals. The subscriber is the subject of the system that manages their relationship with the platform, and they aren't a party to that process.
The viewing record isn't ancillary to the service, but the system through which the service operates. The subscriber who pays $207 a year for access to Netflix's content library is simultaneously contributing to a behavioral intelligence system whose commercial value is woven into every content, retention, and product decision the platform makes. The subscription fee is disclosed on the billing statement. The behavioral contribution isn't disclosed anywhere.
Spotify Technology S.A. reported total revenue of €15.67 ($16.9) billion for fiscal year 2024. As of the fourth quarter of 2024, Spotify had 263 million premium subscribers and 675 million monthly active users globally. Premium revenue for the full year was driven by subscriber growth and price increases, with a quarterly average revenue per user of approximately €4.85 per month ($59 per year). The platform's free tier, which is supported by advertising, reached 425 million users in the fourth quarter of 2024, an increase of 12% year over year.
Spotify's listening data is organized around a commercial insight the platform has pursued systematically since at least 2014, when it acquired Echo Nest, a music intelligence company, giving it access to the behavioral data embedded in more than 1.5 billion user-generated playlists. What that acquisition revealed, and what Spotify has built a significant part of its advertising business upon, is that listening behavior doesn't merely indicate musical preference. It indicates emotional state, daily routine, activity context, and purchasing psychology with a precision that conventional demographic targeting cannot match.
In 2016, Spotify entered into a multi-year partnership with WPP, at the time the world's largest advertising and public relations holding company, structured around audience segmentation and mood-based targeting derived from listening behavior. The commercial rationale was stated plainly: music attributes including tempo, energy, and valence are measurable proxies for emotional state, and emotional state is a more precise targeting variable than age, location, or declared interest. The partnership gave WPP's clients access to targeting built on what Spotify's subscribers were listening to, when, and in what emotional context.
Spotify has also filed patents for technology designed to infer a user's emotional state, age, gender, and accent from their voice and ambient background audio. A patent granted in January 2021, United States Patent No. 10,891,948, described a system that would analyze content metadata extracted from a user's voice to determine attributes that could then be used to recommend content. The commercial implication is straightforward: a platform that can infer a subscriber's emotional state from audio signals can time commercial messages to moments of specific psychological receptivity.
The premium subscriber who pays to remove advertising hasn't paid to remove any of this. The removal of advertisements from the listening experience is what the premium fee purchases. The listening record, the mood inference, the behavioral profiling, and the commercial application of the subscriber's emotional and psychological data continue regardless of subscription tier. The person paying €4.85 a month ($59 per year) to listen without interruption is still contributing a continuous record of their emotional life to a platform that has built commercial partnerships specifically around the value of that record. The interruptions are gone but the extraction is not.
Microsoft Corporation reported that Microsoft 365 Consumer reached 82.5 million subscribers as of the fiscal year ended June 30, 2024, with consumer subscription revenue growing year over year. Microsoft 365 is priced at $99.99 annually for an individual subscription, producing a confirmed per-subscriber annual revenue baseline of approximately $100, against which the behavioral record each subscriber generates through their use of Word, Excel, Outlook, Teams, and OneDrive runs continuously in the background.
Productivity telemetry is embedded throughout Microsoft 365. The platform records document creation and editing patterns, email composition and response behavior, meeting attendance and participation signals, search queries within the suite, and usage patterns across applications. This data feeds into product development, into Microsoft's AI systems including Copilot, and into the commercial intelligence infrastructure that Microsoft operates across its enterprise and consumer products. The subscriber paying for the ability to write documents and send email is also contributing a detailed behavioral record of how they work, when they work, what they work on, and how their productivity patterns change over time. That record isn't described in the subscription terms as a contribution they are making to a second commercial system but classified as telemetry necessary for the service to function.
LinkedIn presents the dual transaction structure more plainly than any other platform in this paper. Microsoft confirmed in its earnings call for the second quarter of fiscal year 2025 that LinkedIn Premium subscription revenue had passed $2 billion on an annual basis, with premium subscribers up 25% year over year. LinkedIn Premium is sold to subscribers as a career acceleration tool: it provides expanded search visibility, direct messaging to people outside a subscriber's network, and insights into who has viewed their profile. The subscriber pays for these features because they are trying to find a job, build a business relationship, or advance professionally.
While they're doing that, they're also generating one of the most commercially precise behavioral signals in the paid platform economy: professional intent. A LinkedIn Premium subscriber's search behavior, profile views, connection requests, message content, and engagement with job listings constitute a detailed record of their career aspirations, their professional relationships, their skills, and the employers and roles they are actively considering. LinkedIn organizes this record and sells access to it through its Talent Solutions, Sales Solutions, and Marketing Solutions products, each of which allows employers, recruiters, and advertisers to target LinkedIn's user base based on the behavioral signals those users generate.
The subscriber paying for LinkedIn Premium to find a job is simultaneously generating the career intent signal that LinkedIn sells to the same employers they are trying to reach. The subscriber is both the applicant and the input to the system being sold to the same market. They've paid for visibility but they haven't been told that their search behavior is the product being sold to the other side of the transaction they are trying to complete.
The Introduction to this series established four conditions whose simultaneous presence is required for participation to function as voluntary exchange: survivable refusal, recognized standing, transparency of terms, and independent jurisdiction. Papers One through Five showed these conditions failing through engagement optimization, institutional compulsion, hardware activation, and the cumulative dependency that free platforms engineer over time. This is the first paper in the series where payment is present but recognition of the second transaction is not. That distinction changes which conditions fail and how.
Survivable refusal is more plausible on paid platforms than in any domain examined so far. A person can cancel Netflix without losing access to essential services. They can stop using Spotify without losing their social connections or their professional relationships. They can decline LinkedIn Premium and continue using the free tier. The survivable refusal condition is weak rather than absent, because the network effects and interwoven dependencies that make refusal costly on free platforms operate more mildly here. A Netflix subscriber who cancels loses their viewing history and their recommendations but they don't lose their income, relationships, or the ability to participate in economic life. This is a meaningful distinction from the conditions documented in Papers Three, Four, and Five.
Recognized standing fails clearly and completely. The subscriber is recognized as a paying customer whose billing relationship with the platform is well-documented and legally enforceable. They aren't recognized as a behavioral contributor whose viewing record, listening history, productivity patterns, or career intent signals generate commercial value that exists entirely outside the subscription fee they paid. The platform records the subscription payment with precision. It doesn't record the behavioral contribution at all — not in the subscriber's account, terms of service, nor in any disclosure the subscriber receives at any point in the relationship.
Transparency of terms fails for the same structural reason it has failed across every paper in this series, but with an additional layer specific to paid platforms. On free platforms, the absence of a price is itself a signal that something else is being exchanged, even if that something isn't disclosed. On paid platforms, the presence of a price suppresses that signal. The subscriber who sees a monthly charge on their bank statement has received apparent evidence that the transaction is complete, that the charge and the service account for each other. What the billing statement doesn't show is the second transaction: the behavioral record that was generated during the billing period, what it contained, what it was used for, and what it was worth to the platform beyond the subscription fee already collected.
Independent jurisdiction is present in formal terms and limited in practical ones. Subscription contracts are governed by consumer protection law, and regulatory frameworks exist for challenging deceptive enrollment and cancellation practices. They don't provide any mechanism through which a subscriber could claim recognition as the origin of the behavioral record they generate, or any standing in the commercial arrangements that record makes possible. The regulatory architecture addresses the subscription transaction but it doesn't address the behavioral transaction running alongside it.
The Personal Data Royalty (PDR) formula established in the Introduction to this series, and applied across Papers One through Five, produces a per-user participation value baseline from disclosed financial data. In the paid platform domain, the formula operates differently from earlier papers in one important respect. The primary revenue figure is subscription revenue, not advertising revenue. The subscriber is the direct revenue source rather than the audience being sold to advertisers. This means the PDR calculation captures only one of the two transactions this paper documents. It measures the subscription transaction, whose terms are disclosed and whose value is confirmed from primary sources. It can't measure the behavioral transaction, whose value is neither disclosed nor recognized.
In the Origin Economics framework, Y = λ · f(H, K, T) expresses output as a function of human-origin participation, capital, and technology, multiplied by whether the legitimacy conditions of the exchange were satisfied. For paid platforms, lambda doesn't fail through compulsion or hardware activation, it fails through structural omission. The subscriber knows what they are paying but they don't know what they are generating beyond the payment. The legitimacy conditions for the behavioral transaction were never established because the behavioral transaction was never disclosed as a transaction.
The Netflix United States and Canada per-member annual revenue baseline of approximately $207, derived from the company's 10-K, represents the subscription transaction. It's confirmed, precise, and directly attributable to the subscriber's relationship with the platform. The gap between that figure and the total value the platform realizes from each subscriber — including the contribution of the viewing record to content decisions, retention systems, recommendation training, and AI development — isn't calculated to precision in this paper, but is established as present and unaccounted for. That gap is what the PDR formula is designed to make visible, and in this domain its visibility is complicated by the fact that platforms whose primary revenue is subscription-based have less incentive to disclose behavioral data value than platforms whose advertising revenue makes the connection between user behavior and platform revenue explicit.
For Spotify, the premium subscriber baseline can be estimated from publicly disclosed figures. Spotify's full-year 2024 premium revenue of approximately €13.6 ($16.9) billion divided across an average premium subscriber base of approximately 250 million globally produces a per-subscriber annual revenue figure of approximately €54, or roughly $59 at average 2024 exchange rates. This figure, like the Netflix baseline, represents the disclosed transaction only. The listening record that each subscriber generates, the mood inference systems built from that record, and the audience segmentation products sold to advertisers based on it constitute a second commercial activity whose value to the platform isn't captured in premium subscription revenue and isn't disclosed to the subscriber at any point.
The PDR baselines for this paper are therefore floors in a more specific sense than in earlier papers. They're not floors because data is missing from the calculation, they're floors because the calculation is structurally limited to one of the two transactions this paper documents: the subscriber's recognized contribution to the platform is their subscription fee, their unrecognized contribution is the behavioral record. The formula measures the first. The second is established as real, commercially significant, and entirely outside any accounting that acknowledges the subscriber as its origin.
The primary objection to the analysis presented in this paper is that the subscriber chooses to pay, the terms were disclosed, and the relationship is therefore legitimate. This objection correctly identifies that paid platform subscribers exercise more meaningful choice than users of free platforms, compelled students, or households whose participation began at the moment of hardware purchase. However, it doesn't address the misclassification argument, because the misclassification doesn't occur in the subscription transaction. It occurs in the behavioral transaction that runs alongside it.
A subscriber who understands and accepts Netflix's subscription terms has consented to pay for access to Netflix's content library. They haven't consented to anything about the viewing record, because the viewing record isn't described in the subscription terms as a contribution they're making to a second commercial system. The consent that the subscription transaction produces doesn't extend to a transaction whose existence hasn't been disclosed. Choosing to pay does not constitute consent to an arrangement the subscriber wasn't told about.
The second objection is that the behavioral record benefits the subscriber through better recommendations, more relevant content, and a more personalized experience. This is accurate. The recommendation system that Netflix builds from viewing data does improve the subscriber's experience. The mood-based playlist construction that Spotify builds from listening data does make the service more useful and the career insights that LinkedIn builds from professional behavior data do benefit the subscriber's job search.
The benefit argument has been addressed in every paper in this series and the answer is the same here. The existence of value on both sides of an exchange doesn't establish that the exchange has been accurately recorded or that both parties have been recognized as contributors. A Netflix subscriber receives a better recommendation. Netflix receives a behavioral intelligence system that informs content decisions worth billions of dollars annually, built from the collective viewing records of hundreds of millions of subscribers, none of whom are recognized as having contributed to it. The subscriber received something but that doesn't establish that what they contributed was recognized, valued, or accounted for as the productive input it actually is.
The paid platform sector faces regulatory exposure across two distinct frameworks. Neither addresses the behavioral transaction this paper documents, but both establish that the subscription relationship is within active regulatory scope.
The Federal Trade Commission's action against Amazon under the Restore Online Shoppers' Confidence Act, filed in June 2023 and settled in September 2025 for $2.5 billion, is the most significant regulatory event in the subscription economy to date. The FTC alleged that Amazon used deceptive interface designs to enroll consumers in automatically renewing Prime subscriptions without their informed consent, and deliberately complicated the cancellation process to prevent subscribers from leaving. The settlement included a $1 billion civil penalty and $1.5 billion in consumer redress. It established that subscription enrollment and cancellation practices are within active FTC enforcement scope and that the penalty for systematic deception in the subscription economy is now measured in billions. The settlement doesn't address behavioral data extraction, but only one layer of the subscription transaction: whether the subscriber understood and consented to what they were paying for. The behavioral transaction running alongside the subscription remains entirely outside its scope.
State auto-renewal laws, now enacted in the majority of American states, require that automatic subscription renewals be disclosed clearly at the point of enrollment and that cancellation be made at least as easy as enrollment. These frameworks, like the FTC action, address the subscription transaction. They regulate whether the subscriber understood the terms of the financial commitment they were making. They don't regulate whether the subscriber understood that their behavioral record was being collected, organized, and applied commercially as a second transaction running throughout the subscription period.
The regulatory direction that most directly implicates the behavioral transaction is data portability. The European Union's Digital Markets Act requires designated gatekeepers to provide users with the ability to take their data with them when they leave a platform. If a subscriber's behavioral record has commercial value, the question of what rights the subscriber has to that record on departure becomes a concrete legal question rather than a theoretical one. The regulatory architecture for answering it doesn't yet exist in the United States. Its development in European frameworks points toward the direction in which the gap between subscription regulation and behavioral transaction regulation may eventually close.
The boundary of current regulation is clear: subscription practices are regulated and the behavioral transaction is not addressed. The subscriber who was deceived about the terms of their Prime enrollment has a remedy, yet the subscriber whose viewing record, listening history, and productivity patterns generate commercial value that was never disclosed as part of their subscription relationship has none.
The PDR calculation in this paper rests on confirmed figures from Netflix, Spotify, Microsoft 365, and LinkedIn. It doesn't include the following platforms. Each exhibits the same dual transaction structure: a disclosed subscription fee running alongside an undisclosed behavioral record whose commercial value isn't attributable to individual subscribers at the level the formula requires. Their exclusion is methodological. Their behavioral transactions are real.
The categories below aren't exhaustive, they're representative of the system.
Streaming and Entertainment
Apple TV Plus, Disney Plus, Max, Paramount Plus, Peacock, and YouTube Premium each charge subscription fees and each assemble viewing records whose commercial application to recommendation systems, content decisions, and behavioral profiling is structurally identical to the Netflix model documented in this paper. Apple TV Plus operates within Apple's broader ecosystem, where viewing data connects to device usage, purchase behavior, and health and location signals across the subscriber's entire Apple relationship. Disney Plus operates within a content and theme park ecosystem whose commercial intelligence value extends far beyond streaming revenue. None disclose per-subscriber behavioral data value at the level the formula requires.
Creative and Professional Tools
Adobe Creative Cloud is the dominant subscription platform for professional creative work, with more than 30 million subscribers paying for access to Photoshop, Illustrator, Premiere Pro, and the broader Creative Suite. Every design decision, every editing pattern, every tool selection, and every project workflow that a subscriber produces inside Adobe's environment generates behavioral telemetry informed by signals generated through professional use, including the development of Adobe's AI systems. Autodesk, the dominant subscription platform for engineering and CAD software including AutoCAD, Revit, and Fusion 360, assembles equivalent telemetry from the engineering and design workflows of its professional subscribers. In both cases, the behavioral record being generated is a record of professional creative and technical labor whose commercial value to AI development and product intelligence is substantial and undisclosed.
Music Production
Avid Pro Tools, the dominant subscription platform for professional audio production, and Native Instruments capture detailed workflow telemetry from professional music producers and audio engineers. The production patterns, tool selections, and creative decisions of professional audio subscribers constitute behavioral records whose value to AI audio development is significant and entirely outside the scope of the subscription fee those subscribers pay.
Gaming Subscriptions
Xbox Game Pass, PlayStation Plus, Nintendo Switch Online, and EA Play each charge subscription fees while assembling detailed behavioral records of play patterns, session lengths, social interactions, and in-game decision making. Xbox Game Pass has more than 34 million subscribers. PlayStation Plus has more than 47 million. Neither Microsoft nor Sony discloses per-subscriber behavioral data value at the level the formula requires.
News and Publishing
The New York Times, The Wall Street Journal, The Washington Post, and The Atlantic each charge subscription fees while assembling reading behavior records: what articles subscribers read, for how long, in what sequence, at what time of day, and with what engagement patterns. Reading behavior is among the most precise signals of political preference, intellectual interest, and commercial intent available to any platform. None disclose per-subscriber behavioral data value.
Learning and Skills Platforms
Duolingo Plus, Masterclass, Skillshare, Coursera, and LinkedIn Learning charge subscription fees while capturing detailed behavioral records of learning patterns, topic interests, skill development trajectories, and engagement signals. The subscriber paying to develop a professional skill is generating a record of their career aspirations and learning behavior that has commercial value in the same markets their LinkedIn profile data reaches.
Cloud Storage and Productivity
Dropbox, Box, iCloud, and Google One charge subscription fees while storing and analyzing the documents, files, and behavioral patterns of their subscribers. Cloud storage platforms have access to the complete record of what their subscribers create, save, and organize, a behavioral signal whose commercial value is embedded in AI training, product development, and the broader data ecosystems of the companies that operate them.
The confirmed per-subscriber baselines established in this paper represent the subscription transactions of the platforms that disclose revenue with sufficient precision for calculation. They don't represent the total value each subscriber's behavioral record contributes to each platform's commercial operations. The floor is not the number; it's the portion that has been disclosed.
Netflix reported $17.4 billion in United States and Canada revenue for fiscal year 2024, against 84.1 million average paying memberships, producing a per-member annual baseline of approximately $207. Spotify reported €15.67 ($16.9) billion in total revenue for fiscal year 2024 across 263 million premium subscribers. Microsoft 365 Consumer had 82.5 million subscribers. LinkedIn Premium passed $2 billion in annual subscription revenue. Every subscriber in every one of these figures paid for access to a service and received it. The transaction was real, the billing was transparent, and the service was delivered. Alongside every one of those transactions, a second transaction ran without disclosure, without recognition of the subscriber as its origin, and without any accounting of what the subscriber's behavioral record was worth to the platform beyond the fee already collected.
The subscriber has standing in the subscription transaction, but no recognized standing in the behavioral transaction. Both are real. Only one appears on the receipt. The price was never limited to what appeared on the receipt.
Running total after Paper Six. Add only the lines that apply to you.
Netflix, Inc., Annual Report on Form 10-K, fiscal year ended December 31, 2024, filed with the United States Securities and Exchange Commission, January 2025.
Netflix Technology Blog, Netflix Research: Recommendations, various publications 2012 to present. research.netflix.com.
Spotify Technology S.A., Q4 2024 Shareholder Letter and Earnings Materials, filed with the United States Securities and Exchange Commission, February 2025.
Spotify Technology S.A., United States Patent No. 10,891,948, Identification of Taste Attributes from an Audio Signal, granted January 12, 2021. Filed February 8, 2018.
Liz Pelly, Big Mood Machine, The Baffler, 2019. Documents Spotify's mood data commercial infrastructure and WPP partnership.
Microsoft Corporation, Annual Report on Form 10-K, fiscal year ended June 30, 2024, filed with the United States Securities and Exchange Commission, July 2024.
Microsoft Corporation, Earnings Release FY2025 Q2, January 29, 2025. LinkedIn Premium subscription revenue confirmed as exceeding $2 billion on an annual basis.
Federal Trade Commission v. Amazon.com, Inc., Case No. 2:23-cv-00932, United States District Court, Western District of Washington. Complaint filed June 21, 2023. Stipulated Order for Permanent Injunction, Monetary Relief, Civil Penalty Judgment, and Other Relief entered September 2025. Civil penalty $1 billion; consumer redress $1.5 billion; total settlement $2.5 billion. ftc.gov/legal-library/browse/cases-proceedings/2123050-amazoncom-inc-rosca-ftc-v.
AT&T Internet Preferences program. Pricing confirmed in AT&T product materials, 2013 to approximately 2016. Reported by Ars Technica, 2015. GigaPower fiber service: $70 per month with data sharing, $99 per month without. Differential of $29 per month, or $348 per year, represents AT&T's disclosed valuation of subscriber browsing data for advertising purposes.
Federal Communications Commission, Protecting the Privacy of Customers of Broadband and Other Telecommunications Services, FCC 16-148, adopted October 27, 2016. 81 Fed. Reg. 87274, December 2, 2016.
Congressional Review Act repeal of FCC broadband privacy rules. Senate vote March 23, 2017. House vote March 28, 2017. Signed by President Trump April 3, 2017. Public Law 115-22.
Imanol Arrieta-Ibarra, Leonard Goff, Diego Jiménez-Hernández, Jaron Lanier, and E. Glen Weyl, Should We Treat Data as Labor? Moving beyond Free, AEA Papers and Proceedings 108, 2018, pp. 38–42.
Eric Posner and E. Glen Weyl, Radical Markets: Uprooting Capitalism and Democracy for a Just Society, Princeton University Press, 2018.
Shoshana Zuboff, The Age of Surveillance Capitalism, PublicAffairs, 2019.