Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 1 of 118
`
`
`
`
`
`
`BATHAEE DUNNE LLP
`Yavar Bathaee (CA 282388)
`yavar@bathaeedunne.com
`Edward M. Grauman (p.h.v. forthcoming)
`egrauman@bathaeedunne.com
`Andrew C. Wolinsky (p.h.v. forthcoming)
`awolinsky@bathaeedunne.com
`445 Park Avenue, 9th Floor
`New York, NY 10022
`Tel.: (332) 322-8835
`
`Brian J. Dunne (CA 275689)
`bdunne@bathaeedunne.com
`633 West Fifth Street, 26th Floor
`Los Angeles, CA 90071
`Tel.: (213) 462-2772
`
`Attorneys for Plaintiffs
`
`
`UNITED STATES DISTRICT COURT
`
`NORTHERN DISTRICT OF CALIFORNIA
`
`SAN JOSE DIVISION
`
`
`TODD CROWDER, KEVIN SCHULTE, and
`GARRICK VANCE, on behalf of themselves and all
`others similarly situated,
`
` Plaintiffs,
`
`
` Case No. 5:22-cv-00237
`
`
`CLASS ACTION COMPLAINT
`
`
`DEMAND FOR JURY TRIAL
`
`v.
`
`LINKEDIN CORPORATION,
`
` Defendant.
`
`
`
`
`
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 2 of 118
`
`
`
`TABLE OF CONTENTS
`
`III.
`
`IV.
`
`V.
`
`INTRODUCTION ....................................................................................................................................... 1
`PARTIES ................................................................................................................................................... 12
`I.
`PLAINTIFFS ................................................................................................................................. 12
`II.
`DEFENDANT ............................................................................................................................... 13
`JURISDICTION AND VENUE ................................................................................................................ 15
`DIVISIONAL ASSIGNMENT ................................................................................................................. 16
`I.
`LINKEDIN: THE BEGINNING OF A WINNER-TAKE-ALL BUSINESS ............................... 17
`A.
`SocialNet: A Social Network Before Its Time .................................................................. 17
`B.
`PayPal and the Power of Network Effects ......................................................................... 18
`C.
`“Blitzscaling”: Hoffman’s Insight from PayPal ................................................................ 20
`THE RISE OF LINKEDIN: A SPRINT TO DOMINANCE ........................................................ 22
`A.
`The Founding of LinkedIn ................................................................................................. 22
`B.
`LinkedIn Begins Monetizing Its Network of Users ........................................................... 24
`LINKEDIN IS ACQUIRED BY MICROSOFT WITH AN EYE TOWARD AI-DRIVEN
`MONETIZATION ......................................................................................................................... 27
`LINKEDIN’S DATA AND AI JUGGERNAUT .......................................................................... 29
`A.
`LinkedIn’s Extraction of Structured User Data to Train Machine-Learning Models ....... 30
`B.
`LinkedIn’s Data Centralization and Distribution Infrastructure ........................................ 34
`C.
`LinkedIn’s Social, Telemetry, and Inferred Data .............................................................. 37
`D.
`LinkedIn’s Acquisition of Drawbridge .............................................................................. 39
`THE LINKEDIN API AND THE SALE OF USER DATA TO PARTNERS ............................. 42
`A.
`LinkedIn’s REST and JSON API System ......................................................................... 42
`B.
`LinkedIn Privatized Its APIs in 2015 ................................................................................ 44
`C.
`LinkedIn Enters into Agreements with Select “Partners” for User Data Purchases
`through the APIs. ............................................................................................................... 46
`VII. THE DATA, MACHINE LEARNING, AND INFERENCE BARRIER TO ENTRY ................. 46
`A.
`Data Aggregation, Centralization, and Infrastructure Effects ........................................... 47
`B.
`The Effects of LinkedIn’s Machine-Learning Systems ..................................................... 50
`C.
`The Effects of Inference .................................................................................................... 54
`VIII. LINKEDIN ANTICOMPETITIVELY MAINTAINS ITS MONOPOLY ................................... 56
`LinkedIn Enters into Agreements to Anti-Competitively Sell User Data to “Partners”
`A.
`Through Its Privatized APIs .............................................................................................. 56
`
`VI.
`
`
`
`
`
`i
`Class Action Complaint
`
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 3 of 118
`
`B.
`
`C.
`
`IX.
`
`X.
`
`
`
`LinkedIn Employs Technical Countermeasures to Prevent Rivals or Potential Entrants
`from Obtaining Even Data Users Have Decided to Make Public ..................................... 62
`LinkedIn Strengthened the Data, Machine Learning, and Infrastructure Barrier to
`Entry by Integrating with Parent Microsoft’s Azure Cloud and AI Systems .................... 64
`LINKEDIN AGREES WITH FACEBOOK TO DIVIDE THE MARKET .................................. 67
`A.
`LinkedIn and Facebook Were Poised to Compete Before 2015 ....................................... 68
`B.
`Internally, LinkedIn and Facebook Secretly Negotiated an Agreement ........................... 71
`C.
`After the Negotiations, Facebook Never Entered the Market with a Rival Product ......... 74
`D.
`Facebook Has a Practice of Dividing Markets with Potential Competitors, which Is
`Consistent with the Apparent Agreement Between LinkedIn and Facebook .................... 79
`THE RELEVANT MARKETS ..................................................................................................... 81
`A.
`The Professional Social Networking Market ..................................................................... 81
`1.
`The Professional Social Networking Market Is a Distinct Submarket. ..... 83
`2.
`LinkedIn Possesses Monopoly and Market Power in the Professional
`Social Networking Market ......................................................................... 94
`Relevant Geographic Market ..................................................................... 97
`3.
`Barriers to Entry ........................................................................................ 98
`4.
`HARM TO COMPETITION ......................................................................................................... 99
`XI.
`CLASS ACTION ALLEGATIONS ........................................................................................................ 102
`CLAIMS FOR RELIEF ........................................................................................................................... 107
`PRAYER FOR RELIEF .......................................................................................................................... 114
`JURY DEMAND ..................................................................................................................................... 115
`
`
`
`
`
`ii
`Class Action Complaint
`
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 4 of 118
`
`
`
`
`INTRODUCTION
`
`1.
`
`This lawsuit concerns the unlawful monopolization of the professional social networking
`
`market by defendant LinkedIn. Plaintiffs are LinkedIn Premium subscribers who have been overcharged
`
`due to LinkedIn’s unlawful conduct, which has enabled LinkedIn to extract supracompetitive profits from
`
`its subscribers through inflated subscription prices and data sale revenues.
`
`2.
`
`After emerging as the unchallenged leader in the professional social networking market
`
`through what its founder Reid Hoffman called “Blitzscaling”—the rapid race to capture network and
`
`lock-in effects by scaling at any cost—LinkedIn quickly turned to protecting and monetizing its position,
`
`including by using sophisticated data acquisition and analysis to maximize user attention and revenues.
`
`By 2015, LinkedIn’s subscription business was protected by a powerful barrier to entry, which was the
`
`net sum of LinkedIn’s data centralization and aggregation, its machine learning and AI infrastructure,
`
`and the inferred data it produced. This Data, Machine Learning, and Inference Barrier to Entry
`
`(“DMIBE”) became LinkedIn’s greatest asset, and in 2016 drew a $26.5 billion acquisition of the
`
`company by Microsoft—owner of one of world’s largest and most powerful arsenals of massively
`
`scalable on-demand computational hardware. Combining LinkedIn’s unrivaled professional data trove
`
`and infrastructure with its parent Microsoft’s high-end cloud computing arrays, the companies are
`
`developing an AI and machine-learning-backed monopoly of enormous scale—fortifying and profoundly
`
`strengthening the DMIBE. At the time of this Complaint, the DMIBE represents a near-insurmountable,
`
`and growing, barrier to meaningful entry in the professional social networking market, let alone entry at
`
`sufficient scale to effectively check LinkedIn’s pricing and subscription terms.
`
`3.
`
`Since the Microsoft acquisition, LinkedIn has engaged in affirmative anticompetitive
`
`conduct that has strengthened (and continues to strengthen) the DMIBE, that has reduced consumer
`
`choice, and that has allowed LinkedIn to charge and maintain inflated Premium subscription prices and
`
`
`
`
` 1
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 5 of 118
`
`
`
`subscription terms with no competitive check. This conduct has prevented—indeed, effectively
`
`precluded—entry by others into the professional social networking market, insulating prices from
`
`competition. Among this anticompetitive conduct was (and is): (i) LinkedIn’s non-optional sale of
`
`Premium user data to unnamed “partners,” which forcibly grafts a negative value feature—one that
`
`materially harms competition in the professional social networking market and at the same time lacks
`
`measurable procompetitive effects, even aside from its lack of consumer benefit—onto LinkedIn’s
`
`Premium subscription product; (ii) deploying sophisticated technological countermeasures specifically
`
`designed by LinkedIn to prevent users’ public data from being accessed by potential or actual
`
`competitors, thereby maintaining and fortifying the DMIBE and hindering potential entry at scale; (iii)
`
`aggressively integrating LinkedIn’s unmatched professional social networking data repository and
`
`pipeline and its powerful AI and machine-learning data and infrastructure with its parent company
`
`Microsoft’s Azure cloud servers and arrays of Graphical Processing Units (“GPUs”)—an internationally
`
`scarce hardware resources necessary for complex AI and machine learning computation at scale; and (iv)
`
`expressly or tacitly dividing markets with LinkedIn’s most natural potential competitor, Facebook—an
`
`agreement that, as explained in detail in this complaint, apparently continues to this day.
`
`4.
`
`Plaintiffs seek trebled damages for the price overcharge they have experienced (and
`
`continue to experience) for LinkedIn Premium subscriptions due to LinkedIn’s monopolization of the
`
`professional social networking market. They also seek injunctive relief to stop LinkedIn’s anticompetitive
`
`conduct, including, among other things, injunctive relief allowing Premium subscribers to opt out of
`
`LinkedIn’s parasitic data sale to unnamed partners and injunctive relief halting and unwinding the
`
`unprecedented and anticompetitive integration of LinkedIn’s professional data, machine learning, and AI
`
`infrastructure with Microsoft’s powerful cloud computing hardware. Absent abatement by this Court,
`
`
`
`
` 2
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 6 of 118
`
`
`
`LinkedIn’s will continue to fortify and strengthen the DMIBE, potentially sealing off the professional
`
`social networking market from competition for years and perhaps decades to come.
`
`*
`
`*
`
`*
`
`5.
`
`Reid Hoffman, the enigmatic co-founder of PayPal, took away an important insight from
`
`his wildly successful early venture: rapidly scaling a online business can create powerful network effects,
`
`and those network effects can be durable. Hoffman didn’t just take this lesson to heart; he designed his
`
`professional life around it; evangelized it to everyone who would listen; and spent more than a decade
`
`trying to harness this phenomenon on his own. Hoffman called his business strategy of deploying massive
`
`first-mover resources to capture durable network effects, “Blitzscaling.”
`
`6.
`
`As Hoffman explained about his success at PayPal, “the faster we got to scale, the stronger
`
`we created network effects. . . .” In 2003, Hoffman took this key insight and applied it to his new venture,
`
`a professional social network called LinkedIn.
`
`7.
`
`Hoffman rapidly scaled LinkedIn, and by the middle of 2005, the company had 1.7 million
`
`users. By the summer of that same year, LinkedIn’s user base had doubled to 3.3 million. In Hoffman’s
`
`pursuit of blitzscaling for LinkedIn, nothing was off limits—even the reviled (and likely unlawful) tactic
`
`of commandeering LinkedIn users’ contact lists and spamming those contacts with invitations to
`
`Hoffman’s new venture.
`
`8.
`
`Hoffman recognized that LinkedIn’s exponential growth was the result of powerful
`
`network effects stemming from user growth, the inelastic ubiquity of job hunting and professional
`
`connection, and LinkedIn’s role as a source of professional identity. As Hoffman explained, “Just one of
`
`these network effects would probably be enough to create first-scaler advantage; all three working
`
`together built a massive strategic moat that protected the LinkedIn business from any new entrants.”
`
`
`
`
` 3
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 7 of 118
`
`
`
`After a few years, LinkedIn had become the de facto standard for online professional
`
`9.
`
`identity, and the unchallenged market leader in professional social networking, both domestically and
`
`worldwide. But although LinkedIn had rapidly, and in some sense successfully, “blitzscaled,” two
`
`significant obstacles remained: user engagement and monetization.
`
`10.
`
`On the issue of monetization, the specific market niche LinkedIn had grown to dominate
`
`was (and is) unique from other social networking markets: professional social networking participants
`
`will pay upfront for important business services like advanced hiring tools and direct connections to other
`
`businesspeople. LinkedIn introduced Premium subscription products, and quickly accreted subscribers
`
`and subscriber revenue. But by the mid-2010s, as data aggregation and monetization technology—fueled
`
`by the rise of powerful machine learning and artificial intelligence tools, including powerful GPU arrays
`
`available in the cloud—transformed the business of high technology, LinkedIn realized that its monopoly
`
`could be supercharged (and supra-monetized) by leveraging data anticompetitively, including against its
`
`own Premium subscribers.
`
`11.
`
`12.
`
`That is exactly what LinkedIn did—and continues to do.
`
`As the 2010s reached their midpoint, it was clear to LinkedIn that the real profit-center in
`
`its business going forward was data-driven—and the company took a series of coordinated actions to
`
`leverage its market dominance in professional social networking to develop, deploy, and weaponize a
`
`massive-scale data acquisition, machine learning, and artificial intelligence apparatus that would both
`
`supercharge its overall profits, but also fortify and maintain the barrier to entry around its business. not
`
`only to serve content to users that drove engagement, but to monetize user data for profit.
`
`13.
`
`By 2015, LinkedIn had begun developing cutting edge machine learning and artificial
`
`intelligence infrastructure that could algorithmically serve content to users that drove engagement—and
`
`monetize user data for profit. That infrastructure included the vectorization, centralization, and
`
`
`
`
` 4
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 8 of 118
`
`
`
`structuring of user data; the streamlining of real-time updates of user data and the development of
`
`machine learning and artificial intelligence models; the collection of detailed telemetry information
`
`derived from user interactions with LinkedIn’s app; and a massive trove of “inferred” data about users—
`
`data derived from predictions by machine learning and artificial intelligence models.
`
`14.
`
`As these new data tools and practices began to transform LinkedIn’s backend, a powerful
`
`barrier to entry formed around the company’s business—exponentially more powerful, indeed, than the
`
`network effects that had facilitated LinkedIn’s blitzscaling in the first instance.
`
`15.
`
`This new, incredibly powerful barrier to entry comprised (and comprises) three primary
`
`aspects: (1) LinkedIn’s data centralization, (2) LinkedIn’s machine learning models, and (3) the resulting
`
`trove of inferred data. What’s more, these three aspects of LinkedIn’s business together reinforced—and
`
`reinforce—each other, resulting in a Data, Machine Learning, and Inference Barrier to Entry (“DMIBE”).
`
`16.
`
`Standing alone, LinkedIn’s AI and machine-learning juggernaut was already extremely
`
`valuable (and ripe for misuse, given LinkedIn’s monopoly position in professional networking)—but in
`
`the hands of a company with a ubiquitous presence in the office, it could become an impenetrable source
`
`of monopoly rents for decades. This is precisely what happened after office software and cloud computing
`
`giant Microsoft acquired LinkedIn.
`
`17.
`
`By the middle of 2016, Microsoft, which had dominated office productivity software for
`
`decades, saw LinkedIn as a missing piece of its new business—artificial intelligence and machine
`
`learning in the cloud. LinkedIn needed more user engagement, and Microsoft’s cloud infrastructure and
`
`direct input into corporations around the world created powerful synergies.
`
`18.
`
`In June 2016, Microsoft announced that it would acquire LinkedIn for $26.2 billion—then
`
`the largest acquisition in Microsoft’s history. As Microsoft’s CEO Satya Nadella explained, the purpose
`
`of the merger was the “coming together of the professional cloud and the professional network.”
`
`
`
`
` 5
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 9 of 118
`
`
`
`As the Financial Times described the merger, the acquisition was the product of “data
`
`19.
`
`gravity”—the “tendency for large bodies of data to attract suppliers of services and applications.” Despite
`
`concerns raised by Europe’s competition chief; the CEO of Salesforce, Marc Benioff; and others, the
`
`FTC did not challenge the acquisition out of a fear that if it challenged any merger based on data
`
`aggregation, it would have to challenge a host of other similar deals.
`
`20.
`
`In Microsoft’s hands, the DMIBE protecting LinkedIn’s monopoly in the professional
`
`social networking market became even more powerful. LinkedIn quickly began taking measures to
`
`further strengthen the DMIBE to keep out new entrants and competitors, but also began rapid integration
`
`of LinkedIn’s AI and machine-learning systems with Microsoft’s cloud computing service, Azure.
`
`21.
`
`From 2017 to the present, LinkedIn aggressively protected the DMIBE, including through
`
`anticompetitive conduct. Among them are four primary courses of action, which have irreparably sealed
`
`off LinkedIn from competition.
`
`22.
`
`First, LinkedIn aggregates and structures user data, then sells that data through its
`
`application programming interfaces (“APIs”) to private, undisclosed “Partners.” This paid access creates
`
`public-facing, permissioned endpoints that provide external access to LinkedIn’s massive trove of user
`
`data, which has been carefully structured, cleaned, and centralized for programmatic use and
`
`consumption.
`
`23.
`
`These data sale agreements with Partners dangerously expose user data. LinkedIn has no
`
`control over the data while it is in the hands of its partners; it does not disclose who its “partners” are
`
`with respect to the LinkedIn APIs; and the APIs create a massive security vulnerability by centralizing
`
`large amounts of user data and exposing that data to outward-facing interfaces. Indeed, massive troves of
`
`data corresponding to LinkedIn users—data purportedly gleaned at least in part from LinkedIn APIs—
`
`are currently being sold on the dark web.
`
`
`
`
` 6
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 10 of 118
`
`
`
`LinkedIn’s data sale to Partners provides no value to Premium subscribers, and those
`
`24.
`
`subscribers cannot opt out of the data sale—or subscribe to a Premium product that does not include the
`
`sale of their data. At the same time, this data sale reinforces the DMIBE and allows LinkedIn to extract
`
`monopoly rents from subscribers and from contracting API Partners. Moreover, by exclusively partnering
`
`with certain consumers of its API, LinkedIn ensures that a new entrant—or even a firm that does not have
`
`a contract with LinkedIn—cannot surmount the DMIBE. These data sales reduce consumer choice,
`
`provide no pro-competitive effects for consumers, inflate prices, and prevent competition and entry,
`
`including by reinforcing the DMIBE.
`
`25.
`
`Second, LinkedIn protects the DMIBE and maintains its monopoly by deploying
`
`technological countermeasures that prevent any other party, competitor, or potential entrant from
`
`accessing public data on its site—even data that users want to make public. Indeed, a LinkedIn page is a
`
`public-facing professional resume and identity. Many (if not most) users elect to make these profiles
`
`public. LinkedIn, however, uses technological countermeasures, including technology it calls Fuse and
`
`Org Block, to prevent access to even public data. At the same time, LinkedIn whitelists certain companies
`
`such as Google, exempting them from LinkedIn’s technological countermeasures. By doing this,
`
`LinkedIn hoards millions of public identities provided by premium subscribers, preventing any potential
`
`entrant or competitor from developing a rival product. This has ensured that no price check has appeared
`
`or imminently will appear, allowing LinkedIn to maintain its DMIBE, monopoly, and unchallenged
`
`subscription prices.
`
`26.
`
`Third, despite maintaining its own servers before 2016, since the Microsoft acquisition
`
`LinkedIn has integrated—and is currently integrating—its standardized and structured trove of user data
`
`(and inferred data) with Microsoft’s powerful AI technology and Graphics Processing Unit (“GPU”)
`
`hardware, which it provides through its Azure cloud computing business. These cloud-based arrays of
`
`
`
`
` 7
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 11 of 118
`
`
`
`GPUs are a scarce resource and are necessary for developing cutting-edge AI and machine learning
`
`models, including large language models. Only Google, Amazon, and Microsoft have such cloud-based
`
`hardware, and LinkedIn’s integration of its massive data infrastructure with Microsoft’s cutting-edge
`
`hardware, including NVIDIA A100 Tensor Core GPUs, creates a full-scale AI and machine-learning
`
`juggernaut acquiring, processing, analyzing, and re-analyzing LinkedIn’s unique pipeline of professional
`
`social networking data—permanently sealing off LinkedIn from competition, let alone competition at
`
`scale.
`
`27.
`
`Upon completion of the integration of LinkedIn’s AI and machine learning tools and data
`
`infrastructure with Azure, the DMIBE will be irreversibly strengthened, and LinkedIn’s integration
`
`efforts to date have already significantly fortified the DMIBE. This integration lacks procompetitive
`
`benefits, let alone benefits that outweigh the massive anticompetitive effects in the Professional Social
`
`Networking market. The net result is that LinkedIn will become virtually unchallengeable, and its
`
`subscription prices and terms will remain unchecked for years—and perhaps decades—to come.
`
`28.
`
`Finally, LinkedIn agreed with its most obvious natural competitor—social networking
`
`juggernaut Facebook—to divide markets, sealing off the last remaining source of potential rivalry to
`
`LinkedIn’s professional social networking dominance. As a result of this agreement, which appears to
`
`have grown of data access negotiations between the two companies in the early 2010s, LinkedIn has
`
`maintained its monopoly in professional social networking without the threat of entry by Facebook, and
`
`Facebook has fortified its dominance and control over personal social networking, perhaps through data
`
`assistance from LinkedIn—similar to an agreement that Facebook recently struck with its would-be
`
`competitor Google.
`
`29.
`
`By 2015, competition between Facebook and LinkedIn seemed inevitable. Press
`
`speculation that Facebook was building a professional social network to rival LinkedIn had reached a
`
`
`
`
` 8
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 12 of 118
`
`
`
`fever pitch, as Facebook secretly developed a new product called Facebook at Work. For example, in
`
`November 2014, the Financial Times reported based on communications with “individuals familiar with
`
`the matter” that Facebook at Work would “allow users to chat with colleagues, connect with professional
`
`contacts and collaborate over documents, competing with Google Drive and Microsoft Office.” In
`
`December 2015, Forbes predicted that “LinkedIn could face intense competition from ‘Facebook at
`
`Work,’” which was at the time to be imminently released.
`
`30.
`
`Internally, Facebook was preparing for scorched earth competition with LinkedIn.
`
`According to publicly reported internal Facebook documents, Facebook had identified LinkedIn as a
`
`competitive threat, and was threatened by LinkedIn’s access to Facebook user data through Facebook
`
`APIs. According to Facebook Vice President Chris Daniels, representatives of the companies had met in
`
`2013, and LinkedIn had agreed with Facebook “not to access [Facebook’s] APIs until [the companies]
`
`worked out an agreement both ways . . . .” Nonetheless, LinkedIn appeared to be aggressively scraping
`
`Facebook user data as part of an arms race for user identities.
`
`31.
`
`Facebook’s senior executives scrambled to audit all of the applications using Facebook’s
`
`Platform APIs, bucketing each of the thousands of apps by whether they were competitive threats.
`
`Facebook identified LinkedIn as a competitive threat in the “reputation” category. After the audit,
`
`Facebook prepared to remove API functionality, breaking thousands of competitive apps, sparing only
`
`companies that entered into secret agreements with Facebook. Facebook internally debated what it would
`
`do about LinkedIn, including whether it would demand data from LinkedIn in exchange for its access to
`
`Facebook’s APIs beyond “status updates.” As Facebook Vice President Konstantinos Papamiltiadis
`
`internally stated, such a meager amount of data from LinkedIn was “not good enough for [him].”
`
`Facebook also sought a bargaining chip in its negotiations with LinkedIn, including a trumped-up policy
`
`
`
`
` 9
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 13 of 118
`
`
`
`violation that it could use as a fig leaf for denying LinkedIn access to its APIs and the removal of “work
`
`history” information from LinkedIn’s scope of access.
`
`32.
`
`Although the specific result of the then-ongoing negotiations between the companies is
`
`not known, what happened after Facebook removed API functionality for thousands of developers speaks
`
`volumes. From April 2015 until the present, LinkedIn suffered no public problems or deprecation after
`
`Facebook privatized its APIs, indicating to some degree of certainty that a deal of some sort was in fact
`
`struck between the companies. And the other side of the equation—Facebook’s behavior—has been even
`
`more striking. Despite entering every adjacent market within its grasp, and seeking to leverage its
`
`professional social networking data and tools in products from messaging, to video, to gaming, and
`
`beyond, Facebook never entered the market for professional social networking. Even the expected
`
`Facebook at Work product, which was released under the name Workplace, had a glaring omission—it
`
`was released without any professional social networking functionality, and there was no competing
`
`subscription social networking product to rival LinkedIn. Facebook had carefully excised from its
`
`Workplace product features that would compete with LinkedIn’s professional social networking products
`
`and services.
`
`33.
`
`From 2016 until the present, Facebook aggressively diversified its lines of business,
`
`entering virtually every technology market. Facebook challenged ad giant Google in internet advertising
`
`and user tracking; Facebook acquired Instagram, directly competing with photo sharing and
`
`microblogging apps, including Twitter; Facebook aggressively moved into video sharing and streaming
`
`with its Facebook Live product; it launched Messenger and even spent billions acquiring WhatsApp;
`
`Facebook made several forays into payments, taking on the likes of PayPal and Venmo; Facebook took
`
`on eBay and Amazon with its Marketplace product; Facebook planned its own crypto currency; and
`
`Facebook even bought a virtual reality company, Oculus.
`
`
`
`
` 10
`Class Action Complaint
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`30
`31
`
`

`

`Case 4:22-cv-00237-KAW Document 1 Filed 01/13/22 Page 14 of 118
`
`
`
`The only market it did not even attempt to enter was the one LinkedIn controlled—a
`
`34.
`
`market where users paid expensive subscriptions out of pocket for professional social networking. That
`
`market could have supported an additional competitor, and Facebook could have undercut prices to obtain
`
`market share, but Facebook never set foot near LinkedIn’s monopoly.
`
`35.
`
`This lack of entry is not reasonably explicable absent an agreement between LinkedIn and
`
`Facebook not to compete with each other, particularly given the state of the companies’ negotiations in
`
`2013 to 2015, which included potential data reciprocity terms.
`
`36.
`
`To this day, both companies deny that they compete with each other. Indeed, even though
`
`Facebook internally considered LinkedIn potentially competitive in late 2013, and it was widely reported
`
`that Facebook was preparing to launch a professional social networking product in 2015, Facebook CEO
`
`and founder Mark Zuckerberg told the U.S. Senate in 2018 that Facebook “do[es]n’t consider LinkedIn
`
`to be one of our direct competitors.” LinkedIn did not list Facebook as a competitor in its SEC filing
`
`when it went public in 2015.
`
`37.
`
`All of this evidence indicates that LinkedIn and Facebook