IntegrityMessageBoards.com v. Facebook, Inc.

Opinion

Case No. 18-cv-05286-PJH Plaintiff,

v. ORDER DENYING MOTION TO STRIKE, DENYING MOTION FOR FACEBOOK, INC., CLASS CERTIFICATION, AND GRANTING IN PART AND DENYING Defendant. IN PART MOTIONS TO SEAL Re: Dkt. Nos. 155-3, 156, 158, 169,

176, 179, 184

Plaintiff IntegrityMessageBoards.com’s (“plaintiff” or “IMB”) motion for class certification (Dkt. 155-3) and motion to strike (Dkt. 176) came on for hearing on April 29, 2021. Plaintiff appeared through its counsel, Steven F. Molo. Defendant Facebook, Inc. (“defendant”) appeared through its counsel, Christopher Chorba. Having read the papers filed by the parties and carefully considered their arguments and the relevant legal authority, and good cause appearing, the court hereby DENIES both the motion for class certification and motion to strike. The court also GRANTS IN PART and DENIES IN PART defendant’s motions to seal (Dkts. 158, 169, 184) various materials filed in connection with the motion for class certification and motion to strike. Defendant operates Facebook. Dkt. 148 (first amended complaint (“FAC”)) ¶ 19. Facebook is a social media platform with two billion users. Id. ¶¶ 19-20. The platform permits users to share content. Id. ¶ 19. Defendant does not charge users for access to or use of its platform. Id. ¶ 21. Instead, third party “advertisers,” i.e., businesses, pay Years ago, plaintiff paid defendant approximately $1,500 to display its advertisements to users on Facebook. Id. ¶¶ 49-69. The instant putative class action followed. Dkt. 1. In its operative complaint, plaintiff alleges that defendant has misled advertisers about its ability to accurately deliver advertisements to users according to an advertiser’s specifications. FAC ¶¶ 2-4, 8. Plaintiff alleges that defendant has done so through some combination of its: (1) public remarks, id. ¶¶ 19-25; (2) statements on its website, id. ¶¶ 26-34; and (3) representations contained in defendant’s so-called “targeting interface” programs, id. ¶¶ 35-43. Advertisers use the targeting interface programs to create and publish their advertisements. Id. ¶¶ 6, 19-43 Plaintiff alleges that, despite defendant’s statements, defendant cannot, in fact, display advertisements to users according to an advertiser’s specifications. Id. ¶¶ 2-4, 8, 78-108. Plaintiff explains that the reason for defendant’s inability to deliver on its purported statements is straightforward: defendant relies on data that does not accurately capture the actual attributes of users. Id. ¶¶ 8, 88-108. What’s worse, plaintiff says, is that defendant’s agents understood and appreciated that inability. Id. ¶¶ 78-80, 124-31. And plaintiff does not make that contention lightly. To substantiate it, plaintiff cites a “writeup” sent on February 16, 2016 by defendant’s ads targeting team to its Vice President of Ads, Andrew Bosworth (“Bosworth”). Id. ¶¶ 8, 127. The court will refer to that write-up as the “Bosworth report.” That report informed Bosworth that, in the United States, the “precision” of a key sort of user attribute criteria that advertisers may select when designing a target audience (namely, “Interests” criteria): is only 41% - that means more than half the time we’re showing ads to someone other than the advertisers’ intended audience. And it is even worse internationally. We don't feel we’re meeting expectations today and want to invest this half to grow interest precision. Dkt. 155-8 at 264 (email), 269 (Bosworth report) (emphasis added). Based on the above, plaintiff alleges the following three claims against defendant: • Violation of California Business & Professions Code § 17200. Id. ¶¶ 143-53. • Common law fraud. Id. ¶¶ 163-70. Plaintiff premises each of these claims on defendant’s representations (affirmative statements and omissions) about defendant’s “intention and ability to accurately display ads to users who matched the targeting criteria advertisers selected through Facebook’s targeting interface.” Id. ¶¶ 145, 155, 164. The supposed misstatements contained in the targeting interface programs are at the heart of this motion. Given that, the court will first explain what those programs are and how they work. I. The Ads Manager Website Targeting Interface Presently, defendant offers five different self-serve digital programs for advertisers to create, publish, and track their advertisements. Dkt. 169-9 ¶ 6. These programs include the following: • Ads Manager website interface. • Ads Manager App. • Facebook pages themselves. • Facebook Business Suite. • Automated Ads. Id. Throughout its motion, plaintiff generally refers to all five of these programs as defendant’s “targeting interface.” At oral argument, however, plaintiff’s counsel acknowledged plaintiff’s ambiguous use of that term in the motion. He requested that the court “construe the class definition . . . to be limited to the class [using] the Ads Manager [website] interface.” Dkt. 192 at 33-34. The court will adopt that construction for purposes of these motions. Given that, the court will focus on describing the Ads Manager website interface and detail its counterparts only as necessary. The Ads Manager website interface is the interface that advertisers most commonly use. Dkt. 169-9 ¶ 6. To create an advertisement using the Ads Manager website interface, an advertiser must follow various steps. At the first step, the advertiser selects a “campaign objective.” Dkt. 170-1 at 29-33. Those objectives are aimed at advertiser. Id. at 30. At the second step, an advertiser selects a so-called “target audience.” To do so, the advertiser chooses among various categories of criteria describing the user who the advertiser wants to see its advertisements. Id. at 34-44. Those categories include the user’s physical location, age, gender, and language. Id. Other categories include a so-called “Detailed Targeting” category. Id. That category includes (or included) the following three sub-categories: (1) “Interests”; (2) “Behaviors”; and (3) “Partner Categories.” Id. at 37-44. The Ads Manager website interface describes the first two sub-categories as follows: • For Interests: “Reach specific audiences by looking at their interests, activities, the Pages they have liked and closely related topics.” Id. at 37. • For Behaviors: “Reach people based on purchase behaviors or [sic], device usage and more. Some behavior data is available for US audiences only.” Id. at 38. In 2018, defendant discontinued use of its Partners Categories sub-category. Dkt. 169-5 at 46. That category concerned certain user financial demographics, id. at 36; Dkt. 155-8 at 61, including, for example household income, Dkt. 157-1 at 192. When using the Ads Manager website interface, an advertiser could type into a search bar to identify various potential Interests, Behaviors, and Partner Categories. Id.; Dkt. 170-1 at 37-38. When searching and browsing those sub-categories, the interface includes a “hovercard” description of the users corresponding to a criterion. Id. at 38. The hovercard notes how defendant determines whether a user falls within a particular Interest, Behavior, or Partner Category criterion. For example, a hovercard for an Interest in “Cats” states “people who have expressed an interest in or like pages related to Cats.” Id. at 39. A hovercard for a “Travel” Behavior states “People who have traveled abroad more than once in the past 6 months.” Id. at 40. A hovercard for the “$150,000-$250,000” Income Demographic states “People in households that have an estimated household income of between $150K and $250k.” Dkt. 157-1 at 192. meet the criterion’s description. Dkt. 170-1 at 39 (527,342,350 users who have an Interest in Cats); Id. at 40 (677,733,885 users who have a Travel Behavior); Dkt. 157-1 at 192 (4,266,100 users who fit within the $150,000-$250,000 Income Demographic). Defendant’s ability to accurately target users who fit within any particular criterion of the above Interests, Behaviors, and Partner Categories is a major focus of this litigation. Once an advertiser selects the criteria defining its target audience, the Ads Manager website interface generates a “potential reach” estimate showing the number of users who purportedly fall within the target audience. Dkt. 155-8 at 163-84. Often (but not always), the interface would display that estimate alongside an “audience definition” gauge. Id. That gauge signaled whether a target audience was “specific” or “broad.” Id. At the third step, an advertiser selects a payment method. Dkt. 170-1 at 44. From there, the interface directs an advertiser to a box stating: “By clicking the ‘Publish’ button, you agree to Facebook’s Terms and Advertising Guidelines.” Id. At that point, an advertiser may then choose between “Close,” “Back,” and “Publish.” Id. An advertiser may elect for defendant to charge it at least two ways: (1) when a user viewing its advertisement acts in a way the advertiser hoped the user would act; or (2) when a user is shown an advertisement (referred to as “impressions”). Dkt. 169-9 ¶ 17. II. Plaintiff’s Advertisement Campaigns1 Plaintiff runs an online forum, “Investor Village.” FAC ¶ 44. Investor Village offers a space for investors to discuss stock sales. Id. Plaintiff’s “managing member” is Ralph Kidd (“Kidd”). Id. ¶ 46. Plaintiff has a Facebook page for Investor Village. Id. In June 2015, Kidd browsed Facebook. Id. Defendant showed Kidd a “teaser” advertisement depicting what an actual advertisement for Investor Village could look like. Id. Kidd bit. Id. He was redirected to the Ads Manager website interface. Id. Kidd “browsed” that interface. Id. He “saw various representations regarding the 1 Originally, the FAC named Retour, Inc. (“Retour”) as a second plaintiff in this action. FAC ¶ 116. However, on September 29, 2020, Retour filed a notice voluntarily display of ads to users matching an advertiser’s targeting criteria.” Id. “Based on those representations,” Kidd “concluded that Facebook’s targeting interface presented a unique opportunity to target ads to IMB’s demographic of highly compensated and educated investors.” Id. Kidd “found the promise of using Facebook’s targeting interface to narrowly define a specific audience to be very attractive.” Id. ¶ 47. Plaintiff mounted three advertisement campaigns on Facebook. Id. ¶¶ 49-69. Only the latter two campaigns are relevant to this motion. Id. ¶ 48. According to plaintiff, the “objective” of each campaign was to increase the number of “likes” that its Facebook pages received from users who fit its target audience. Id. ¶¶ 50-51, 63-67. Using the Ads Manager website interface, plaintiff defined its second campaign’s target audience as users with all of the following characteristics: (1) located in the United States; (2) with Interests in “Investment”; (3) with a college degree; (4) who earn an annual household income above $250,000; (5) who own a home; and (6) are over 45 years old. Id. ¶ 50. Plaintiff defined its third campaign’s target audience with materially similar attributes, except it did not call for users with an Investment Interest. Id. ¶ 63. Plaintiff observed the progress for its campaigns. Id. ¶¶ 54-60, 67. Plaintiff noticed that various users who “liked” its pages lacked some characteristics that plaintiff selected in its target audiences. Id. ¶¶ 54, 58, 67. For its second campaign, plaintiff paid defendant based on “impressions.” Id. ¶ 49. For its third campaign, plaintiff paid defendant based on the “likes” that defendant’s advertising generated for plaintiff’s pages. Id. ¶ 66. In total, plaintiff paid defendant $1,409.68 for its second campaign and $242.17 for its third campaign. Id. ¶ 62. The court will detail further factual background as necessary in its analysis below. III. Relevant Procedural Events and the Instant Motions On July 27, 2020, defendant moved to dismiss the FAC’s § 17200 claim, as well as all requests for equitable relief. Dkt. 118. On November 6, 2020, the court granted in part and denied in part defendant’s motion. Dkt. 147. Importantly, the court dismissed past harm (i.e., restitution and disgorgement). Id. at 10. The court permitted plaintiff to pursue such relief to extent premised on future harm. Id. at 13. Thus, in light of the court’s November 6, 2020 order, plaintiff’s requests for monetary damages (actual, nominal, and punitive) as well as its requests for declaratory and injunctive relief are live. On December 23, 2020, plaintiff filed the instant motion for class certification. Dkt. 155-3 (sealed); Dkt. 156 (redacted). In it, plaintiff asks the court to certify a class under Rule 23(b)(2) and Rule 23(b)(3) comprising the following: All individuals or entities within the United States who, from August 28, 2014 to the present, targeted Facebook users in one or more “interest,” “behavior,” or “partner” categories selected using Facebook’s self-serve targeting interface, and who paid Facebook for at least one ad for which they neither (a) expressly authorized Facebook to disregard their targeting criteria by opting into “Target Expansion” nor (b) elected to be charged only when a user made a purchase or downloaded an app. Dkt. 155-3 at 3. On March 3, 2021, defendant filed its opposition to that motion. Dkt. 169-5 (sealed); Dkt. 170 (redacted). On March 24, 2021, in response to certain declarations proffered by defendant in support of its opposition, plaintiff filed a motion to strike. Dkt. 176. The court will detail and analyze that motion at Section I. When briefing these motions, the parties filed more than 1,000 pages of documents under seal. Dkt. 155 (plaintiff’s motion to provisionally seal its motion for class certification opening brief materials); Dkt. 169 (defendant’s motion to permanently seal its motion for class certification opposition materials); Dkt. 179 (plaintiff’s motion to provisionally seal its motion for class certification and motion to strike reply materials). Defendant filed three motions to permanently seal some of that information. Dkt. 158 (motion to permanently seal motion for class certification opening materials); Dkt. 169 (noted above); Dkt. 184 (motion to permanently seal motion for class certification and motion to strike reply materials). The court will address these motions at Section III. I. Motion to Strike certification, defendant disclosed to plaintiff ten additional witnesses as part of its second supplemental initial disclosure. Dkt. 177 ¶ 9. These ten witnesses comprise advertisers, advertisement agencies, advertising consultants, and defendant’s marketing partners. Dkt. 170-5 at 169-205. To support its opposition to the motion for class certification, defendant offers a declaration from each witness (the “challenged declarations”). Id. Defendant primarily offers this evidence to refute plaintiff’s Rule 23(b)(3) predominance showing. Dkt. 169-5 at 32-34, 40-41. In its motion, plaintiff asks the court to strike the challenged declarations under Rule 37(c) for failure to timely disclose. Dkt. 176 at 2-3. The court denies the motion to strike. On September 23, 2020, the court approved a stipulation between the parties to extend various class certification-related deadlines. Dkt. 143. At footnote one, that stipulation states: [B]oth parties reserve the right to introduce rebuttal evidence in their opposition and reply briefs that they could not have reasonably anticipated by [the extended initial disclosure deadline] that they would need. . . . The parties agree to produce the information as soon as it becomes clear they will need to rely on this information. Id. at 4-5 n.1. The challenged declarations qualify as the sort of rebuttal evidence contemplated by the above footnote. Plaintiff has altered its theory of liability throughout this litigation. For example, in its motion for class certification, plaintiff advances a proposed class definition that is different than that alleged in its FAC. Compare Dkt. 155-3 at 3 with FAC ¶ 132. Shifts like that support finding that defendant could not have anticipated that it would need the challenged declarations until after it had the opportunity to review plaintiff’s opening brief. Accordingly, the court will consider the challenged declaration in its analysis of plaintiff’s motion for class certification. II. Motion for Class Certification A. Legal Standard A proposed class must satisfy Rule 23(a)’s numerosity, commonality, typicality, prerequisites of Rule 23(a) are satisfied, the court then determines whether to certify the class under one of the three subsections of Rule 23(b), which requires plaintiffs to show either: (1) a risk of substantial prejudice from separate actions, (2) that declaratory or injunctive relief benefitting the class as a whole would be appropriate, or (3) that common questions of law or fact to the proposed class predominate and that a class certification is superior to other methods available for adjudicating the controversy at issue. Fed. R. Civ. Pro. 23(b)(1)-(3). “Before certifying a class, the trial court must conduct a rigorous analysis to determine whether the party seeking certification has met the prerequisites of Rule 23.” Mazza v. American Honda Motor Co. Inc., 666 F.3d 581, 588 (9th Cir. 2012). To the extent there are “any factual disputes necessary to determine” a Rule 23 criterion, a district court is required to resolve them. Ellis II v. Costco Wholesale, 657 F.3d 970, 983 (9th Cir. 2011) (emphasis in the original). The party seeking certification bears the burden of proof in demonstrating that it has satisfied all four Rule 23(a) requirements and that their action falls within one of the three types of actions permitted under Rule 23(b). Zinzer v. Accufix Research Inst., Inc., 253 F.3d 1180, 1186 (9th Cir. 2001). B. Analysis The court finds that plaintiff satisfies Rule 23(a)’s numerosity and commonality requirements. However, the court concludes that plaintiff failed to proffer sufficient evidence showing that it satisfies the remaining Rule 23(a)’s typicality and adequacy requirements. The court also concludes that plaintiff failed to show that common questions predominate to justify certification under Rule 23(b)(3) or that the court can issue injunctive relief on a classwide basis to justify certification under Rule 23(b)(2). 1. Numerosity Rule 23(a)(1) requires that a proposed class be so numerous that joinder of all members is impracticable. Fed. R. Civ. Pro. 23(a)(1). While there is no fixed number members meets such requirement. Californians for Disability Rights, Inc. v. California Dep't of Transportation, 249 F.R.D. 334, 346 (N.D. Cal. 2008). In its opening brief, plaintiff proffers a statement by defendant’s Chief Executive Officer Mark Zuckerberg (“Zuckerberg”) indicating that, as of 2015, defendant had “more than 2.5 million active advertisers.” Dkt. 157-1 at 258. Defendant does not dispute that the proposed class is sufficiently numerous. Given the above, the court concludes that plaintiff showed that its proposed class satisfies the numerosity requirement. 2. Commonality Rule 23(a)(2) requires questions of law or fact common to the class. Fed. R. Civ. Pro. 23(a)(2). Under this requirement, plaintiff must “demonstrate that the class members have suffered the same injury,” not merely violations of “the same provision of law.” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 349-50 (2011). Given that, plaintiffs’ claims “must depend upon a common contention” such that “determination of [their] truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.” Id. “What matters to class certification . . . is not the raising of common questions—even in droves—but rather the capacity of a classwide proceeding to generate common answers apt to drive the resolution of the litigation.” Id. To that end, the Ninth Circuit has explained that “plaintiffs need not show that every question in the case, or even a preponderance of questions, is capable of classwide resolution. So long as there is even a single common question, a would-be class can satisfy the commonality requirements of Rule 23(a)(2).” Wang v. Chinese Daily News, Inc., 737 F.3d 538, 544 (9th Cir. 2013). Given that, “where the circumstances of each particular class member vary but retain a common core of factual or legal issues with the rest of the class, commonality exists.” Evon v. Law Offices of Sidney Mickell, 688 F.3d 1015, 1029 (9th Cir. 2012). The court concludes that plaintiff satisfies the commonality requirement. As an initial matter, the parties agree that whether defendant had knowledge about the plaintiff to substantiate its claims. Compare Dkt. 155-3 at 12 with Dkt. 169-5 at 29. Plaintiff proffered evidence demonstrating that that element is capable of classwide resolution. As already noted, plaintiff proffered the Bosworth report to support its motion. Dkt. 155-8 at 264 (email), 269 (report). That report itself shows that, as of February 2016, a key corporate officer appreciated (or should have appreciated) the possibility or existence of material limits on defendant’s ability to target users by Interests. If defendant represented to a class member that its targeting ability on that sub-category exceeded 41 percent (and the report’s stated “precision” for that sub-category is true), then that report itself would show on the merits that defendant knew its representation was false. Given that a single question capable of classwide resolution may satisfy the commonality requirement, Wang, 737 F.3d at 544, plaintiff has done so here. 3. Typicality Rule 23(a)(3) requires that the claims of the named plaintiffs be typical of those of the proposed class. Fed. R. Civ. Pro. 23(a)(3). “The test for typicality is whether other members have the same or similar injury, whether the action is based on conduct which is not unique to the named plaintiffs, and whether other class members have been injured by the same course of conduct.” Sandoval v. Cty. of Sonoma, 912 F.3d 509, 518 (9th Cir. 2018), cert. denied sub nom. Cty. of Sonoma, California v. Sandoval, 140 S. Ct. 142, 205 L. Ed. 2d 35 (2019). The Ninth Circuit has acknowledged that a party representative fails these requirements if the representative faces “unique defenses” that “threaten to become the focus of the litigation.” Hanon v. Dataproducts Corp., 976 F.2d 497, 508 (9th Cir. 1992) (collecting cases). Those defenses include issues that are “potentially dispositive” of the representative’s claims. In re Facebook, Inc., PPC Advert. Litig., 282 F.R.D. 446 (N.D. Cal. 2012), aff'd sub nom. Fox Test Prep v. Facebook, Inc., 588 F. App'x 733 (9th Cir. 2014). In its opening brief, plaintiff asserts that it satisfies the typicality requirement interface to build the audiences for its ads.” Dkt. 155-3 at 25. Plaintiff says that, during that process, it “was exposed to substantially the same materially false and misleading representations and developed the same false impression of Facebook’s targeting accuracy as every other member of the class.” Id. According to plaintiff, “[w]hen it paid for ads shown to users who did not actually have the characteristics that Facebook claimed,” plaintiff “suffered the same injury as other class members.” Id. at 25-26. The court concludes that plaintiff failed to show that its claims are typical of those of the class it seeks to represent. Three reasons support this conclusion. First, plaintiff’s purportedly typical injury assumes two important facts: (1) all persons using the Ads Manager website interface were exposed to the materially similar statements; and, based on such exposure (2) all such persons developed the same “false impression” of defendant’s targeting abilities. Dkt. 155-3 at 25. Plaintiff, however, fails to substantiate those assumptions. Plaintiff further fails to cite any authority in its opening brief explaining why the existence of some purported misstatements in an interactive website interface warrants an inference of exposure to all misstatements it might contain. These failures are particularly fatal when, as here, defendant proffered evidence showing two countervailing inferences. First, many putative class members were not “exposed” to the interface in the first instance. Dkt. 169-9 ¶ 8 (explaining that many advertisers engage third party ad agencies or consultants to place advertisement for them). Second, even if such members were exposed to the interface, many maintain a degree of preexisting marketing experience that tends to preclude them from reasonably developing a “false impression” about defendant’s advertisement targeting abilities. Dkt. 169-17 ¶ 35 (expert report detailing advertiser survey results showing that “[o]ver half of respondents (55%) reported having run ad campaigns on as many as five or more different advertising platforms and forms of media.”). As this court has previously held, a material difference between the sophistication of a class representative and that of absent class members supports denying certification. Compare In re Facebook, Inc., Second, plaintiff failed to show that its exposure to the interface’s purported misstatements is materially similar to that experienced by other advertisers. As plaintiff acknowledges in its FAC, plaintiff relied on only a handful of Interests and Partner Categories criteria when selecting its target audience. FAC ¶¶ 50-51, 66. Defendant also proffered evidence showing that plaintiff did not select any Behaviors criteria when developing its target audience. Dkt. 169-11 ¶ 7. Based on the above, plaintiff lacks any basis to assert that it viewed at least one major Detailed Targeting sub-category that other putative class members might have been exposed to when forming their “impression” about defendant’s (in)abilities. This shortcoming itself prevents the court from finding that plaintiff qualifies as “typical” of the class it seeks to represent. Third, plaintiff’s principal, Ralph Kidd, faces unique defenses and criticisms that will distract from litigating the merits of any class claims. Critically, when asked at his deposition whether he “would [] run advertisement on Facebook again?”, Kidd testified that “[a]t this time, no.” Dkt. 169-7 at 166. Regardless of whether this testimony deprives plaintiff of Article III standing to pursue forward-looking relief,2 Kidd’s position supports the inference that plaintiff has a greater interest in litigating his requests for monetary relief based on past harms. That inference is further bolstered by Kidd’s acknowledgment that he has repeatedly attempted to profit from his supposed discovery of defendant’s targeting inaccuracies. Dkt. 169-7 at 174. In particular, Kidd solicited hedge funds to participate in a scheme in which the fund would short defendant’s stock and Kidd would then “leak” his theory about defendant’s supposed fraud. Id. at 108-110. Sometime after that idea floundered, Kidd authored a book, “Like Fraud? You Decide.” Id. 171. Once that fell through, Kidd shopped his case to law firms as a possible class action. Id. at 69-70. These attempts also indicate that, rather than using defendant’s services for advertising plaintiff’s business, Kidd is primarily motivated by making money. That motivation serves as another ground to conclude that plaintiff is atypical of the class it seeks to represent. 4. Adequacy Rule 23(a)(4) requires that the party representative fairly and adequately protect the interests of the class. Fed. R. Civ. Pro. 23(a)(4). The Ninth Circuit has established the following two-part test for this requirement: • “[D]o the named plaintiffs and their counsel have any conflicts of interest with other class members?” • “[W]ill the named plaintiffs and their counsel prosecute the action vigorously on behalf of the class?” Staton v. Boeing Co., 327 F.3d 938, 957 (9th Cir. 2003). The court concludes that plaintiff fails these requirements. Two reasons support this conclusion. Both are grounded in Kidd’s deposition testimony. First, at his deposition, Kidd demonstrated disbelief in plaintiff’s ability to represent all advertisers using defendant’s services. In particular, Kidd testified to the following: Q Are you -- through this lawsuit, is [plaintiff] seeking to represent all advertisers on Facebook? [objection] A: I mean, as I said, I would say similarly situated to us. So I -- if that's -- if that's everybody, I'd be surprised. Q: Why would you be surprised? A: I'd be shocked. Q: Why would you be shocked? A: Because if it's what I think it is, then that would really be shocking. Q: But when you say "similarly situated," are you talking about -- you identified some small businesses here. Are you primarily thinking of small businesses? A: Yeah. That's -- that's the bulk of what -- of who uses this platform, this Ads Manager, and it's the bulk of the revenue that Facebook derives, is from small business, SMBs. Right? And it's millions of small businesses like me. Dkt. 169-7 at 78 (emphasis added). entities within the United States who . . . targeted Facebook users . . .” Dkt. 155-3 at 3. Plaintiff’s proposed class does not limit its membership based on business size. Thus, to the extent plaintiff seeks to represent businesses that are larger than so-called “SMBs” (i.e., small to medium sized busineses), Kidd’s testimony itself establishes that plaintiff is inadequate. Compare In re Facebook, Inc., PPC Advert. Litig., 282 F.R.D. at 454 (finding party representative inadequate when it maintains a relatively smaller stake in the litigation than its absent putative class member counterparts). Second, the disparity between Kidd’s testimony and the proposed class suggests that plaintiff’s counsel—not plaintiff—makes the key strategic decisions in prosecuting plaintiff’s claims. That suggestion, in turn, shows that plaintiff is not sufficiently engaged in monitoring this action and, therefore, inadequate for representing its proposed class. * * * The court concludes that plaintiff has failed to satisfy Rule 23(a)’s typicality and adequacy requirements. On that basis alone, the court may deny plaintiff’s motion for class certification. That said, for the reasons noted below, plaintiff also failed to satisfy the requirements for class certification under Rule 23(b)(2) and Rule 23(b)(3). These failures provide an independent basis for denying plaintiff’s motion. 5. Rule 23(b)(3) – Predominance If a plaintiff satisfies Rule 23(a), the plaintiff may maintain a class action under Rule 23(b)(3) if the court finds that “questions of law or fact common to class members predominate over any questions affecting only individual members.” Fed. R. Civ. Pro. 23(b)(3). “The predominance inquiry under Rule 23(b)(3) tests whether proposed classes are sufficiently cohesive to warrant adjudication by representation.” In re Hyundai and Kia Fuel Economy Litig., 926 F.3d 539, 557 (9th Cir. 2019). “It presumes that the existence of common issues of fact or law have been established pursuant to Rule 23(a)(2)” and “focuses on whether the common questions present a significant aspect of the case and they can be resolved for all members of the class in a single adjudication.” than on an individual basis.” Id. The predominance analysis is not a matter of “nose-counting.” Id. Instead, “more important questions apt to drive the resolution of the litigation are given more weight in the predominance analysis over individualized questions which are of considerably less significance to the claims of the class.” Id. Given that, “even if just one common question predominates, the action may be considered proper under Rule 23(b)(3) even though other important matters will have to be tried separately.” Id. To determine if common questions predominate, court first look to the elements underlying the proposed class’s claims. Castro Valley Union 76, Inc. v. Vapor Sys. Techs., Inc., 2012 WL 5199458, at *7 (N.D. Cal. Oct. 22, 2012). The parties agree that each of plaintiff’s claims include some version of each of the following five elements: • Defendant made either a “false” or “likely to deceive” statement or omission. • The purported misstatement was material. • Defendant had knowledge about the misleading nature of the statement. • Defendant intended to induce reliance on the misstatement • Plaintiff suffered harm as a result of relying on the purported misstatement. Compare Dkt. 155-3 at 12 with Dkt. 169-5 at 29 To support its predominance showing, plaintiff identifies four factual questions relevant to the above elements that it contends are capable of classwide resolution. Those questions include the following: • Whether defendant’s “ad-targeting interface” conveyed the “false impression” that it would show ads only to users who matched the criteria advertisers selected. • Whether defendant acted with fraudulent intent in conveying that impression. • Whether class members relied on defendant’s misstatements. • Whether class members were injured as a result of their purported reliance on such misstatements. Dkt. 155-3 at 29. seeks to represent only those advertisers that used the Ads Manager website interface. Dkt. 192 at 33-34. Given that, the court will construe plaintiff’s unspecified reference to defendant’s “ad-targeting interface” to mean only the Ads Manager website interface. With that clarification, the court analyzes the parties’ position on each question below. a. Individualized Issues Predominate the Actionability of the Purported Misstatements Contained in the Ads Manager Website Interface The court concludes that plaintiff’s first question does not qualify as a common issue capable of classwide resolution. First, the court understands that plaintiff contends that the Ads Manager website interface itself serves as the misrepresentation underlying plaintiff’s claims. Plaintiff does not, however, plainly allege that theory of liability in its FAC. At best, plaintiff implies as much in a handful of allegations.3 But plaintiff does not offer those suggestions in a vacuum. Instead, when discussing the Ads Manager website interface in its FAC, plaintiff repeatedly characterizes its supposed misstatements in the plural form and as appearing “in” the interface. Salient examples of those characterizations include the following: • “Mr. Kidd browsed Facebook's targeting interface and saw various representations regarding the display of ads to users matching an advertiser's targeting criteria. Based on those representations, he concluded that Facebook's targeting interface presented a unique opportunity to target ads to IMB's demographic of highly compensated and educated investors.” FAC ¶ 46

3 FAC ¶ 35 (“The targeting interface Facebook provided to Plaintiff and millions of other self-serve advertisers reinforced the false impression that Facebook would display ads only to users who matched advertisers’ targeting criteria.”); id. ¶¶ 78, 80 (similar); id. ¶ 40 (“Facebook understood this interface created the expectations that ads would be displayed with a high degree of accuracy to users who matched the targeting criteria advertisers selected.”); id. ¶ 43 (“Facebook knew that the visuals in its targeting interface led advertisers to expect that Facebook would display ads only to users in the specified target audience, and that it would do so with a high degree of accuracy.”). The “visuals” referred to in paragraph 43 include a so-called “Audience Definition” meter, the purported (emphasis added). • “Relying on Facebook's promise of narrow targeting to a specific audience— including as conveyed by the visuals in the targeting interface—IMB expected that Facebook would display its ad to users within the target audience that it had defined.” Id. ¶ 52 (emphasis added). • “There are questions of law and fact common to the Class that predominate over any questions affecting only individual members, including . . . Whether Facebook defrauded members of the Class by misrepresenting during the Class Period— through public statements by its officers, on its website, and in its targeting interface—that it would accurately display ads . . .” Id. ¶¶ 138 (first “common question” posed in class action allegations) (emphasis added). • “Facebook engaged in a fraudulent practice under the UCL when it misrepresented—both by affirmative statements and material omissions of information it had a duty to disclose—its intention and ability to accurately display ads . . . Id. ¶ 145 (§ 17200 claim) (emphasis added). • “Plaintiffs also saw and reasonably relied upon specific representations by Facebook (in its targeting interface and elsewhere on its website) that it would display Plaintiffs' ads . . .” Id. ¶ 149 (§ 17200 claim) (emphasis added). • “Facebook knew that Plaintiffs and other members of the Class expected—based on Facebook's public statements and the visuals in its targeting interface—that Facebook would display their ads . . .” Id. ¶ 157 (deceit claim) (emphasis added). • “Facebook engaged in common law fraud when it misrepresented (both through affirmative statements and material omissions of information it had a duty to disclose) its intention and ability to accurately display ads . . .” Id. ¶ 164 (common law fraud claim) (emphasis added). With the above context in mind, the court finds that plaintiff’s contention that the Ads Manager website interface (as a whole) serves as a misstatement underlying its motion. The court rejects that attempt and will not consider it. dotStrategy, Co. v. Facebook, Inc., 2021 WL 2550391, at *5 (N.D. Cal. June 22, 2021) (“This order rejects plaintiff's attempt to shift its theory of the merits at the class certification stage.”). But even if the court were to consider it, plaintiff fails to proffer any authority showing that it can simply point to the entire interface and say the whole thing (in the abstract) qualifies as misleading to any advertiser. Instead, to show “uniform conduct likely to mislead [an] entire class,” plaintiff must proffer evidence establishing that all advertisers viewed the same specific misrepresentation. Kaldenbach v. Mut. of Omaha Life Ins. Co., 178 Cal. App. 4th 830, 849-50 (2009), as modified (Oct. 26, 2009) (affirming trial court determination that plaintiff failed to show that insurer acted with uniform misleading conduct under § 17200 when “any given purchaser” could have relied on different “materials, disclosures, representations, and explanations” from sales agents). With that threshold matter decided, the court finds that determining whether the interface includes actionable misstatements poses countless individual issues that predominate their common counterparts. Two sets of evidence support this finding. First, defendant proffered evidence showing that the Interests, Behaviors, and Partner Categories’ sub-categories each comprise numerous criteria. Dkt. 169-13 at 3-4 (declaration acknowledging over 500,000 Interest criteria); Dkt. 170-1 at 40 (exhibit listing various sub-categories of Behavior criteria); Dkt. 170-1 at 76-86 (chart noting user online Behavior criteria selected by example advertisers); Dkt. 169-15 ¶¶ 66, 81-86 (expert report indicating existence of “hundreds” of Partner Categories criteria). Each criterion maintains its own description in an associated hovercard. Dkt. 157-1 at 191-96; Dkt. 170- 1 at 39-40; Dkt. 170-2 at 1-8. To be sure, plaintiff presented evidence suggesting that all hovercard descriptions for the Interests sub-category maintain similar descriptions. Dkt. 179-11 at 3; Compare Dkt. 170-1 at 39 with Dkt. 157-1 at 165. Other evidence, however, shows that various of the Behaviors and Partner Categories criteria descriptions contain conditional phrases 12 (user Demographics), “likely,” Dkt. 170-2 at 4 (user Demographics), “suggests,” Dkt. 157-1 at 192; Dkt. 170-2 at 6 (user Behavior), “primarily,” Dkt. 170-2 at 8 (user Behavior), or that the description is based on an “estimate,” Dkt. 157-1 at 192 (user Demographics). When taken together, the evidence proffered by defendant shows that the court would, at minimum, need to review the hovercards for all Behaviors and Partner Categories criteria to determine whether a given hovercard includes a conditional phrase about a user attribute. In that event, the court would then need to determine whether such phrase renders the statement non-actionable opinion. That two-step process would require the court to resolve countless individualized questions. The fact that an advertiser could combine any particular Behavior or Partner Categories criterion with over 500,000 other Interests criteria (regardless of whether all criteria in that sub-category maintain similar descriptions) would further complicate that inquiry. Second, defendant proffered evidence showing that its Terms and Advertising Guidelines (the “Terms and Guidelines”) expressly state that “[w]hen serving your ad, we do our best to deliver the ads to the audience you specify, although we cannot guarantee in every instance that your ad will reach its intended target.” Dkt. 170-3 at 437-54 (emphasis added). Defendant presented evidence showing that all self-serve advertisers were required to accept the Terms and Guidelines. Dkt. 169-9 ¶ 18. Defendant also presented evidence showing that it included a link to the Terms and Guidelines prior to an advertiser’s decision to “publish” its campaign. Dkt. 170-1 at 44. The latter two pieces of evidence suggest that a material number of putative class members may have reviewed the Terms and Guidelines prior to publishing their campaign. Such a review affects whether any purported misstatement contained in the interface qualifies as actionable. Thus, to decide that question, the court would need to determine whether an advertiser reviewed this disclaimer prior to publishing its campaign. That determination would generate up to 2.5 million individualized inquiries. For the above reasons, the court concludes that plaintiff’s first question concerning not qualify as a common issue capable of classwide resolution. b. Individualized Issues Predominate Determining Defendant’s Fraudulent Intent Under California law, whether a person acted with “fraudulent intent” depends on whether that person intended to lead a second person into relying on the first person’s misleading statement. Lazar v. Superior Ct., 12 Cal. 4th 631, 638 (1996) (“The elements of fraud . . . are . . . (c) intent to defraud, i.e., to induce reliance . . .”). To show such reliance, a plaintiff must show a change in his or her position because of the purported deceit. Cal. Civ. Code § 1709 (“One who willfully deceives another with intent to induce him to alter his position to his injury or risk, is liable for any damage which he thereby suffers.”) (emphasis added). The court rejects plaintiff’s contention that it can show defendant’s fraudulent intent on a classwide basis. Three reasons support this conclusion. First, plaintiff failed to show that, prior to purchasing defendant’s services, advertisers operate on a blank slate in terms of their understanding about defendant’s advertisement targeting abilities. Second, defendant proffered evidence showing that, when creating their campaigns, many advertisers retain third party advertisement agencies or consultants, Dkt. 169-9 ¶ 8, or communicate with Facebook Help Center personnel, id. ¶ 12. Defendant also proffered evidence showing that many advertisers maintain different agreements with defendant. Id. ¶¶ 18-21. The above variabilities preclude plaintiff from showing that all advertisers shared a uniform understanding about defendant’s targeting abilities prior to their purchase of defendant’s services. Because plaintiff must establish an advertiser’s initial position in order to show a change in that position, plaintiff cannot show fraudulent intent on a classwide basis. Third, the court understands plaintiff’s assertion in its reply that the court may 179-3 at 18. That assertion does not salvage plaintiff’s argument. As decided in Section II.B.5.a., the court cannot determine the actionability of the targeting interface’s purported misstatements on a classwide basis. But even if the court could, plaintiff does not proffer any authority showing that defendant may not rebut such an inference by presenting evidence that negates its supposed fraudulent intent. As noted above, such evidence includes, for example, that defendant understood that numerous advertisers retain third party advertisement agencies and consultants when creating their campaigns. The court finds it plausible that those third parties would advise advertisers about the limits on defendant’s targeting abilities. Thus, even if the court were to infer fraudulent intent based on defendant’s unspecified “misrepresentations,” the court would still need to consider other evidence similarly showing defendant’s mental state about what any particular advertiser understood about defendant’s targeting abilities. That need would raise countless individual questions. For the above reasons, the court concludes that plaintiff’s second question concerning defendant’s purportedly fraudulent intent does not qualify as a common issue capable of classwide resolution. c. Individualized Issues Predominate Whether Class Members Relied on the Ads Manager Website Interface’s Purported Misstatements In its opening brief, plaintiff asserts that the court owes its proposed class a “presumption of reliance” because plaintiff limited its class to those advertisers who were “exposed” to the interface’s purportedly “materially misleading” misstatements. Dkt. 155- 3 at 31. To support that position, plaintiff cites the Ninth Circuit’s decision Mazza v. Am. Honda Motor Co., 666 F.3d 581 (9th Cir. 2012). Id. In its opposition, defendant asserts that plaintiff cannot offer common proof of such exposure to or materiality of the misstatements at issue. Dkt. 169-5 at 31-37. The court addresses each contention below. i. Exposure to the Purported Misstatements In its opening brief, plaintiff asserts that its proposed class definition “guarantees” that “every single class member saw and interacted with the misrepresentations and omissions” that are “at the heart of this case.” Dkt. 155-3 at 31. According to plaintiff, those misrepresentations include the Ads Manager website interface’s “menu” of targeting criteria, its hovercards, and its reach estimate per targeting category. Id. Plaintiff says there is “no dispute” that the purported misstatements were communicated to all members because those statements were made as part of an online purchasing process. Id. at 31. Plaintiff then asserts that any differences in the interface’s purported misstatements “make no difference to the predominance analysis.” Id. at 32. Instead, plaintiff explains, the critical question is whether there is a “classwide practice of non- disclosure.” Id. Plaintiff cites Torres v. Mercer Canyons Inc., 835 F.3d 1125, 1137-38 (9th Cir. 2016) and In re First Alliance Mortgage Co., 471 F.3d 977, 991 (9th Cir. 2006) for the proposition that such a practice qualifies as a common question that may predominate over the interactions between defendant and a particular advertiser. Id. The court finds that plaintiff failed to establish that it can prove that its proposed class members were exposed to the challenged statements on a classwide basis. First, as detailed in Section II.B.5.a., defendant presented evidence showing that the Interests, Behavior, and Partner Categories subcategories comprise criteria that detail different hovercard descriptions. Some of those descriptions use qualifying phrases (e.g., “likely” or “suggesting”) that would require the court to make an individualized inquiry into the actionability of a description using such phrase. Because some descriptions might not be actionable, plaintiff must identify which advertisers viewed which description to show exposure. Plaintiff failed to show how it might do so on a classwide basis. Second, plaintiff overstates the import of Torres and In re First Alliance Mortgage Co. to the case at hand. As an initial matter, plaintiff is correct that the court in Torres Rule 23(b)(3)’s predominance inquiry. However, that same court went on to observe that a “disparate series of affirmative statements” might preclude predominance. More specifically, the court in Torres explained: [Defendant] counters that whether the omission was misleading, and therefore injurious, will “turn on the minutia[e] [sic] of individual interactions and circumstances” between [defendant] and class members. However, [defendant] has not shown that the class as a whole was exposed to “disparate information from various representatives of the defendant” . . . creating materially different impressions about the availability of H-2A work. Rather, the conduct at issue is reasonably uniform as the crux of Plaintiffs' legal challenge involves a common failure to disclose information, and not merely a disparate series of affirmative statements. 835 F.3d at 1137- 38. Here, despite its characterization in this motion, plaintiff does not base its theory of liability solely on omissions. To the contrary, as memorialized in its FAC, plaintiff alleges that defendant made numerous affirmative misstatements in its targeting interface. FAC ¶¶ 26-34 (discussing website representations); id. ¶¶ 40-42 (listing questions detailed at certain stages in the targeting interface); id. ¶ 46 (explaining that plaintiff “saw various representations regarding the display of ads to users matching an advertiser’s targeting criteria” and decided to use Facebook “based on” those representations); id. ¶¶ 78, 127 (alleging various “assurances” by defendant to advertisers about its targeting abilities); id. ¶¶ 145, 155, 164 (based all three of its claims on defendant’s “affirmative statements and material omissions”) (emphasis added). In light of these differences, the court finds that the Ninth Circuit’s decision in Torres does not control. That leaves In re First Alliance Mortgage Co. In that case, the Ninth Circuit acknowledged that “[c]lass treatment has been permitted in fraud cases where, as in this case, a standardized sales pitch is employed.“ 471 F.3d at 991. It cautioned that courts should avoid a “talismanic rule” that class certification for fraud claims requires “identical” misrepresentations. Id. Here, plaintiff fails to explain how the purported misstatements in the Ads Manager may choose among thousands of criteria to customize its own audience. Defendant proffered evidence showing that each criterion maintains its own hovercard description. When taken together, those features show that an interface’s statements do not qualify as a standardized sales pitch. Given the above, the court finds that plaintiff failed to show that it can prove exposure to the Ads Manager website interface’s purported misstatements on a classwide basis. ii. Materiality of the Purported Misstatements Under California law, a misstatement qualifies as “material” if “a reasonable [person] would attach importance to its existence or nonexistence in determining his [or her] choice of action in the transaction in question.” Engalla v. Permanente Med. Grp., Inc., 15 Cal. 4th 951, 977 (1997), as modified (July 30, 1997). Given that, California courts have explained that “materiality is generally a question of fact” unless the “fact misrepresented is so obviously unimportant that the jury could not reasonably find that a reasonable [person] would have been influenced by it.” Id. In its opening brief, plaintiff again focuses on the materiality of the Ads Manager website interface (as a whole), as opposed to the particular misstatements in it. Dkt. 155- 3 at 33-34. Plaintiff asserts that the interface was materially misleading because it suggested that defendant could “distinguish among its users with an extraordinary degree of precision” when, in reality, “more than half the users who would ultimately see their ad would not actually fit the categories they sought to reach.” Id. at 33. In its opposition, defendant asserts that a statement’s materiality depends on the specific advertiser. Dkt. 169-5 at 34-36. To substantiate that assertion, defendant relies in part on an expert report from a University of California Los Angeles professor of marketing, Dominque Hanssens (“Hanssens”). Dkt. 169-17. The court rejects plaintiff’s contention that it can show materiality on a classwide basis. Again, as decided in Section II.B.5.a., plaintiff may not now shift its theory of itself. But even if the court considered that theory, plaintiff still fails to show that it could establish that all advertisers generally interacting with that interface would also “attach importance” to the fact that defendant could target users with advertisements at less than 50 percent accuracy. As defendant aptly points out, plaintiff fails to proffer any evidence from other advertisers on the materiality issue in its opening brief. Instead, plaintiff principally relies on the Bosworth report’s recognition that “[w]hen advertisers select” targeting criteria, they “expect [defendant] to have high confidence that we will only show ads to [their] intended audience.” Dkt. 155-3 at 34 (emphasis in the original). The Bosworth report certainly shows that defendant’s key personnel were aware about limits on defendant’s targeting abilities. Dkt. 155-8 at 269. That report also indicates that defendant did not “feel” that it was “meeting expectations” of advertisers concerning accuracy. Id. However, these admissions (and others contained in similar internal documents cited by plaintiff) speak to defendant’s understanding of how important advertisers find accuracy. That singular understanding does not establish that plaintiff can show—by common proof—that all 2.5 million advertisers in its proposed class would attach materially similar significance to defendant’s actual targeting abilities. Defendant, on the other hand, proffered three sorts of evidence undermining the possibility of such showing. First, in his report, Hanssens surveyed advertisers on the following question: Now please think again about the most recent paid Facebook advertising campaign you were involved in. What sources did you primarily rely on in deciding to use Facebook for this campaign? Select all that apply. Dkt. 169-17 at 192. According to Hanssens, the survey yielded the following results by respondents: • 44.8 percent answered “prior knowledge about the effectiveness of advertising on Facebook from previous Facebook advertising campaigns.” • 38.8 percent answered “researched online outside of Facebook’s website.” • 31.5 percent answered “recommended by an advertising agent/third-party consultant.” • 26.1 percent answered “statement made by Facebook on its website.” • 21.2 percent answered “statements made by Facebook representatives.” Id. These answers show significant variation in the sources of information that an advertiser relies on when determining whether to use Facebook for an advertising campaign. That variance, in turn, raises individualized questions about the materiality of the interface (or its particular misstatements) to any given advertiser’s purchase decision. Second, defendant proffered evidence showing that advertisers maintain different objectives. Those objectives include increasing one of the following metrics: (1) brand awareness; (2) website traffic; or (3) in-store traffic. Dkt. 170-1 at 29-33. These differing objectives might render the interface’s purported misstatements more or less material to any particular advertiser. If, for example, an advertiser’s goal is to increase its brand awareness (as opposed to sales), then that advertiser’s objective in increasing awareness is still advanced when defendant serves an advertisement to a user outside the advertiser’s target audience. Third, defendant proffered the challenged declarations. Dkt. 170-5 at 169-205. Those declarations show that advertisers might have different expectations about defendant’s targeting abilities. Compare Id. at 177 (“I don’t expect the targeting we use to be 100% accurate because that would be impossible”) with Id. at 192 (acknowledging that “the audience selected and the people who ultimately see the ads may not always be a perfect match”). While anecdotal, these declarations illustrate that advertisers do not make purchase decisions based solely on defendants’ statements. Instead, as common sense suggests, advertisers (like anyone in business) enter such transactions with prior knowledge, experience, and expectations. The differences in such preexisting beliefs raise individualized inquiries on whether defendant’s failure to disclose the accuracy rate of its targeting were material to all advertisers’ purchase decisions. and materiality on a classwide basis. Thus, plaintiff is not entitled to a presumption that its proposed class relied on the Ads Managers website interface’s purported misstatements. Accordingly, the court concludes that plaintiff’s third question concerning class member reliance does not qualify as a common issue. d. Plaintiff Failed to Proffer a Damages Model Consistent with Its Theory of Liability and Capable of Measurement on a Classwide Basis Rule 23(b) requires that a putative class representative show that “damages are capable of measurement on a classwide basis.” Comcast Corp. v. Behrend, 569 U.S. 27, 34 (2013). In Comcast, the Supreme Court ruled that courts “must conduct a ‘rigorous analysis’ to ensure at the class-certification stage that 'any model supporting a plaintiff's damages case [is] consistent with its liability case.’” 569 U.S. at 35. In effect, that analysis requires “that the model ‘measure[s] only those damages attributable to that theory’ of liability.” Rikos v. Procter & Gamble Co., 799 F.3d 497, 523 (6th Cir. 2015). The Ninth Circuit has clarified, however, that Comcast “did not alter our holding that individualized damages issues do not alone defeat certification.” Nguyen v. Nissan N. Am., Inc., 932 F.3d 811, 817 (9th Cir. 2019) (citing Pulaski & Middleman, LLC v. Google, Inc., 802 F.3d 979, 988 (9th Cir. 2015)). “The mere fact that there might be differences in damage calculations is not sufficient to defeat class certification.” In re Hyundai and Kia Fuel Economy Litig., 926 F.3d at 560-61. In its opening brief, plaintiff identifies two damages models that it says are consistent with its theory of liability and capable of proof on a classwide basis. Dkt. 155-3 at 34-37. Plaintiff’s expert, Gregory Pinsonneault (“Pinsonneault”), describes those methods as follows: (1) the “proportional or R/N methodology” (the “R/N method”) and (2) the “revenue share methodology” (the “revenue share method”). Dkt. 155-5 (Pinsonneault December 23, 2020 Declaration) ¶ 7. Pinsonneault explains that the R/N method focuses on “allocat[ing] the revenue Id. ¶ 92.4 Pinsonneault claims that he can “calculate the diminution in value based on” the R/N method by the following two steps: • Allocate revenue among all criteria used for a particular advertisement. Id. ¶ 112. • Calculate the diminished value for inaccurate criteria. Id. ¶¶ 112, 119-20 (diminished value equals the value of a particular criterion reduced by the likelihood of its inaccuracy). According to Pinsonneault, both steps are supported by defendant’s refund policies. Id. ¶ 143. Pinsonneault explains that the first step is consistent with defendant’s “practice of identifying what proportion of the ‘ad spend’ was affected by the targeting problem.” Id. Pinsonneault adds that the second step is consistent with defendant’s “policy of ‘considering refunds if more than 5% of spend was for delivery outside of targeted areas, regardless of the cause of the issue.” Id. Turning to the revenue share method, Pinsonneault explains that that method measures “the amount of revenue associated with a partner category targeting segment” and “the revenue share was the portion of revenue a data partner received as compensation for providing the segment.” Id. ¶ 146. Pinsonneault notes that this method reflects defendant’s approach “for compensating third-party vendors for the data they provided” to support the (now defunct) Partner Categories sub-category. Id. ¶ 145. Pinnsonneault claims that he can measure the diminished value for a Partners Categories criterion by applying the following two steps: • Measure the value of each Partner Categories criterion used in an advertisement by taking its revenue share. Id. ¶ 158. • Calculate the diminution in value for each inaccurate criterion “based on the observation that the advertiser did not receive any value” from it. Id. ¶ 161. Pinsonneault suggests that he may use the revenue share method to calculate only the diminution in value resulting from use of the Partner Categories criteria (as opposed to its Interests and Behavior counterparts). Id. ¶ 166. The court concludes that plaintiff failed to identify a damages model that is both consistent with its theory of liability and capable of measurement on a classwide basis. Three reasons support this conclusion. First, plaintiff does not disagree that it must show that defendant’s conduct caused each class member’s purported injury. Dkt. 169-5 at 38. Instead, plaintiff asserts only that “[t]he presumption of reliance takes care of causation.” Dkt. 179-3 at 23. As decided in Section II.B.5.c., plaintiff failed to show that the proposed class is entitled to such a presumption. Thus, contrary to its bare assertion, plaintiff failed to explain how defendant’s conduct caused each proposed class member’s purported injury. Second, contrary to plaintiff’s contention, there is a “practical basis” to consider any given advertiser’s “subjective value,” “expectations,” or “motivations” when considering whether defendant’s purported misstatement injured such advertiser. Those preexisting beliefs inform how much an advertiser is willing to pay for defendant’s services. Logically speaking, in the event an advertiser values defendant’s inaccurate targeting at some rate greater than the amount the advertiser paid for such services, then the advertiser might still purchase the service. Why? The advertiser would, in that event, nonetheless profit from the subject services. Of course, that possibility then raises individualized issues concerning how each advertiser values those services (and why). To be sure, plaintiff is correct that the Ninth Circuit in In re Hyundai and Kia Fuel Econ. Litig., (9th Cir. 2019) has acknowledged that: Nor is it clear why the damages here would need to be calculated based on each consumer's willingness to pay for higher fuel efficiency. Fraud damages do not normally correlate with the degree of reliance. . . . If a consumer were to establish the threshold level of reliance, the automaker would be liable for the consumer's entire loss from higher-than- expected fuel costs—an amount that can easily be calculated on an individual basis, as it was in the Reimbursement Program. 926 F.3d at 561 (emphasis added). In this case, though, it is “clear” why the damages would need to be calculated on plaintiff adopts that economic logic to support its claims. FAC ¶ 161 (“Had Plaintiffs and other members of the Class known the truth about the material inaccuracy of the data Facebook used for targeting advertisements, they would not have agreed to pay anything at all for those ads, or at a minimum would have paid Facebook substantially less for those ads”) (emphasis added); Id. ¶¶ 169, 151 (substantially similar allegations). Moreover, in his declaration, Pinsonneault acknowledges the relationship between accuracy and an advertiser’s willingness to pay for defendant’s services. When explaining the importance of accuracy to advertisers, Pinsonneault cites a February 2016 email by Jian Yuan (“Yuan”). Dkt. 155-5 ¶ 63. In that email, Yuan observed that: Interest accuracy has a direct correlation of the adoption/spend on interest. There will be no interest targeting if we have 0 accuracy, which currently accounts for 1/3 of our revenue. The more accurate we are, the more advertiser would choose to spend on interest, even though the effect may be long term and not measurable in A/B test. Id. (emphasis added). Given its own pleadings and evidence, plaintiff cannot now credibly take the position that considering a particular advertiser’s subjective willingness to pay for defendant’s services (however inaccurate) lacks any “legal or practical basis” when determining damages. Dkt. 179-3 at 23. Third, defendant is correct that plaintiff cannot rely on a damages model that calculates damages based on “average accuracy rates.” As defendant points out, that approach would arbitrarily under- or over-award certain class members without regard to the accuracy of the specific criteria they chose. The court understands plaintiff’s citation to Pulaski for proposition that California law permits fraud damages based on an “approximation.” Pulaski & Middleman, 802 F.3d at 989 (“In calculating damages, here restitution, California law requires only that some reasonable basis of computation of damages be used, and the damages may be computed even if the result reached is an approximation.”). That said, plaintiff’s theory of liability is that it and putative class members overpaid for defendant’s services because of Given that significance, plaintiff may not, consistent with Comcast, dismiss its damages model’s endemic under- and over-inclusiveness as a mere rounding error “approximation.” In short, the court finds that that plaintiff failed to identify a damages model that is both consistent with Comcast’s requirements and capable of classwide measurement. Accordingly, plaintiff’s fourth question concerning class member injury does not qualify as a common issue.5 * * * The court finds that plaintiff failed to show that questions central to the validity of its claim and capable of classwide resolution predominate over their individualized counterparts. As a result, the court concludes that plaintiff failed to satisfy the Rule 23(b)(3)’s predominance requirement. On that alternative basis, the court denies plaintiff’s request that the court certify its proposed class pursuant to Rule 23(b)(3). 6. Rule 23(b)(2) – Injunctive Relief If a plaintiff satisfies Rule 23(a), he or she may maintain a class action under Rule 23(b)(2) if “the party opposing the class has acted or refused to act on grounds that apply generally to the class, so that final injunctive relief or corresponding declaratory relief is appropriate respecting the class as a whole.” Fed. R. Civ. Pro. 23(b)(2). In Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338 (2011), the Supreme Court explained that “the key to the (b)(2) class is the indivisible nature of the injunctive or declaratory remedy warranted—the notion that the conduct is such that it can be enjoined or declared unlawful only as to all of the class members or as to none of them.” Dukes, 564 U.S. at 360. The Ninth Circuit has explained that this requirement is “unquestionably satisfied when members of a putative class seek uniform injunctive or declaratory relief

5 The court agrees with plaintiff that defendant effectively concedes that the court may determine punitive damages on a classwide basis. Dkt. 169-5 at 41-42. That said, plaintiff also concedes that, to open the punitive damages issue, plaintiff must first from policies or practices that are generally applicable to the class as a whole.” Parsons v. Ryan, 754 F.3d 657, 688 (9th Cir. 2014). In Parsons, Judge Reinhardt emphasized that: [The Rule 23(b)(2)] inquiry does not require an examination of the viability or bases of the class members' claims for relief, does not require that the issues common to the class satisfy a Rule 23(b)(3)-like predominance test, and does not require a finding that all members of the class have suffered identical injuries. Id. (footnote and citations omitted). Instead, Judge Reinhardt explains: as the text of the rule makes clear, this inquiry asks only whether “the party opposing the class has acted or refused to act on grounds that apply generally to the class.” Id. The court denies plaintiff’s request to certify its proposed class under Rule 23(b)(2). As an initial matter, the court rejects defendant’s contention that plaintiff lacks Article III standing to pursue forward-looking equitable relief. In Davidson v. Kimberly- Clark Corp., 889 F.3d 956 (9th Cir. 2018), the Ninth Circuit ruled that a “previously deceived” consumer may satisfy Article III’s “actual and imminent” harm requirement by the consumer’s “plausible allegations that she will be unable to rely on the product's advertising or labeling in the future, and so will not purchase the product although she would like to.” Id. at 969-70. The holding of that case controls here. Following his deposition, Kidd submitted a declaration stating that “[i]f [pliantiff] could rely on Facebook’s representation concerning its targeting accuracy,” then plaintiff “would resume advertising on Facebook.” Dkt. 156-1 ¶ 5.6 In light of that follow-up statement, the court concludes that plaintiff may continue to pursue forward-looking injunctive relief despite Kidd’s deposition testimony about his intent to advertise on Facebook in the future. 6 The court rejects defendant’s contention that the sham affidavit doctrine precludes the court from considering Kidd’s declaration. Courts narrowly construe that doctrine and require a clear inconsistency between a declarant’s prior and subsequent statement for it to apply. SM 10000 Prop., LLC v. Allianz Glob. Risks US Ins. Co., 2021 WL 1299071, at That short-lived victory aside, the court finds that plaintiff failed to show how the court could issue an order awarding its requested equitable relief on a classwide basis. As decided in Section II.B.5.c.i., plaintiff fails to show that all proposed class members were exposed to the same purported misstatements when using the Ads Manager website interface. Again, that interface details different descriptions per Interest, Behavior, and Partner Category criteria. Some of those descriptions include qualifying phrases that might render their statements non-actionable. Thus, the court would need to review each description to determine whether it qualifies as actionable. Such description-specific review precludes this court from enjoining any misstatements on that interface on a uniform basis. For the above reasons, the court finds that plaintiff failed to show that certification under Rule 23(b)(2) is improper. The court denies plaintiff’s request for Rule 23(b)(2) certification on that alternative basis. III. Motions to Seal “There is a general principle in favor of public access to federal court records.” Nixon v. Warner Commc'ns, Inc., 435 U.S. 589, 597 (1978). “The proponent of sealing bears the burden with respect to sealing. A failure to meet that burden means that the default posture of public access prevails.” Kamakana v. City & City of Honolulu, 447 F.3d 1172, 1182 (9th Cir. 2006). The Ninth Circuit has recognized that two different standards may apply when a request to seal a document is made in connection with a motion—namely the “compelling reasons” standard or the “good cause” standard. Ctr. For Auto Safety v. Chrysler Grp., LLC, 809 F.3d 1092, 1096-97 (9th Cir. 2016). The compelling reasons standard applies to any sealing request made in connection with a motion that is “more than tangentially related to the merits of a case.” Id. at 1101. Courts in this district have ruled that the compelling reasons standard controls sealing requests made in connection with a motion for class certification. Yan Mei Zheng Kellogg Sales Co., 2018 WL 7814785, at *2 (N.D. Cal. Sept. 5, 2018). Under the compelling reasons standard, a court may seal a record only if it finds a “compelling reason” to support such treatment and articulates “the factual basis for its ruling, without relying on hypothesis or conjecture.” Ctr. for Auto Safety, 809 F.3d at 1096-97. If it has made such finding, the court “must then conscientiously balance the competing interests of the public and the party who seeks to keep certain judicial records secret.” Id. at 1097. Factors relevant to that balancing test include the public interest “in understanding the judicial process,” Pintos v. Pac. Creditors Ass'n, 605 F.3d 665, 679 n.6 (9th Cir. 2010), as well as the volume of material sought to be sealed, Algarin v. Maybelline, LLC, 2014 WL 690410, at *3 (S.D. Cal. Feb. 21, 2014). The Ninth Circuit has acknowledged that “[w]hat constitutes a ‘compelling reason’ is best left to the sound discretion of the trial court.” Ctr. for Auto Safety, 809 F.3d at 1097. “Examples include when a court record might be used to gratify private spite or promote public scandal, to circulate ‘libelous’ statements, or ‘as sources of business information that might harm a litigant's competitive standing.’” Id. In its November 6, 2020 order, the court held that qualifying business information might also include the following: • Sufficiently specific descriptions of a company’s technical processes and function. Dkt. 147 at 22. • Internal studies suggesting that they might qualify as trade secrets. Id. 21. • Financial detail if the subject numbers provide some glimpse into the company’s mental impressions about how it dedicates its organizational efforts. Id. at 20. Defendant filed three motions for the court to permanently seal various filings made in connection with the motion for class certification and motion to strike. Dkt. 158; Dkt. 169; Dkt. 184. Plaintiff opposes those motions. Dkt. 160; Dkt. 175; Dkt. 191. The court addresses each motion in turn. A. Motion to Seal Plaintiff’s Motion for Class Certification Materials contained in plaintiff’s motion for class certification filings. Dkt. 158 at 2. The subject information includes parts of plaintiff’s opening brief, expert reports, and supporting exhibits. Dkt. 158-1 at 8-20. According to defendant, the subject portions concern some combination of the following five categories of information: 1) Defendant’s internal testing and assessments of its ads targeting products. 2) Defendant’s revenue streams. 3) Defendant’s refunds and the reasons for such refunds. 4) The identities of defendant’s business partners. 5) The technical processes and functions of defendant’s systems. Id. at 6. In its motion, defendant asserts that the information sought for sealing “reflect or quote” its “internal proprietary, commercially sensitive, and confidential information.” Id. Defendant claims that it “would be at a competitive disadvantage if this information were to be publicly released.” Id. To support its claims of potential harm, defendant proffers the declarations of two employees: (1) Associate General Counsel Holly Tambling (“Tambling”), Dkt. 158-1; and (2) Data Science Manager David Amsallem (“Amsallem”), Dkt. 158-2. The court will detail these declarations as necessary in its analysis below. 1. Internal Testing and Assessments of Ads Targeting Products In her declaration, Tambling explains that the information in this category includes the following: [1] Confidential information about specific studies that Facebook conducted on various ads targeting features, such as interest targeting and targeting using third-party data, and internal discussions about those studies. . . . [2] The specific methods Facebook employed to run its studies, specific formulas Facebook developed to test its features and products, and/or the results of these studies. Dkt. 158-1 ¶ 8. Tambling states that defendant “believes” that disclosure of this information “would allow Facebook’s competitors to gain an unfair advantage, thereby causing Facebook category relies on “proprietary methods” that defendant “has developed and . . . does not share with competitors.” Dkt. 158-2 ¶ 4. He also generally asserts that “[t]he disclosure of the information above would subject Facebook to competitive harm” because disclosure would reveal to competitors: [1] how Facebook’s products and system works . . . [2] what Facebook is doing to evaluate, analyze, and improve [its products and system] . . . [3] allow [competitors] to anticipate Facebook’s business decisions and . . . react in ways they could not have otherwise. Id. ¶ 8. The court denies defendant’s motion to seal the information in this category for three reasons. First, Tambling does not provide any factual basis to support her statement that defendant will suffer competitive harm if this information were unsealed. Tambling’s citation to Amsallem’s declaration, Dkt. 158-1 ¶ 8, does not make up for that shortcoming. Critically, in his declaration, Amsallem only generally states that if disclosed, internal testing and assessment information might result in competitive harm. Dkt. 158-2 ¶ 8. He does not connect that statement to any specific information at issue. Second, in at least one circumstance, defendant seeks to seal portions of documents that appear to be nothing more than embarrassing discussions among defendant’s personnel about its targeting inaccuracy. Dkt. 155-8 at 28 (“Last but not least, one of the questions asked in our most recent research is ‘how important is interest accuracy to you,’ and almost all advertisers choose ‘very important.’ They may not be able to tell how good we are doing due to the lack of transparency, but it just doesn’t feel right we don’t invest in something that advertisers think as one of the most important. . . .”) (emphasis added). Thus, defendant failed to narrowly tailor its sealing request to protect only the supposed testing and assessment information in this category. Third, defendant understates the significance of the public’s interest in understanding the merits of this action. The Facebook platform has changed how assessments of its advertisement targeting abilities show the efficacy of such abilities and, perhaps more importantly, what defendant understood about such efficacy. Given the public remarks made by defendant about those abilities, FAC ¶¶ 19-34, the court finds that the public has a significant interest in accessing this information. 2. Revenue Streams In her declaration, Tambling explains that the information in this category concerns the “size and relative importance of certain non-public revenue streams.” Dkt. 158-1 ¶ 9. According to Tambling, defendant would suffer a “competitive disadvantage” if the information in this category were disclosed because such disclosure “would allow Facebook’s competitors to gain unfair insight into Facebook’s business decision-making and to adjust their strategies accordingly.” Id. ¶ 9. The court finds Tambling’s explanation persuasive. The court further finds that the public does not need to see the subject financial details to understand this action. Accordingly, the court grants defendant’s motion to seal the information in this category. Compare Dkt. 147 at 20 (granting motion to seal similar revenue-related information). 3. Refunds and Reasons for Refunds In her declaration, Tambling explains that the information in this category “identif[ies] and discuss[es] the amounts of refunds to certain clients and/or the reasons for them.” Dkt. 158-1 ¶ 10. Tambling says that disclosing “details about the refunds (and the reasons for them)” would “reveal sensitive pricing information of the advertiser.” Id. She adds that the subject information “could” reveal “proprietary information about Facebook’s ad auction process.” Id. ¶ 12. She then adds that disclosing defendant’s ad auction process “would give competitors unfair insight into how Facebook conducts its business,” thereby “subjecting it to competitive harm.” Id. The court denies defendant’s motion to seal the information in this category for two reasons. First, courts deny requests to seal refund information absent evidence of specific harm. McArdle v. AT&T Mobility LLC, 2018 WL 6803743, at *6 (N.D. Cal. Aug. proffer any declaration from an advertiser whose legitimate privacy interest would be affected by disclosure of the subject information. Second, as detailed in Section II.B.5.d., defendant’s refund policy serves as a premise underlying plaintiff’s R/N method for showing damages on a classwide basis. Even though the court rejects plaintiff’s position on that issue, the subject policy nonetheless maintains a notable role in this order’s Rule 23(b)(3) predominance analysis. Given that, the court finds that the public’s interest in understanding defendant’s refund policy outweighs defendant’s (or any given advertiser’s) private interest in hiding that policy from the public record. 4. Business Partner Identities In her declaration, Tambling explains that the information in this category includes the identifies of “advertisers, contractors, and computer software companies.” Dkt. 155-8 ¶ 13. Tambling says that disclosing the identities of these third parties would “undermine Facebook’s business relationship and would put it at a disadvantage in the marketplace,” thereby “causing it competitive harm.” Id. ¶ 13. The court denies defendant’s motion to seal this category of information. Critically, Tambling fails to explain how disclosure of a business partner’s identity would harm defendant’s placement in the marketplace. At core, defendant asks this court to seal the subject information simply because defendant does not want their relationships publicized. The court rejects such a blanket position. 5. Technical Processes and Functions of Defendant’s Systems In her declaration, Tambling explains that the information in this category includes the following: [1] the technical functionalities of Facebook’s ad targeting systems and . . . [2] the data and databases that inform those functionalities, including the names of these databases and descriptions of their content. Dkt. 158-1 ¶ 14. competitors to gain an unfair advantage by giving [them] insight into Facebook’s internal processes,” thus “causing Facebook competitive harm.” Id. In his declaration, Amsallem states that the information in this category is “particularly sensitive because Facebook uses this data to improve its advertising offerings” and its disclosure “would give competitors unfair insight into how Facebook is improving its products.” Dkt. 158-2 ¶ 7. The court grants defendant’s motion to seal the information in this category. Unlike the business information at issue in the above categories, the information in this category appears technical in nature. Given that nature, the court will credit Amsallem’s statement of competitive harm more weight. Further, the court also finds that, under the circumstances at hand, the public does not need to access the details of defendant’s technical processes to understand this litigation. Dkt. 147 at 22. B. Motion to Seal Defendant’s Opposition Materials Defendant’s second motion to seal closely tracks its first motion to seal. In its second motion, defendant seeks to seal 37 portions of information contained in its opposition to plaintiff’s motion for class certification. Dkt. 169 at 2. The subject information includes parts of its opposition brief, expert reports, and supporting exhibits. Dkt. 169-2 at 8-14. The subject portions concern five categories of information that are substantially similar to that at issue in defendant’s first motion to seal. Compare Dkt. 158-1 at 6 with Dkt. 169-2 at 4. To support this motion, defendant proffers another declaration from Tambling (Dkt. 169-1) and Amsallem (Dkt. 169-2). In her second declaration, Tambling modifies her description of the information in certain categories. Dkt. 169-2 ¶ 8. However, neither she nor Amsallem proffer any new, material claim of harm to justify sealing other than that already noted above. Compare Dkt. 158-1 ¶¶ 8-15 with Dkt. 169-2 ¶¶ 8-15 (Tambling declarations); Compare Dkt. 158-2 with Dkt. 169-1 (Amasallem declarations). Given these similarities, the court grants defendant’s motion to seal any information in categories two and five. The court otherwise denies defendant’s request. C. Motion to Seal Plaintiff’s Reply Materials Defendant’s third motion to seal closely tracks its first and second motions to seal. In its third motion, defendant seeks to seal 51 portions of information contained in plaintiff’s reply briefs filed in support of its motion for class certification and motion to strike. Dkt. 184 at 2. The subject information includes parts of those briefs, expert reports, and supporting exhibits. Dkt. 184-1 at 8-16. The subject portions concern the same five categories of information at issue in defendant’s second motion to seal. Compare Dkt. 169-2 at 4 with Dkt. 184-1 at 3. To support this motion, defendant proffers a third declaration from Tambling (Dkt. 184-1) and Amsallem (Dkt. 184-2), respectively. Each is materially similar to its second declaration counterpart. Compare Dkt. 169-2 with Dkt. 184-1 (Tambling’s declaration); Compare Dkt. 169-1 with Dkt. 184-2 (Amsallem’s declarations). Again, given these similarities, the court grants defendant’s motion to seal any information in categories two and five. The court otherwise denies defendant’s request. * * * As an appendix to her declarations, Tambling notes the portions of the various documents that defendant seeks to permanently seal. Dkt. 158-1 at 8-20 (appendix to first motion); Dkt. 169-2 at 8-14 (appendix to second motion): Dkt. 184-1 at 8-16 at (appendix to third motion). In its right column, each appendix specifies the basis for sealing each such portion. Id. Any portion that contains any information concerning categories two or five noted in defendant’s motions to seal may remain under seal. Defendant must file on the public docket any portion of any document sought for sealing that contains any information concerning only the remaining categories (categories one, three, or four). Of course, defendant must also file all portions of any document filed under provisional seal for which defendant did not seek permanent sealing. For the above reasons, the court DENIES plaintiff’s motion to strike and motion for motions to permanently seal the above noted information. Incident to that decision, the court TERMINATES all motions filed by plaintiff to provisionally seal any information that it filed under seal in connection with its motions. Within ten days of this order, defendant must file on the public docket all materials subject to disclosure as set forth above. The court will not reconsider its sealing decisions. Dated: August 24, 2021 /s/ Phyllis J. Hamilton PHYLLIS J. HAMILTON United States District Judge