Gurminder Singh v. Google LLC

Opinion

GURMINDER SINGH, Case No. 16-cv-03734-BLF

Plaintiff, ORDER DENYING MOTION FOR v. CLASS CERTIFICATION

GOOGLE LLC, [Re: ECF No. 134] Defendant.

This case concerns the AdWords program1 run by Defendant Google LLC. Through AdWords, Google sells to individuals and businesses of all sizes pay-per-click advertisements that are displayed on the Google Display Network, which consists of Google.com, other Google properties (such as YouTube and Gmail), and third-party sites who enroll in Google’s separate AdSense program. Plaintiff Gurminder Singh, a small business owner, signed up for AdWords in January 2008 and now controls multiple AdWords accounts. Singh alleges that Google deceives advertisers who use AdWords by making false and misleading statements concerning (1) how effectively Google identifies and filters out invalid and fraudulent clicks on advertisements; and (2) the proportion of total AdWords clicks that constitute invalid and fraudulent clicks. These misrepresentations allegedly induced him to sign up for AdWords and then pay for more invalid and fraudulent clicks than Google represented he would pay for. Over five years after filing this lawsuit, Singh seeks to represent an expansive class of all 1 As of July 24, 2018, AdWords is known as “Google Ads.” The Court uses “AdWords,” as the Parties do, recognizing that this was the name of the program for much of the putative class persons and entities who advertised and paid for clicks through AdWords since June 1, 2012, where the clicks originated from the Google Display Network. See ECF No. 134 (“Motion”); see also ECF No. 150 (“Reply”). Google opposes the Motion, arguing that Singh cannot satisfy the requirements of Federal Rule of Civil Procedure 23. ECF No. 142 (“Opp.”). The Court held a hearing on the Motion on December 2, 2021. ECF No. 160. For the reasons discussed on the record and explained below, Singh’s motion for class certification is DENIED. This case has an extensive history that is familiar to the parties and chronicled thoroughly in the Court’s previous orders. See ECF Nos. 64 (dismissing SAC); 85 (dismissing TAC); 104 (dismissing Fourth Amended Complaint, ECF No. 86 (“4AC”)). The Court here only recounts the alleged misrepresentations on which Singh’s case is based. Singh’s general theory is that Google misled advertisers participating in its AdWords program by making false and misleading statements concerning (1) how effectively Google identifies and filters out invalid and fraudulent clicks on advertisements; and (2) the proportion of total AdWords clicks that constitute invalid and fraudulent clicks. Singh focuses on statements made on two different pages of Google’s website. The first set of alleged misrepresentations is in the “Ad Traffic Quality Resource Center” (“ATQRC”), which specifies Google’s process for identifying invalid traffic. ECF No. 134-2 (“ATQRC”). The ATQRC states:

The relationship between Google, advertisers, and publishers is built on trust. Advertisers rely on the relevance of our ad placement, our reporting statistics, and the quality of clicks their ads receive. Publishers in turn count on advertiser participation, relevant ads which create a good experience for users, and an accurate and reliable source of income which contributes to the success of their websites and business. We take this trust seriously and we know that the Google advertising networks couldn’t exist without it. Id. The ATQRC explains the difference between what Google terms “click fraud”—“clicks generated with malicious or fraudulent intent”—and “invalid traffic”—“both clicks and impressions on AdWords ads that Google suspects to not be the result of genuine user interest.”2 Id. Google does not charge advertisers for invalid traffic. Id. Google then claims:

The vast majority of all invalid clicks on AdWords are caught by our online filters. These filters are constantly being updated and react to a wide variety of traffic patterns and indications of click fraud attacks. On average, invalid clicks account for less than 10% of all clicks on AdWords ads.” Id. The second set of challenged statements are in a February 28, 2007 post on the AdWords Blog. ECF No. 142-19 (“Blog Post”). The Blog Post contains similar statements as the ATQRC. The Blog Post describes Google’s recent efforts and performance in detecting click fraud, including the same methods explained in the ATQRC. The Blog Post describes how Google’s filters “[a]ccount for the vast majority of invalid click detection” and that “invalid clicks fluctuate constantly but average less than 10% of all clicks.” Id. The Blog Post also states that “the overall invalid clicks rate, as well as its day-to-day fluctuations, has almost no relation to the invalid clicks rate for an individual advertiser.” Id. Individual advertisers should “refer to [their] invalid clicks report for that data,” the Blog Post says. Id. According to Singh, the two pages make claims that are “intended to convince advertisers to sign up for AdWords, impress[] upon the reader that AdWords’ [pay-per-click] system was adequately tackling the scourge of [c]lick [f]raud.” Motion at 4. In fact, Singh says, his expert has found click fraud accounts for 14% of all clicks on the online advertising platforms, including on Google’s platform, which “significantly exceeds” Google’s 10% claim. ECF No. 134-3 ¶ 37. Singh claims that Google knows of this disparity and “conceal[s] the prevalence of [c]lick [f]raud on [its] platform, [which is] material information affecting all consumers.” Id. at 6. On July 13, 2021, Singh moved for class certification. See Motion. Singh seeks to certify and represent the following class: 2 The term “invalid traffic” used to be called “invalid clicks” before AdWords impressions were added to the definition of the term. See Opp. at 6. The parties do not dispute that this addition is All persons and entities throughout the United States who advertised through Google’s AdWords program and paid for clicks on their Google AdWords advertisement(s) at any time since June 1, 2012 (the “Class Period”), where such clicks originated from Google’s Display Network. Id. at Notice of Motion. His request for class certification is based on two claims asserted in the Fourth Amended Complaint for violations of the UCL and FAL. Id. Singh also seeks appointment of Miller Shah LLP and Edgar Law Firm LLC as class counsel. Id. The Court held a hearing on this Motion on December 2, 2021. ECF No. 160. A class action is maintainable only if it meets the four threshold requirements of Rule 23(a): (1) the class is so numerous that joinder of all members is impracticable; (2) there are questions of law or fact common to the class; (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class; and (4) the representative parties will fairly and adequately protect the interests of the class. Fed. R. Civ. P. 23(a); Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 613 (1997). “In addition to satisfying Rule 23(a)’s prerequisites, parties seeking class certification must show that the action is maintainable under Rule 23(b)(1), (2), or (3).” Amchem, 521 U.S. at 614. Certification under Rule 23(b)(1) is proper “where prosecuting separate actions by or against individual class members would create a risk of: (A) inconsistent or varying adjudications with respect to individual class members that would establish incompatible standards of conduct for the party opposing the class; or (B) adjudications with respect to individual class members that, as a practical matter, would be dispositive of the interests of the other members not parties to the individual adjudications or would substantially impair or impede their ability to protect their interests.” Rule 23(b)(2) requires that “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.” Rule 23(b)(3) requires that “questions of law or fact common to class members predominate over any questions affecting only individual members,” and that “a class action is superior to other available methods for fairly and “A party seeking class certification must affirmatively demonstrate his compliance with the Rule – that is, he must be prepared to prove that there are in fact sufficiently numerous parties, common questions of law or fact, etc.” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350 (2011). A. Standing Google’s first argument in opposition to class certification is that both Singh and other members of his putative class lack standing. Google says that Singh has not met his burden to demonstrate statutory standing as a potential class representative by showing that he actually paid for any invalid traffic or undetected click fraud. Opp. at 11. Google also says that Singh cannot show that he relied on the alleged misrepresentations and that Singh’s continued use of AdWords forecloses standing. Id. at 12. These problems will also plague members of the putative class, Google says. Id. at 12–13. Singh responds that the Court can defer consideration of standing until after ruling on his motion for class certification. Reply at 1–2. If the Court reaches the issue of standing, Singh argues that he has both statutory and Article III standing. He says Google’s argument to the contrary mischaracterizes his claims, which seek restitution for overpayment for the clicks he bought based on Google’s alleged misrepresentations about the level of average click fraud for AdWords ads. Id. at 3. This overpayment theory also forecloses Google’s argument about continued use of AdWords after seeing the representations and learning that they were misleading. Id. Courts generally must determine whether parties have standing prior to reaching the merits of a case. FW/PBS Inc. v. City of Dallas, 493 U.S. 215, 230–31 (1990). In multiple cases, however, the Supreme Court has considered class certification before standing. See Amchem, 521 U.S. at 612–13; Ortiz v. Fibreboard Corp., 527 U.S. 815, 831 (1999). There is a “growing consensus” among lower courts that class certification can be decided first “in situations in which the certification decision will itself shed light on the standing question.” 1 Newberg on Class Actions § 2:2 (5th ed. 2021) (“Newberg”) (citing cases); In re Carrier IQ, Inc. Customer Privacy “ha[d] discretion to defer questions of standing until after class certification”). Class certification is usually determined before standing where “it is apparent that class members will have standing even if the named class representative does not.” Svenson v. Google Inc., 2016 WL 8943301, at *4 (N.D. Cal. Dec. 21, 2016). This is “because once a class has been certified, ‘the class of unnamed persons described in the certification acquires a legal status separate from the representative.” Senne v. Kansas City Royals Baseball Corp., 114 F. Supp. 3d 906, 922 (N.D. Cal. 2014) (quoting In re Carrier IQ, Inc., 78 F. Supp. 3d 1051, 1072 (N.D. Cal. 2015)). The Court will exercise its discretion to determine Singh’s individual standing now. Because this order denies class certification, no class will “acquire[] a legal status separate from” Singh, the suit will become an individual lawsuit, and the Court need not consider whether all members of the class would have standing. Senne, 114 F. Supp. 3d at 922. With Singh the only plaintiff left, he must possess standing to proceed, so deferring consideration of his standing would be inappropriate. The Court finds that Singh has standing to pursue this case. The Ninth Circuit reversed this Court’s previous determination that Singh inadequately pled standing in the operative complaint. In the operative complaint, Singh alleges that he purchased clicks through AdWords, that Google has misrepresented the efficacy of its click filters, and that he would not have purchased clicks but for relying on those misrepresentations. 824 F. App’x 512, 513–14 (9th Cir. 2020). Singh has produced adequate evidence to support these allegations for standing purposes. “In a false advertising case, plaintiffs meet [the] [standing] requirement if they show that, by relying on a misrepresentation on a product label, they ‘paid more for a product than they otherwise would have paid, or bought it when they otherwise would not have done so.’” Reid v. Johnson & Johnson, 780 F.3d 952, 958 (9th Cir. 2015). Singh has provided evidence to support that argument, even if Google disputes it. Singh has testified that he read alleged misrepresentations about invalid clicks on the AdWords blog and that he relied on those statements before signing up for AdWords. See ECF No. 140-12 at 163:20–164:1. Singh’s position that he would not have paid as much for AdWords clicks if he had known that the average statements were in fact misrepresentations. See Pirozzi v. Apple, Inc., 966 F. Supp. 2d 909, 920 (N.D. Cal. 2013) (overpayment for a service is “lost money or property” for UCL standing. His continued use of AdWords is not a bar to pursuing this case. Robinson v. Unilever U.S., 2019 WL 8012687, at *4 (C.D. Cal. Aug. 21, 2019). Singh thus has individual standing to pursue his claims. B. Rule 23(a) i. Numerosity The prerequisite of numerosity is discharged if “the class is so large that joinder of all members is impracticable.” Hanlon v. Chrysler Corp., 150 F.3d 1011, 1019 (9th Cir. 1998), overruled on other grounds by Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338 (2011) (quoting Fed. R. Civ. P. 23(a)(1)). “Numerosity is presumed where the plaintiff class contains forty or more members.” Arroyo v. Int’l Paper Co., 2019 WL 1508457, at *2 (N.D. Cal. Apr. 4, 2019) (brackets and internal quotation marks omitted). There is no dispute here that the putative class is sufficiently numerous, as Google has far more than 40 customers of AdWords. ii. Commonality A class has sufficient commonality “if there are questions of fact and law which are common to the class.” Fed. R. Civ. P. 23(a)(2). Indeed, this requirement has been construed permissibly, and “[a]ll questions of fact and law need not be common to satisfy the rule.” Hanlon, 150 F.3d at 1019. “[P]laintiffs 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). In other words, the claims “must depend upon a common contention” such that “determination of its truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.” Dukes, 564 U.S. at 349–50. In his motion, Singh asserts that the common questions—which are “many and predominate”—stem from Google’s “misrepresentations and omissions regarding the general questions for the Court in that section, referring to his discussion of predominance despite the different standard. See id.; Amchem, 521 U.S. at 624 (predominance inquiry “far more demanding” than commonality requirement). In opposition, Google identifies the two common questions from Singh’s introduction: (1) “whether the filters’ actual efficacy captured 10% or more invalid traffic”; and (2) “whether that information was material to the reasonable advertiser.” Opp. at 13 (citing Motion at 2, 15). Singh first asserts that the actual efficacy of Google’s filters is a common question among the class members’ claims. Motion at 10; Reply at 4. In reply, he frames the relevant common question as “whether the network average of invalid clicks differed materially from Google’s representations.” Reply at 4. Google asserts that this issue is not subject to classwide proof because “actual efficacy” of Google’s filters varies for each ad campaign and thus each advertiser. Opp. at 13–14. Because each advertiser will experience different rates of click fraud, Google says, the fact-finder would necessarily have to compare the experience of each advertiser to the average rate claimed by Google. Id. at 14. The Court agrees with Singh that—at least for the purpose of commonality—this is a common question subject to classwide proof. Singh is correct that he asserts that Google has misrepresented the average level of invalid clicks on AdWords. On the ATQRC, Google represents that “[o]n average invalid clicks account for less than 10% of all clicks on AdWords ads.” See ATQRC; 4AC ¶ 47. Singh contends that the actual average amount of click fraud on the AdWords platform is 14% of clicks, which “significantly exceeds” Google’s claim, as shown by his proffered expert report. Motion at 4 (citing ECF No. 134-3 (“Cavazos Report”) ¶¶ 37, 40– 57). If this is indeed true, then the findings in the expert report would show that Google made a misrepresentation about the average level of click fraud. The answer to whether Google made a misrepresentation would, of course, “resolve an issue that is central to the validity of each one of the claims [of putative class members] in one stroke.” Dukes, 564 U.S. at 349–50. Google’s arguments revolve around a different framing of the question: determining the actual efficacy of Google’s filters for each advertiser and for each campaign. See Opp. at 13. invalid traffic for each advertising campaign run by different advertisers. Each advertising campaign will thus experience different rates of undetected click fraud. But this framing of the question is too granular and would defeat a commonality inquiry in every false advertising class action. While the answer to this Singh’s question alone will not fully establish Google’s liability for an alleged misrepresentation to each of the members of the putative class, the answer to that common question would resolve part of each of the class members’ claims. This suffices for a showing of commonality—although Google’s arguments are well-taken later in the Court’s class certification inquiry. This finding is in accord with the finding of commonality in a different advertising class action brought against Facebook in this district and on which Google relies. In IntegrityMessageBoards.com v. Facebook Inc., the plaintiff alleged that Facebook misled advertisers about its ability to accurately deliver advertisements to certain categories of users based on an advertiser’s specifications. 2021 WL 3771785, at *1 (N.D. Cal. Aug. 24, 2021) (“IMB”). Facebook allegedly made the misrepresentations in public remarks, statements on its website, and in the “targeting interface” through which advertisers specified to which categories of users its advertisements should be delivered. Id. The court found the commonality requirement satisfied based on the common question of whether Facebook “had knowledge about the misleading nature of its purported misstatements” because that question was “an element necessary for plaintiff to substantiate its claims.” Id. at *6–7. The court found that the plaintiff’s expert report, which showed that a “key corporate officer appreciated (or should have appreciated) the possibility or existence of material limits on [Facebook’s] ability to target users by Interests,” could provide an answer to that common question. Id. at *7. Thus, there was at least a “single question capable of classwide resolution” and the commonality requirement was satisfied. Id. (citing Wang, 737 F.3d at 544). The same is true here. The common question—whether Google misrepresented the average level of click fraud on AdWords—is “an element necessary for plaintiff to substantiate [his] claims” and “a single question capable of classwide resolution.” Id. at *6–7. claims, it need not consider for commonality purposes whether a reasonable advertiser considered Google’s representation about the average efficacy of its filters to be material. Wang, 737 F.3d at 544 (only a “single common question” required for commonality). This requirement is thus satisfied. iii. Typicality The typicality requirement looks to whether the claims of the class representative are typical of those of the class. The requirement is satisfied when each class member’s claim arises from the same course of events, and each class member makes similar legal arguments to prove the defendant’s liability.” Rodriguez v. Hayes, 591 F.3d 1105, 1124 (9th Cir. 2010) (brackets and internal quotation marks omitted). This requirement is “permissive and requires only that the representative’s claims are reasonably co-extensive with those of absent class members; they need not be substantially identical.” Id. (internal quotation marks omitted). “The purpose of the typicality requirement is to assure that the interest of the named representative aligns with the interests of the class.” In re Yahoo Mail Litig., 308 F.R.D. 577, 593 (N.D. Cal. 2015) (internal quotation marks omitted). The Ninth Circuit has said that that “class certification should not be granted if there is a danger that absent class members will suffer if their representative is preoccupied with defenses unique to [him].” Hanlon, 976 F.2d at 508 (internal quotation marks omitted). Singh argues that he meets the typicality because he has “suffered the same or similar injury as the class members through the same course of conduct by Google.” Motion at 10–11. He says that Google’s misrepresentations about the prevalence of click fraud caused him, like members of the putative class, to pay more for AdWords clicks than he otherwise would have. Id. at 11. Google offers several arguments that Singh does not meet the typicality requirement. The Court focuses on two: first, that Singh is atypical because he opted out of the AdWords arbitration agreement, unlike most advertisers in the putative class; and second, that Singh is atypical because not all advertisers in the putative class were exposed to or had the same understanding of Google’s allegedly misleading statements in the ATQRC or Blog Post. Opp. at 16–18. The Court agrees a. Arbitration Agreement with Class Action Waiver First, Google argues that Singh is atypical because he opted out of the AdWords arbitration agreement, unlike most advertisers in the putative class. See Opp. at 17. Singh himself alleges that the arbitration clause was introduced into the AdWords agreement in September 2017 and admits that he opted out of the arbitration provision. 4AC ¶ 11. Nevertheless, Singh says this is not a typicality problem because Google has not met its burden to make it “clear” that the issue of the arbitration clause would “swamp the litigation.” Reply at 6. The Court agrees with Google that Singh is atypical because he opted out of the arbitration clause to which many or most of the putative class is subject. “[C]lass certification is properly denied based upon the existence of an arbitration agreement and class action waiver appliable to unnamed class members but not the proposed class representative.” Farr v. Acima Credit LLC, 2021 WL 2826709, at *7 (N.D. Cal. Jul. 7, 2021); see also Tan v. Grubhub, Inc., 2016 WL 4721439, at *3 (N.D. Cal. Jul. 19, 2016) (plaintiff who opted out of arbitration clause and class action waiver not typical because he lacked standing to challenge applicability of those provisions, which would affect members of the putative class); Avilez v. Pinkerton Gov’t Servs., Inc., 596 F. App’x 579, 579 (9th Cir. 2015) (vacating class certification order that certified classes and subclasses that included employees who signed class action waivers but had a class representative who signed an agreement that did not contain a class action waiver). The Court finds that Singh cannot represent a class made up of advertisers who will largely be subject to a mandatory arbitration clause and class action waiver. Signatories to those agreements would have to overcome them to be included in the class, but Singh could not challenge the agreements because he did not sign them. Tan, 2016 WL 4721439, at *3. The Court disagrees with Singh that this issue is not relevant to class certification because it would not “threaten to become the focus of the litigation.” Reply at 6 (citing Rodriguez v. Hayes, 591 F.3d 1105, 1124 (9th Cir. 2010), and Hanon v. Dataproducts Corp., 976 F.2d 497, 508 (9th Cir. 1992)). Neither Rodriguez nor Hanlon considered the specter of an arbitration clause or class action waiver to which many putative class members but not the named plaintiff are subject. become the focus of the litigation.” Hanon, 976 F.2d at 508. Whether the arbitration clause and class action waiver apply here is a necessary condition for many members of the putative class to join this class action. If those provisions apply, those advertisers are precluded from participating in this case. Singh would have to litigate this issue to include those advertisers, but he would lack standing to do so because he opted out of the arbitration clause and class action waiver.3 The Court also rejects Singh’s argument that Google has failed to meet its burden to provide the number of opt-outs or produce the terms of the arbitration agreement. Reply at 6. Singh primarily relies on Farr, saying that the court there only found a typicality problem because it considered an “evidentiary submission about the number of opt outs” and “extensively analyz[ed] the terms of the relevant arbitration clause.” Id. That characterization of Farr is inaccurate. In fact, the Farr court expressly noted that the “exact number of [opt outs] is not known,” and that defendant merely indicated that “a number of customers are potentially subject to the arbitration clause and class action waiver.” Farr, 2021 WL 2826709, at *7 n.6. The exact number of opt outs is not known here, but it is very likely that most of the putative class of advertisers did not opt out of the arbitration clause. Additionally, the Farr court only engaged in a lengthy analysis of the arbitration clause because there was a dispute about whether it applied to the claims in the lawsuit. See id. at *4–6 (analyzing plaintiff’s argument that the equitable claims in the lawsuit were not subject to mandatory arbitration). Here, Singh himself recognizes the existence of the arbitration clause and has not disputed its validity or applicability. 4AC ¶ 11; see also Trudeau v. Google, 349 F. Supp. 3d 869, 876–81 (N.D. Cal. 2018) (upholding enforceability of AdWords arbitration clause), aff’d, 816 F. App’x 68 (9th Cir. 2020). 3 If Singh declined to litigate this issue, he would have to exclude any advertiser who did not opt out of the arbitration clause and class action waiver from his class definition. The Court will not modify the putative class definition after the conclusion of briefing on this Motion. Because of the other issues the Court identifies in this order, a renewed class certification motion with a narrower b. Assumptions Underlying Singh’s Injury Second, Google argues that Singh is atypical because his purportedly typical injury, as applied to the putative class, would improperly assume that (1) all AdWords advertisers were exposed to the alleged misrepresentations; and (2) based on that exposure, they all had the same understanding of those statements. Opp. at 16. Google says that there is no evidence favoring the inference that all members of the putative class were exposed to the alleged misstatements because advertisers do not need to view the ATQRC or Blog Post to sign up for AdWords or manage their accounts. Id. Google also says that because many sophisticated advertisers use advertising agencies or have at least previous online advertising experience, they would not have come away with the same impression of the statements as Singh. Id. at 16–17. Singh responds that Google’s argument does not address his omissions theory based on Google’s failure to disclose the actual efficacy of its invalid click filters. Id. Singh also argues that he does not need to prove that all members of the putative class had the same understanding of the alleged misrepresentations, and that differences in sophistication do not implicate whether all class members suffered the same injury. Id. at 5–6. The Court finds that Singh adequately pled two omissions theories, although they are thin. The operative complaint does state that his claims are based partially on Google’s omissions. See, e.g., 4AC ¶ 117 (identifying “material omissions” regarding Google’s click filter mechanisms, including that Google’s filters, “in fact, do not catch a ‘vast majority of all invalid clicks’ before advertisers are charged”). Singh’s first omissions theory is actually a partial misrepresentation theory because it is an outgrowth of the affirmative statements that Google made on the ATQRC and Blog Post regarding the number of fraudulent or invalid clicks. As Google argues, that theory relies on Google’s alleged “obligation to clarify or provide further information as to the filter’s actual efficacy to render its prior representations and silence not misleading.” Opp. at 22 (citing Motion at 2). If a putative class member did not view the statements in the ATQRC or Blog Post, then this theory is not implicated. The issue of exposure to and reliance on the alleged misstatements thus remains for that partial misrepresentation theory. In contrast, Singh’s second misstatements. “[I]rrespective of the affirmative representations,” Singh says, Google had a “general duty to disclose the actual efficacy of its filters” because Google has “exclusive knowledge” of the actual efficacy of the filters, which is material to users of AdWords. Id. Google vigorously disputes the merits of this theory, which the Court discusses infra Section III.C.i, but for present purposes it is adequately pled. Nevertheless, the Court finds that Singh does not have a typical injury under any of his theories. For Singh’s affirmative misrepresentation and partial misrepresentation theories, his typical injury assumes that all AdWords advertisers were exposed to the alleged misstatements. That assumption is unjustified and unsubstantiated. Google has provided evidence showing that advertisers do not need to view the ATQRC or Blog Post to sign up for AdWords or place an advertisement. See Opp. at 16 (citing Bjorke Decl., ECF No. 142-2 ¶ 19). Google has also provided evidence that many AdWords advertisers use advertising agencies to place ads and manage accounts, meaning those members of the putative class never would have seen the AdWords interface at all, much less the alleged misrepresentations. Nelson Decl. Ex. B., ECF No. 153-2 at 73:2–9. Without exposure to the alleged misstatements, a putative class member could not rely on those misstatements (or be misled by Google’s omission of additional information needed to clarify those statements) and thus could not be injured. That makes Singh’s injury atypical of members of the putative class who did not view the alleged misstatements. See IMB, 2021 WL 3771785, at *7–8 (class representative atypical where he failed to substantiate assumption that putative class members viewed the misstatements due to (1) lack of need to view them in typical use of interface and (2) use of third party ad agencies).4 But even if exposure to the alleged misstatements was not an issue—whether under Singh’s two theories that depend on exposure or the one pure omissions theory that does not— Singh has failed to show that his injury would be typical due to the varying levels of advertising sophistication in his class. Even if putative class members were exposed to the misstatements that 4 A presumption of exposure to and reliance on the alleged misstatements is not warranted for the underlie his first two theories, Singh has failed to “make an objective showing of a probability that a ‘significant portion’ of the relevant consumers acting reasonably ‘could be misled’ by the challenged statements.” Hadley v. Kellogg Sales Co., 324 F. Supp. 3d 1084, 1116 (N.D. Cal. 2018).5 For his final theory, Google’s duty to disclose depends on its “exclusive knowledge” of facts that are unknown to members of the class. LiMandri v. Judkins, 52 Cal. App. 4th 326, 336 (1997). But many of the advertisers in the putative class are likely to have “a degree of preexisting marketing experience that tends to preclude them from reasonably developing a ‘false impression’ about [Google’s click filtering] abilities.” IMB, 2021 WL 3771785, at *8 (plaintiff atypical where his false impression about Facebook’s advertisement targeting abilities was not likely to be shared by more sophisticated advertisers in the putative class); see also In re Facebook, Inc., PPC Advert. Litig., 282 F.R.D. 446, 454 (N.D. Cal. 2012). Singh’s own expert admits as much. See, e.g., Nelson Decl., Ex. B at 18:5–9 (“What a small business person up the street from you knows about – you know, click ad fraud or advertising, versus what Procter & Gamble knows or Bank [of] America, their advertising agency knows, are completely different universes.”) These putative class members would not share Singh’s false impression about the efficacy of Google’s click filters—whether based on Google’s affirmative misrepresentations or not—making Singh’s injury further atypical for a potentially “significant portion” of the putative class. Singh thus fails to satisfy the typicality requirement. iv. Adequacy To “fairly and adequately protect the interests of the class,” Fed. R. Civ. P. 23(a)(4), “[a] class representative must be part of the class and ‘possess the same interest and suffer the same injury’ as the class members.” Rodriguez, 563 F.3d at 959 (quoting E. Tex. Motor, 431 U.S. at 403). Courts generally resolve two questions in determining whether plaintiffs will adequately represent the class: “(1) do the named plaintiffs and their counsel have any conflicts of interest 5 Here, where Singh’s putative class includes advertisers of all sizes and levels of sophistication, with other class members; and (2) will the named plaintiffs and their counsel prosecute the action vigorously on behalf of the class?” Ellis, 657 F.3d at 985 (internal quotation marks omitted). “An absence of material conflicts of interest between the named plaintiffs and their counsel with other class members is central to adequacy and, in turn, to due process for absent members of the class.” Rodriguez, 563 F.3d at 959. Google argues that Singh would not fairly and adequately protect class members’ interests because he has misled this court and the Ninth Circuit about the allegations in this case. Google says Singh continues to claim that the real rate of undetected click fraud on Google’s network to be orders of magnitude higher than his own expert’s estimates of 1–4%. Opp. at 18. Google says that Singh has made arguments that he knew were not supported by evidence, including the Oxford tests, and that he cannot represent sophisticated companies that spend much more per month that he does on advertising on AdWords. Id. Singh responds that he satisfies the adequacy inquiry because he does not have conflicts of interest and will vigorously litigate on behalf of the class. The Court agrees with Google that Singh is not an adequate class representative for the same reasons that his claims are atypical of the proposed class. First, Singh’s proposed class includes all AdWords advertisers who vary in level of sophistication and ad spending on AdWords. Singh seeks to represent AdWords advertisers who spend amounts of money that are orders of magnitude higher per month than he spends per year on AdWords. Compare ECF No. 140-20 (records showing Singh spent low four-figure dollar amounts per year on AdWords ads), with ECF No. 142-24 (report stating that Amazon and Home Depot spend $31.4 million and $6.81 million per month on AdWords ads). This large disparity in size and sophistication of Singh and the members of the proposed class represents an adequacy problem. See In re Facebook, Inc., PPC Advert. Litig., 282 F.R.D. at 454. Second, Singh’s election to opt-out of the AdWords arbitration clause (which includes a class action waiver) presents an adequacy problem in addition to a typicality issue. See Farr, 2021 WL 2826709, at *7 (citing Avilez, 596 F. App’x at 579). Singh is thus an inadequate class representative. v. Conclusion Singh does not satisfy the requirements of Rule 23(a) because his claims and defenses are not typical of those of the putative class and because he is an inadequate class representative. C. Rule 23(b)(3) Because Singh has not fulfilled the typicality or adequacy requirements of Rule 23(a), his proposed class cannot proceed. Nevertheless, the Court analyzes the proposed class under Rule 23(b)(3) to identify additional issues with Singh’s Motion. i. Predominance Singh seeks certification under Rule 23(b)(3), which requires that “questions of law or fact common to class members predominate over any questions affecting only individual members,” and that “a class action is superior to other available methods for fairly and efficiently adjudicating the controversy.” Fed. R. Civ. P. 23(b)(3). The “predominance inquiry tests whether proposed classes are sufficiently cohesive to warrant adjudication by representation.” Amchem, 521 U.S. at 623. The predominance test of Rule 23(b)(3) is “far more demanding” than the commonality test under Rule 23(a)(2). Amchem, 521 U.S. at 624. Though common issues need not be “dispositive of the litigation,” In re Lorazepam & Clorazepate Antitrust Litig., 202 F.R.D. 12, 29 (D.D.C. 2001), they must “present a significant aspect of the case [that] can be resolved for all members of the class in a single adjudication” so as to justify “handling the dispute on a representative rather than an individual basis.” Hanlon, 150 F.3d at 1022. Courts must therefore separate the issues subject to “generalized proof” from those subject to “individualized proof” to determine whether plaintiffs have satisfied the predominance requirement. See In re Dynamic Random Access Memory (DRAM) Antitrust Litig., No. M 02-1486, 2006 WL 1530166, at *6 (N.D. Cal. June 5, 2006) (“Predominance requires that the common issues be both numerically and qualitatively substantial in relation to the issues peculiar to individual class members.” (internal quotation omitted)). Whether the predominance requirement is satisfied in a particular case “turns on close scrutiny of ‘the relationship between the common and individual issues.’” In re Wells F.3d at 1022). Singh identifies “two common and predominate issues”: “[1] whether the filters’ actual efficacy captured 10% or more [of] Invalid Traffic; and [2] based on the prior findings on the filters’ actual efficacy, whether that information was material to the reasonable advertiser and Google had an obligation to clarify or provide further information as to the filters’ actual efficacy to render its prior representations and silence not misleading.” Motion at 2, 14–15. Singh says that Google misrepresented the actual efficacy of its filters both affirmatively and by omission, and that the efficacy was material to a reasonable consumer because access to actual customers is the “raison d’être [of] AdWords.” Id. at 12–15. On his pure omissions theory, Singh says that Google—independent of any affirmative representations it made—had a general duty to “disclose the actual efficacy” of its filters because it made a material omission about a defect central to the functioning of AdWords of which it had exclusive knowledge. Id. at 13-14 (citing LiMandri, 52 Cal. App. 4th at 326). Google argues that neither of Singh’s issues is subject to common proof. Opp. at 13–15, 19. But Google also identifies numerous other individual issues that “control the outcome of the claims” and that it argues predominate over the common issues. Id. at 19. These include “whether each advertiser saw the alleged misrepresentations,” “whether they arrived at the same misinterpretation of those statements as [Singh];” and “whether each advertiser attached any significance to the statements;” among multiple others. Id. Google argues in particular that the individual issue of whether an advertiser was exposed to the alleged misrepresentations is a “core issue” that would swamp any common issues in the litigation. Opp. at 20–21. Because the class cannot include individuals who were not exposed to the misrepresentation, the Court would need to conduct individual inquiries to determine if given class members did view them before assessing any of the other requirements for the claims. Id. Singh’s only response to this point is that he need not prove that absent class members relied on particular misrepresentations or omissions to certify a UCL class. Reply at 8 (citing In re Tobacco Cases, 207 P.3d 20, 31 (Cal. 2009)). Google contests that it has any freestanding duty to disclose the actual average efficacy of The Court finds that even assuming Singh’s two issues are common issues subject to classwide proof, individual issues would predominate over those questions. a. Affirmative Misrepresentations and Partial Misrepresentations The most pressing individual issue for Singh’s theory of affirmative misrepresentations and his first omissions theory (which necessarily depends on the affirmative misrepresentations) would be that of exposure to and reliance on the alleged misrepresentations. “Class certification of UCL claims is available only to those class members who were actually exposed to the business practices at issue.” Berger v. Home Depot USA, Inc., 741 F.3d 1061, 1069 (9th Cir. 2014), abrogated on other grounds by Microsoft Corp. v. Baker, 137 S. Ct. 1702 (2017); see also Pfizer Inc. v. Superior Court, 182 Cal. App. 4th 622, 631 (2010) (“One who was not exposed to the alleged misrepresentations and therefore could not possibly have lost money or property as a result of the unfair competition is not entitled to restitution.”). As the Court already explained in the analysis of typicality, advertisers do not need to view the ATQRC or Blog Post to sign up for AdWords or place an advertisement. See Opp. at 16 (citing ECF No. 142-2 ¶ 19). Without exposure to the alleged misstatements, an advertiser could not rely on those misstatements (or be misled by Google’s omission of additional information) and thus could not lose money or property under California consumer protection statutes. Singh is incorrect that he need not show that absent class members were exposed to the alleged misrepresentations to certify a class, and his reliance on Tobacco II is misplaced. In Tobacco II, the California Supreme Court concluded that absent class members did not need to satisfy the actual reliance requirement of the UCL “where class requirements have otherwise been found to exist.” 46 Cal. 4th at 324. But courts since Tobacco II have universally recognized that this holding “was in the context of a ‘decades-long’ tobacco advertising campaign where there was little doubt that almost every class member had been exposed to [the tobacco companies’] misleading statements.” Mazza v. Am. Honda Motor Co., Inc., 666 F.3d 581, 596 (9th Cir. 2012). “In the absence of the kind of massive advertising campaign at issue in Tobacco II, the relevant class must be defined in such a way as to include only members who were exposed to advertising 631 (9th Cir. 2020) (“To establish a reliance presumption, the operative question has become whether the defendant so pervasively disseminated material misrepresentations that all plaintiffs must have been exposed to them.”). A recent advertising class action in this district is instructive. In dotStrategy Co. v. Facebook Inc., 2021 WL 2550391 (N.D. Cal. Jun. 22, 2021), Judge Alsup denied class certification in a case challenging Facebook’s representations that it would not charge advertisers when “fake accounts” clicked on their ads. The three alleged misrepresentations—which said that Facebook would not charge advertisers for invalid clicks, that advertisers would only pay to reach “the right people who’ll love your business,” and that Facebook was a site “where everyone uses the name they go by in everyday life”—were located on separate pages in various help center sections of the Facebook website. Id. at *9. Facebook provided evidence that none of the pages that advertisers had to interact with to place and manage ads on Facebook displayed the alleged misrepresentations, and that “[u]sers who wish[ed] to view such pages must [have] search[ed] for them or navigat[ed] to them in some other way.” Id. Judge Alsup concluded that no presumption of reliance was warranted because “the statements appeared only on several of thousands of webpages on Facebook’s website . . . none of which an advertiser would have necessarily viewed at any point in buying and placing ads on Facebook.” Id. at *10. The Court has no difficulty reaching the same conclusion here. The statements here are buried on two pages on Google’s website, neither of which need to be viewed by an advertiser to sign up for AdWords or place an ad. Singh has provided no evidence suggesting that advertisers would be exposed to those misrepresentations in any way other than by affirmatively accessing one of those two pages on Google’s websites. The alleged misrepresentations are thus not so “pervasively disseminated” as to justify a presumption that putative class members were exposed to them. And Singh’s omissions theory does not save him. As the Court previously explained, while the omissions theory was adequately pled, it necessarily depends on exposure to the misrepresentations on the ATQRC and Blog Post. The analysis thus collapses into analysis of exposure to the misrepresentations and results in the same conclusion. Mazza accordingly exposed to advertising that is alleged to be misleading.” Because Singh’s class definition is not so defined, this issue is one that predominates over common issues. Singh thus does not satisfy the predominance inquiry. b. Pure Omissions Theory Singh’s pure omissions theory—that Google had a general duty to disclose the actual efficacy of its AdWords filters—fares no better and does not allow him to satisfy predominance. Courts in this district have recognized that “California law on pure omissions claims is less settled” than that on affirmative misrepresentations or partial omissions. Anderson v. Apple, 500 F. Supp. 3d 993, 1014 (N.D. Cal. 2020). When a defect does not relate to an unreasonable safety hazard, courts in this district have concluded that a defendant has a duty to disclose when “(1) the omission is material; (2) the defect is central to the product’s function; and (3) at least one of the following four [LiMandri] factors is met: the defendant is the plaintiff’s fiduciary; the defendant has exclusive knowledge of material facts not known or reasonably accessible to the plaintiff; the defendant actively conceals a material fact from the plaintiff; or the defendant makes partial representations that are misleading because some other material fact has not been disclosed.”6 In re Apple Inc. Device Performance Litig., 386 F. Supp. 3d 1155, 1176 (N.D. Cal. 2019) (citing Hodsdon v. Mars Inc., 891 F.3d 857, 863 (9th Cir. 2018)). Singh’s pure omissions theory suffers from its own issues, two of which the Court identifies here. Singh argues, under the LiMandri factors, that Google had “exclusive knowledge” of and “actively concealed” the “actual efficacy of its [c]lick [f]raud filters,” Motion at 13, but the evidence contradicts those assertions. Google does not conceal the actual efficacy of its filters. Each advertiser has access to the percentage of fraudulent clicks detected per campaign. See ECF No. 140-20 (Singh’s campaign statistics, including “[i]nvalid click rate” per campaign). No 6 The last LiMandri factor concerns a partial representation theory, which the Court addressed in the previous subsection. It is accordingly inapplicable to a pure omissions theory and the Court does not consider it here. See Anderson, 500 F. Supp. 3d at 1014. Singh does not contend that the information about actual efficacy was thus concealed. Even if Singh is arguing that Google concealed the average efficacy of its click filters, that omission would not be material to a reasonable advertiser where that advertiser has access to the actual efficacy of the click filters per campaign that the advertiser runs. Additionally, as the Court already discussed, the putative class contains advertisers who are likely to have “a degree of preexisting marketing experience” such that they would not all “attach importance” to a 1–4% difference in average efficacy of the click filters—especially where they would have access to the actual efficacy of those filters for each of their campaigns. See IMB, 2021 WL 3771785, at *16 (plaintiff failed to provide evidence that all advertisers, who varied greatly in level of sophistication, would attach importance to “the fact that defendant could target users with advertisements at less than 50 percent accuracy”). Singh’s pure omissions theory thus does not allow him to satisfy predominance on his two identified issues. Accordingly, Singh fails to satisfy predominance. ii. Superiority To satisfy Rule 23(b)(3), Singh also must demonstrate that “a class action is superior to other available methods for fairly and efficiently adjudicating the controversy.” Fed. R. Civ. P. 23(b)(3). Rule 23 lists the following factors that Courts should consider in making this determination:

(A) the class members’ interests in individually controlling the prosecution or defense of separate actions; (B) the extent and nature of any litigation concerning the controversy already begun by or against class members;

(C) the desirability or undesirability of concentrating the litigation of the claims in the particular forum; and (D) the likely difficulties in managing a class action. The Court notes that some of the same concerns that animate the rulings on typicality and predominance. Individual issues of exposure to the alleged misrepresentations, for example, also present “likely difficulties in managing a class action.” Fed. R. Civ. P. 23(b)(3)(D). Some of the other factors, however, would counsel in favor of a class action. For example, individual class behalf of class here. Valentine v. Carter-Wallace, Inc., 97 F.3d 1227, 1234 (9th Cir. 1996). On balance, had the other Rule 23 factors been met, the superiority requirement would have been satisfied. iii. Damages Model Rule 23(b) also requires that Singh show that “damages are capable of measurement on a classwide basis.” Comcast Corp. v. Behrend, 569 U.S. 27, 34 (2013). Courts are required to “conduct a ‘rigorous analysis’ to ensure at the class certification stage that ‘any model supporting a plaintiff’s damage case [is] consistent with its liability case.’” Id. at 35. This requires “that the model ‘measure[s] only those damages attributable to that theory’ of liability.” IMB, 2021 WL 3771785, at *17 (quoting Rikos v. Procter & Gamble Co., 799 F.3d 497, 523 (6th Cir. 2015)). “The mere fact that there might be differences in damage calculations is not sufficient to defeat class certification.” In re Hyundai & Kia Fuel Economy Litig., 926 F.3d 529, 560–61 (9th Cir. 2019). Singh offers a damages expert, Dr. Roberto Cavazos, in support of his damages model. Dr. Cavazos proposes to calculate damages by multiplying (1) the actual rate of click fraud in excess of the click fraud that reasonable consumers would expect by the total revenue Google derives from AdWords. Singh characterizes the resulting damages—which Dr. Cavazos posits is over $12 billion—as restitution because “the Class would only recover their . . . payments that did not result in reaching real potential customers as Google’s misrepresentations and omissions indicated.” Motion at 15–16. Google argues that the proffered damages model is not in fact restitutionary. Opp. at 23–25. A “true restitutionary model,” Google says, would need to account for the value that class members received from AdWords, “which would vary based on a number of factors” including “the benefits of AdWords over other advertising platforms,” false positives for which Google does not charger advertisers, and the “economic reality that the cost of click fraud is already ‘built-in’ to the price for non-fraudulent clicks.” Id. at 24. The Court need not delve too far into Singh’s damages model given his failure to satisfy predominance. Nevertheless, the Court does conclude that Singh’s damages theory as he has reason than is offered by Google in its opposition. Singh’s damages model subtracts the amount of fraudulent clicks that he says a reasonable consumer would expect Google’s click filters to detect—in other words, 10% of clicks on AdWords ads, as Google says in its ATQRC—from the “actual” percentage of fraudulent clicks on AdWords ads, which Dr. Cavazos estimates to be 14%. This damages model would make Google liable for any undetected fraudulent clicks that pass through its filters. This would make Google a guarantor that advertisers never pay for a fraudulent click—in other words, for any undetected click fraud. But that model is divorced from Singh’s theory of liability, which is not based on any guarantee that Google makes. The two misrepresentations Singh relies upon say that “[o]n average invalid clicks account for less than 10% of all clicks on AdWords ads” and that “invalid clicks fluctuate constantly but average less than 10% of all clicks.” See ATQRC; Blog Post. Those statements are not guarantees that Google will catch all invalid clicks. Indeed, the same paragraph of the ATQRC in which the alleged misrepresentation is found also states that Google captures “the vast majority of all invalid clicks,” not all invalid clicks. See ATQRC. The AdWords Agreement—to which all advertisers agree before using AdWords—also makes clear that Google cannot detect all click fraud, and that if an advertiser believes some click fraud has gone undetected, that it should use Google’s claims process to pursue a refund. See ECF No. 86-1 ¶ 7 (“Customer understands that third parties may generate impressions or clicks on Customer’s Ads for prohibited or improper purposes and that its sole remedy is to make a claim for advertising credits within the Claim Period, after which Google will issue the credits following claim validation . . . .”). Because the alleged misrepresentations on which Singh bases his theory of liability are not guarantees that all click fraud is detected, and Singh’s damages model seeks damages in a way that would make Google the guarantor that all click fraud is detected, Singh’s damages model is not tethered to his theory of liability and violates Comcast. / / / / / / / / / 1 Singh thus fails to meet the requirements of Rule 23(b)(3) and cannot certify a class under 2 that provision of Rule 23. 3 IV. ORDER 4 For the foregoing reasons, IT IS HEREBY ORDERED that Singh’s motion for class 5 certification is DENIED. The existing case schedule (ECF Nos. 121, 128) remains in effect. 6 7 Dated: January 10, 2022 tn) 8 M: BETH LABSON FREEMAN 9 United States District Judge 10 11 a 12

13 14

15 16

= 17 6 Z 18 19 20 21 22 23 24 25 26 27 28