Phhhoto Inc. v. Meta Platforms, Inc.

123 F.4th 592

Opinion

23-763 Phhhoto Inc. v. Meta Platforms, Inc.

UNITED STATES COURT OF APPEALS FOR THE SECOND CIRCUIT

August Term 2023

(Argued: November 16, 2023 Decided: December 10, 2024)

No. 23-763

–––––––––––––––––––––––––––––––––––– PHHHOTO INC.,

Plaintiff-Appellant,

-v.-

META PLATFORMS, INC., FKA FACEBOOK, INC., Defendant-Appellee,

DOES NOs. 1-7,

Defendants. *

––––––––––––––––––––––––––––––––––––

Before: LIVINGSTON, Chief Judge, WESLEY, and CHIN, Circuit Judges.

This appeal requires us to decide whether a plaintiff claiming unlawful monopolization under the Sherman Act sufficiently alleged the fraudulent concealment theory of equitable tolling. Plaintiff-Appellant Phhhoto Inc.

*

The Clerk of the Court is directed to amend the official caption as set forth above.

2

(“Phhhoto”) alleges that one of the world’s largest technology companies, Defendant-Appellee Meta Platforms, Inc. (“Meta”), engaged in a scheme to injure Phhhoto’s business through anticompetitive means, including the adoption of an algorithmic feed for Instagram that purportedly suppressed Phhhoto’s content on that platform. More than four years after the new algorithm was introduced, Phhhoto filed the instant action, alleging in relevant part that Meta’s shift to an algorithmic feed, in combination with certain of its earlier acts, constituted an anticompetitive course of conduct in violation of Section 2 of the Sherman Act. The United States District Court for the Eastern District of New York (Matsumoto, J.) dismissed this claim under Federal Rule of Civil Procedure 12(b)(6) after determining that it had accrued outside of the Sherman Act’s four-year statute of limitations and that equitable tolling could not save it from untimeliness. On appeal, Phhhoto argues that the amended complaint sufficiently alleges Meta’s fraudulent concealment of an anticompetitive scheme and that the district court therefore erred in dismissing the antitrust claim as time-barred. Reviewing the record de novo, we agree with Phhhoto that it adequately alleged that the Sherman Act’s four-year statute of limitations should be equitably tolled until October 25, 2017. Accordingly, we VACATE the district court’s judgment and REMAND for further proceedings.

Judge Chin dissents in a separate opinion.

FOR PLAINTIFF-APPELLANT: SCOTT MARTIN (Irving Scher, on the brief), Hausfeld LLP, New York, NY; Sarah LaFreniere, Hausfeld LLP, Washington, D.C.

(Phillip F. Cramer, Sperling & Slater, LLC, Nashville, TN; Eamon P. Kelly & Nathan A.

Shev, Sperling & Slater, LLC, Chicago, IL, for Josh Davis, Christopher R. Leslie, Robert H.

Lande, Peter C. Carstensen, John B.

Kirkwood, Edward D. Cavanaugh, Darren Bush & Harry First, as amici curiae)

3

(Nada Djordjevic, DiCello Levitt LLP, Chicago IL; Gregory S. Asciolla, DiCello Levitt LLC, New York, NY; Robin A. van der Meulen, Scott & Scott Attorneys at Law LLP, New York, NY; Kristen Marttila, Lockridge Grindal Nauen P.L.L.P., Minneapolis, MN, for The Committee to Support the Antitrust Laws, as amicus curiae)

FOR DEFENDANT-APPELLEE: AARON M. PANNER (Alex Treiger, on the brief), Kellogg, Hansen, Todd, Figel & Frederick, P.L.L.C., Washington, D.C.

DEBRA ANN LIVINGSTON, Chief Judge:

Plaintiff-Appellant Phhhoto Inc. (“Phhhoto”) appeals from a March 31, 2023 judgment of the United States District Court for the Eastern District of New York (Matsumoto, J.) dismissing its amended complaint against Defendant-Appellee Meta Platforms, Inc. (“Meta”) as time-barred. In relevant part, Phhhoto’s amended complaint alleges that Meta engaged in a course of unlawful monopolization in violation of Section 2 of the Sherman Act, 15 U.S.C. § 2. The district court dismissed this claim as untimely, holding that it accrued outside of the Sherman Act’s four-year statute of limitations and that it could not be saved by equitable tolling.

The essence of Phhhoto’s antitrust claim is that Meta used anticompetitive means, starting in or around 2015, to exclude Phhhoto from the personal social

4

networking services market. This alleged anticompetitive conduct included, inter alia, withdrawing Phhhoto’s access to certain features of the Instagram platform on which Phhhoto relied, terminating a joint project to incorporate Phhhoto into the Facebook newsfeed, and releasing an app that replicated Phhhoto’s technology. Even as they called into question the viability of Phhhoto’s business strategy, which heavily relied on Meta’s Instagram platform, these acts were not as significant as Meta’s decision to adopt an algorithmic feed for Instagram in March 2016. The new algorithm represented a shift from the chronological feed that Instagram had used from its inception.

In the wake of the new algorithm, Phhhoto’s popularity faded as quickly as it had previously surged. Meta justified its shift to an algorithmic feed in neutral terms—claiming that the new feed was based on factors such as user interest in the post, user relationship with the posting account, and recency of the post. However, despite the popularity of its platform and the high engagement of its users, Phhhoto experienced a sharp decline in new user registrations and user engagement following Meta’s adoption of the algorithmic feed.

With its user metrics and prospects for funding plummeting, Phhhoto “worked tirelessly” to identify the reason for its sudden decline in popularity. A-

5

116–17, ¶ 92. Owing in part to Meta’s neutral description of the algorithm, Phhhoto did not suspect algorithmic suppression as a potential explanation until October 25, 2017, when one of its co-founders, Champ Bennett, stumbled upon information suggesting a probability that Meta had engaged in anticompetitive behavior. At this point, Phhhoto had gone out of business and its co-founders had rejoined their prior company, Hypno, which had a negligible presence on social media. In an effort to “connect Phhhoto’s remaining Instagram followers to Hypno,” A-119, ¶ 104, Bennett posted an identical video promoting Hypno to Phhhoto’s old Instagram account and Hypno’s new one. This led to two surprising observations. First, Phhhoto’s post “appeared to vanish” from Bennett’s personal Instagram feed. A-120, ¶ 105. Second, while Phhhoto had “approximately 500 times” more followers than Hypno, Hypno’s post received more views and “likes” compared to Phhhoto’s. Id. ¶ 106. Based on these observations, Bennett and his co-founders began to investigate whether Meta was using its algorithmic feed to suppress competitive third-party content on Instagram.

The district court never reached the merits of Phhhoto’s antitrust claim, holding instead that it was time-barred under the Sherman Act’s four-year statute

6

of limitations. Conceding that its antitrust claim is untimely absent equitable tolling, Phhhoto argues on appeal that the district court erred in declining to toll the statute of limitations based on fraudulent concealment. We agree. After conducting our own independent review of the allegations in Phhhoto’s amended complaint, we conclude that Phhhoto is presently entitled to equitable tolling of the Sherman Act’s limitations period until October 25, 2017, such that Phhhoto’s antitrust claim is timely for purposes of Meta’s motion to dismiss. 1 Accordingly, we VACATE the district court’s judgment and REMAND for further proceedings.

BACKGROUND

I. Factual Background 2

1 Before Phhhoto’s complaint was filed, the parties agreed to toll “any applicable statutes of limitations” for fourteen days. Thus, equitably tolling the Sherman Act’s four-year statute of limitations to October 25, 2017 renders Phhhoto’s antitrust claim timely at this stage. 2 The factual background presented here is derived from the allegations in Phhhoto’s amended complaint, which we “accept as true” and construe “in the light most favorable to the plaintiff[]” for purposes of reviewing the district court’s dismissal under Rule 12(b)(6). Nat’l Credit Union Admin. Bd. v. U.S. Bank Nat’l Ass’n, 898 F.3d 243, 252 (2d Cir. 2018) (alteration adopted and citation omitted).

7

A. Phhhoto’s Early Success Phhhoto was founded in 2012 by Champ Bennett, Omar Elsayed, and Russell Armand. Two years later, Phhhoto launched to the public as a photography and social networking app. Through the app, users could “capture[] five frames in a single point-and-shoot burst and link[] them together into a looping video,” creating an animated photo known as a “phhhoto.” A-87, ¶ 1.

Phhhoto quickly gained popularity. Meta’s CEO, Mark Zuckerberg, created an account on Phhhoto’s platform soon after it launched. Other Meta executives quickly followed suit. But Phhhoto’s popularity was not confined to tech circles, as celebrities also joined the platform. At one point, Phhhoto had a rate of new user growth exceeding even that of Instagram.

B. Phhhoto and Meta’s Initial Collaboration In its early days, Phhhoto seemed to have a symbiotic relationship with Meta. After acquiring the popular photo- and video-sharing app, Instagram, in April 2012, Meta continued to facilitate interoperability between Instagram and third-party apps such as Phhhoto. This interoperability was achieved largely through a feature known as iPhone Hooks, which allowed users to post content

8

created on third-party apps directly to Instagram and permitted captions to be “pre-populate[d]” with “a hashtag identifying the developer of the content.” A- 99–100, ¶ 34. These hashtags were a promotional tool for third-party apps, allowing users to identify the app from the caption of an Instagram post and then to “download it [from Apple’s App Store] for their own use.” A-100, ¶ 35. iPhone Hooks “positioned Instagram as a hub for other apps” and thus “as a key platform in the market for personal social networking services.” Id.

Phhhoto relied on this interoperability feature to reach new users and increase user engagement. The ability to share “phhhotos” on Instagram drove much of Phhhoto’s business growth, with “more people seeing . . . phhhotos [on Instagram] and [subsequently] downloading [the Phhhoto] app than through other platforms.” A-104, ¶ 48.

In turn, Phhhoto’s success created an opportunity for Meta to “expand[] its user engagement . . . among younger audiences.” A-107, ¶ 58. Indeed, Meta twice attempted to collaborate with Phhhoto. First, in February 2015, the Strategic Partnerships Manager for Meta’s Facebook business, Bryan Hurren, offered “to incorporate Phhhoto’s technology into the Messenger service on Meta’s Facebook platform.” A-90, ¶ 8. Phhhoto declined, concluding that it would not

9

benefit Phhhoto’s business. A few weeks later, Hurren shared with Bennett a second proposal—this time, “to integrate Phhhoto into the Facebook newsfeed.” A-106–07, ¶ 57. This integration would allow users to “post phhhotos to the Facebook newsfeed directly through the Phhhoto app,” id., and permit phhhotos to appear in their native animated format on Facebook, as they did on Instagram. Excited about the prospect of reaching more Facebook users, Phhhoto offered to lead the technical integration.

C. Meta’s Anticompetitive Scheme Based on this early history, Phhhoto and Meta appeared well positioned to continue developing a seemingly symbiotic business relationship. From Meta, Phhhoto could gain exposure and develop a user base; from Phhhoto, Meta could attract users from new generations and revitalize its brand. At some point in 2015, however, there was a shift in the companies’ relationship, based—as claimed by Phhhoto—on Meta’s concern that Phhhoto presented a competitive threat to Meta’s dominance. This shift might have occurred very early in that year, in which case Meta’s purported interest in collaborating with Phhhoto was merely obscuring Meta’s “scheme to crush Phhhoto and drive it out of business.” A-90, ¶ 9. Or the scheme might have developed more gradually, ultimately

10

superseding—but not rendering disingenuous—Meta’s early efforts to team up with Phhhoto. Either way, the scheme culminated in March 2016, when Meta adopted an algorithmic feed for Instagram that resulted in Phhhoto’s demise as an operational company.

a. Find Friends API

On March 31, 2015, Meta “suddenly withdrew Phhhoto’s access” to an application programming interface (“API”) on Instagram known as the “Find Friends API.” A-109, ¶ 65. The Find Friends API allowed third-party apps, such as Phhhoto, “to access the Instagram friends list” and to create a “social graph” for their users. Id. As a digital representation of a user’s personal network, the social graph “provides the foundation for users” to interact with one another. A- 130–31, ¶ 137. Without the ability to recreate Instagram’s social graph, Phhhoto’s relationship with potential investors would suffer, prompting Bennett to contact Hurren about the withdrawal. Hurren informed Bennett that the change was made because “Meta was . . . upset that Phhhoto was growing in users through its relationship with Instagram.” A-109–10, ¶ 67.

11

b. Facebook Newsfeed Integration Project In another about-face, Meta “surreptitiously terminated the project . . . for integrating Phhhoto’s content into the Newsfeed of Meta’s Facebook platform.” A-90, ¶ 9. Meta strung “Phhhoto along for months without making meaningful progress on the [proposed] integration.” A-107, ¶ 59. Meta repeatedly delayed the proposed launch, citing purported legal and technical concerns, even as “Phhhoto worked diligently to meet all of Facebook’s specifications.” A-108, ¶ 62. At some point between February and June 2015, Meta quietly “decided to abandon the project.” A-108, ¶ 63.

c. Pre-Populated Hashtags Meta abandoned another interoperability feature on August 9, 2015, withdrawing the component of iPhone Hooks that allowed third-party developers to pre-populate captions with hashtags. Meta explained publicly that this change was due to “feedback that the pre-filled captions . . . often feel spammy,” A-111, ¶ 71, and encouraged third-party developers to use watermarks to identify their content instead. This suggestion was fruitless, however, as it ignored a critical distinction between hashtags and watermarks: whereas the former are located

12

entirely in captions, the latter are overlaid on posts and thus “disfigure or cover up content that would not otherwise be impacted by a hashtag.” Id. ¶ 73.

d. Boomerang

Next, in October 2015, Meta introduced a new app called “Boomerang” that mirrored—was a “clone” of—Phhhoto’s technology. A-90, ¶ 9. Meta unveiled Boomerang on the official launch date of Phhhoto’s app for Android devices. To promote the new app, Meta also adopted self-preferencing policies that boosted Boomerang’s performance while lowering that of third-party apps. For example, after eliminating the pre-populated hashtag capability two months earlier, Meta added an automatic “Made with Boomerang” caption to all posts originating from the app. A-113, ¶¶ 81–82 (citation omitted). So, while third-party developers could no longer identify content made on their apps, Meta required users to attribute their posts to Boomerang, even including a link in each caption that led users to download the app.

e. Algorithmic Feed

Finally, in March 2016, Meta adopted an algorithmic feed for Instagram.

From its inception, Instagram had a non-algorithmic feed that displayed posts in reverse chronological order, such that “the newest posts would appear first in a

13

user’s feed” and the oldest would appear last. A-114, ¶ 85. Without disclosing the mechanics of the new algorithm, Meta explained that it was “based on the likelihood [the user will] be interested in the content, [the user’s] relationship with the person posting [the content] and the timeliness of the post.” A-114–15, ¶ 86. Meta attempted to downplay the significance of this change, with Kevin Systrom, the co-founder and former CEO of Instagram, assuring the public that “it’s not like people will wake up tomorrow and have a different Instagram.” A-115, ¶ 88 (alteration adopted).

But this is exactly what happened to Phhhoto. While Phhhoto expected that its popularity and high user engagement would prove beneficial under the new algorithm, the performance of its content suggested otherwise. Following the adoption of the algorithmic feed, Phhhoto’s business began to crumble, with new user registrations and existing user engagement “plummet[ing].” A-126–27, ¶ 125. New user registrations “declined precipitously,” and existing users disengaged, “post[ing] less content, comment[ing] less frequently, and shar[ing] or favorit[ing] content shared by others less frequently.” A-116, ¶ 91. Phhhoto was “rapidly failing,” A-126–27, ¶ 125, and its performance in Apple’s App Store

14

showed it: from April 1, 2016 to May 1, 2016, Phhhoto’s ranking among photo and video apps dropped from 11th to 41st place.

With its metrics continuing to decline, Phhhoto could not rally investors to fund the company. Phhhoto could not resuscitate its business and decided to shut down its app on June 20, 2017.

D. Phhhoto’s Discovery In the early days of the algorithmic feed, “Phhhoto’s team worked tirelessly” to pinpoint the cause of the app’s “sudden[] unpopular[ity].” A-116– 17, ¶ 92. Assuming the decline was related to an internal coding issue, the team “spent months . . . running analytics to determine if Phhhoto had bugs that were causing the app to crash on users.” Id. Phhhoto’s co-founders also considered alternative explanations “such as competition from Boomerang and cyclical usage.” A-117, ¶ 93. These potential causes were “ruled . . . out,” id., leaving Phhhoto’s co-founders no discernible way to save the app from slipping into obscurity.

After Phhhoto shut down in June 2017, its co-founders “returned to work at their prior company, Hypno,” which “provide[s] camera platforms and interactive experiences for live events, retail, and attractions.” A-119, ¶ 104. To boost the

15

“little social media presence” Hypno had, Bennett sought to advertise the company to “Phhhoto’s remaining Instagram followers.” Id. At the time, Phhhoto had approximately 500 times more followers than Hypno. So, on October 25, 2017, Bennett posted the same promotional video to two Instagram accounts—Phhhoto’s old account and Hypno’s new one—with the goal of drawing Phhhoto’s followers to Hypno.

But this strategy did not play out as Bennett anticipated. To start, Bennett observed that “the post from the old Phhhoto account appeared to vanish from [his] own Instagram feed.” A-120, ¶ 105. Additionally, Hypno’s post outperformed Phhhoto’s—with the former receiving 100 views and the latter receiving 36—even though the posts were identical and Phhhoto had significantly more followers on Instagram. Phhhoto’s post was also “liked” 50% less frequently than Hypno’s. Like Bennett, “other Phhhoto followers . . . were not seeing posts from Phhhoto’s old Instagram account.” Id. ¶ 106.

Phhhoto’s team now had reason to doubt the accuracy of Meta’s public description of the algorithmic feed and to investigate whether Meta was engaged in “purposeful suppression of Phhhoto’s content.” A-120–21, ¶ 107 (emphasis omitted). Bennett asked a tech journalist that day whether he had “ever heard

16

anything about Instagram using [its] algorithmic feed to suppress competitive apps in [the] photo/video space.” A-121–22, ¶ 110.

Eventually, public information emerged suggesting that Meta had engaged in algorithmic suppression and other forms of anticompetitive conduct. In December 2018, a committee of the U.K. Parliament “publicly released documents” that were previously “produced confidentially” in federal litigation in California, revealing that some of Meta’s early exclusionary acts toward Phhhoto were part of an anticompetitive scheme that included algorithmic suppression. A-129, ¶ 133. The New York Times later identified the name of this scheme—“Project Amplify”—and was the first to report that “Meta did, in fact, manipulate and reorder posts and content in users’ newsfeeds to benefit Meta.” A-123, ¶ 114.

II. Procedural Background Phhhoto filed the instant action on November 4, 2021, asserting an antitrust claim under the Sherman Act and state-law causes of action for fraud, unfair competition, and deceptive acts or practices. After Meta indicated its intent to move to dismiss Phhhoto’s complaint under Federal Rule of Civil Procedure 12(b)(6), the district court held a pre-motion conference in which it requested that

17

Phhhoto file an amended complaint to address “some . . . deficiencies” in the original one. A-69–70. Phhhoto filed an amended complaint on March 21, 2022. 3 Meta moved to dismiss on June 6, 2022. In opposing Meta’s motion to dismiss, Phhhoto argued in relevant part that the statute of limitations for its antitrust claim was subject to equitable tolling based on fraudulent concealment and that the claim was therefore timely.

The district court disagreed, dismissing as time-barred the Sherman Act claim in Phhhoto’s amended complaint without further leave to amend. 4 Having dismissed the only claim over which it had original jurisdiction, the district court declined to exercise supplemental jurisdiction over the remaining state-law claims. A final judgment to that effect issued on March 31, 2023. Phhhoto appealed.

3 Phhhoto’s amended complaint omits the cause of action for deceptive acts or practices in violation of New York General Business Law § 349, but otherwise replicates the claims that were brought in the original complaint. 4 Although the district court declined to exercise supplemental jurisdiction over Phhhoto’s state-law claims, the district court dismissed those claims “with prejudice.” A-218. This was error. A dismissal after declining to exercise supplemental jurisdiction must be without prejudice. See Kolari v. N.Y. Presbyterian Hosp., 455 F.3d 118, 119 (2d Cir. 2006); see also Carnegie-Mellon Univ. v. Cohill, 484 U.S. 343, 350 (1988). In any event, because we vacate the portion of the district court’s judgment dismissing Phhhoto’s federal-law claim as time-barred—the basis for the district court’s decision to decline to exercise supplemental jurisdiction over the state-law claims—we also vacate the portion of the district court’s judgment dismissing the state-law claims.

18

DISCUSSION

On appeal, Phhhoto argues that the district court erred at each step of the fraudulent concealment analysis and thus in its conclusion that Phhhoto’s antitrust claim is untimely. We agree. Reviewing the district court’s tolling determination de novo, we hold that, for purposes of Meta’s motion to dismiss, the allegations in the amended complaint entitle Phhhoto to equitable tolling of the statute of limitations for its Sherman Act claim until October 25, 2017. These allegations present multiple factual disputes—corresponding to the elements of the fraudulent concealment test—that are not resolvable on a motion to dismiss and instead should be developed through appropriate discovery.

I. Standard of Review In its briefing, Phhhoto argues that the district court erroneously declined to equitably toll the Sherman Act’s four-year statute of limitations, “whether reviewed as an abuse of discretion or de novo.” Appellant’s Reply Br. at 1.

The operative standard of review for equitable tolling determinations depends on “what aspect of the lower court’s decision is challenged: a legal conclusion, a factual finding, or an exercise of discretion.” Doe v. United States, 76 F.4th 64, 70 (2d Cir. 2023) (internal quotation marks and citation omitted). “[W]e

19

review the legal premises for [the district court’s] conclusion de novo, the factual bases for clear error, and the ultimate decision [to deny equitable tolling] for abuse of discretion.” DeSuze v. Ammon, 990 F.3d 264, 268 (2d Cir. 2021). 5 Accordingly, we apply de novo review where the district court denied equitable tolling based “on the belief that the decision was compelled by law,” Phillips v. Generations Fam. Health Ctr., 723 F.3d 144, 149 (2d Cir. 2013) (citation omitted), or where the “asserted error is legal in nature,” Clark v. Hanley, 89 F.4th 78, 104 (2d Cir. 2023).

Here, we conclude that a de novo standard governs our review. To begin, the district court declined to equitably toll the statute of limitations based on its “belief” that this result “was compelled by law.” Phillips, 723 F.3d at 149 (citation omitted). The district court concluded that Phhhoto failed to meet the legal prerequisites for equitable tolling, reaching that conclusion “as a matter of law.” A-214 (determining that Phhhoto could not satisfy the reasonable diligence prong of the fraudulent concealment test “as a matter of law”); see also A-189 (“[T]his Court finds that Phhhoto’s factual allegations are insufficient to satisfy the three

5 It is undisputed that the district court did not make factual findings at this motion to dismiss stage. See Roth v. Jennings, 489 F.3d 499, 509 (2d Cir. 2007) (“[A] ruling on a motion for dismissal pursuant to Rule 12(b)(6) is not an occasion for the court to make findings of fact.”). As such, neither party contends that our review of the district court’s equitable tolling determination is for clear error.

20

elements for fraudulent concealment . . . .”). Next, Phhhoto ultimately challenges the district court’s “legal conclusion[s]” on appeal, Doe, 76 F.4th at 70, identifying multiple “error[s]” in the decision below that are “legal in nature,” Clark, 89 F.4th at 104. These asserted legal errors pertain to the elements of fraudulent concealment, which Phhhoto argues “are more than . . . support[ed]” by the allegations in the amended complaint—contrary to the district court’s holding below. Appellant’s Br. at 29. 6 After determining that Phhhoto failed to plead fraudulent concealment, the district court was not faced with a discretionary call. “Before a court may exercise discretion to grant equitable tolling, a litigant must demonstrate as a factual matter the existence of [various] elements.” Doe, 76 F.4th at 71. “The law prohibits a judge from exercising her discretion where these . . . elements are missing.” Id. In other words, for the equitable tolling decision to become

6 See also Appellant’s Br. at 23 (“[T]he district court wrongly held that Phhhoto failed to plead that Meta engaged in concealment at all.”); Appellant’s Br. at 36 (“Nor do any of the three events identified by the district court as purported ‘additional storm warnings’ . . . suggest the probability of . . . an anticompetitive scheme.”) (citation omitted); Appellant’s Br. at 45 (“In the context of all of Phhhoto’s efforts, the district court’s cited authority concerning appropriate due diligence is plainly inapposite here.”).

21

discretionary, the district court needed to conclude that the doctrine could apply in the first place.

The district court reached the opposite conclusion here, holding that none of the three elements of the fraudulent concealment test was adequately pled. This determination resolved a binary question of law—whether Phhhoto sufficiently pled the elements of fraudulent concealment. The district court was not faced with a “range of possible permissible decisions.” Phillips, 723 F.3d at 149 (citation omitted). Because the district court determined that Phhhoto had not pled the elements of fraudulent concealment—all of which are “necessary predicate[s] for equitable tolling” based on fraudulent concealment, Belot v. Burge, 490 F.3d 201, 207 (2d Cir. 2007)—the district court had no choice but to decline to equitably toll Phhhoto’s antitrust claim. Thus, in dismissing Phhhoto’s antitrust claim as untimely, the district court reached legal conclusions that we review de novo.

II. Fraudulent Concealment Private antitrust claims are “forever barred” if not brought “within four years after the cause of action accrued.” 15 U.S.C. § 15b. A cause of action for an alleged antitrust violation accrues “when a defendant commits an act that

22

injures a plaintiff’s business.” Higgins v. N.Y. Stock Exch., Inc., 942 F.2d 829, 832 (2d Cir. 1991) (quoting Zenith Radio Corp. v. Hazeltine Rsch., Inc., 401 U.S. 321, 338 (1971)). For purposes of this appeal, Phhhoto does not dispute that its cause of action under the Sherman Act accrued no later than April 2016, when Phhhoto felt “the adverse impact of” Meta’s anticompetitive scheme. Id. (citation omitted). Because Phhhoto did not bring suit until November 4, 2021—more than four years later—its antitrust claim is time-barred unless subject to equitable tolling.

As with other statutes of limitations, the Sherman Act’s four-year time bar can be equitably tolled only in “rare and exceptional circumstance[s].” Smith v. McGinnis, 208 F.3d 13, 17 (2d Cir. 2000) (per curiam) (citation omitted). One such circumstance is where the plaintiff can demonstrate fraudulent concealment. See New York v. Hendrickson Bros., Inc., 840 F.2d 1065, 1083 (2d Cir. 1988). To prove fraudulent concealment, an antitrust plaintiff must establish: “(1) that the defendant concealed from him the existence of his cause of action, (2) that he remained in ignorance of that cause of action until some point within four years of the commencement of his action, and (3) that his continuing ignorance was not attributable to lack of diligence on his part.” Id. The plaintiff must plead the

23

fraudulent concealment with particularity, in accordance with Federal Rule of Civil Procedure 9(b). See Armstrong v. McAlpin, 699 F.2d 79, 88–89 (2d Cir. 1983).

The second element has been framed in two different—and potentially conflicting—ways, leading to confusion in the lower courts. See, e.g., Nat’l Grp. for Commc’ns & Computs. Ltd. v. Lucent Techs. Inc., 420 F. Supp. 2d 253, 265 n.15 (S.D.N.Y. 2006) (noting that “[t]here has been some variation in the way that the three elements [of the fraudulent concealment test] have been articulated” by this Court). First, in New York v. Hendrickson Brothers, Inc., we included as a requirement of pleading fraudulent concealment that the plaintiff “remained in ignorance of [its] cause of action until some point within four years of the commencement of [its] action” (hereinafter, the “Hendrickson formulation”). 840 F.2d at 1083 (emphasis added). A decade later, in In re Merrill Lynch Limited Partnerships Litigation, we modified the Hendrickson formulation—inadvertently, in our view—asking whether the concealment “prevented [the plaintiff’s] discovery of the nature of the claim within the limitations period.” 154 F.3d 56, 60 (2d Cir. 1998) (per curiam) (emphasis added). This language has been repeated in subsequent cases, see Corcoran v. N.Y. Power Auth., 202 F.3d 530, 543 (2d Cir. 1999), including a recent one that the district court cited as the governing standard

24

for fraudulent concealment, Koch v. Christie’s Int’l PLC, 699 F.3d 141, 157 (2d Cir. 2012) (hereinafter, the “Koch formulation”).

This difference in phrasing could be outcome determinative in a case where, as here, the plaintiff discovered its cause of action within a limitations period that began at the moment of injury. Applied literally, the Koch formulation would deny relief to this plaintiff, solely because its discovery happened to occur “within the limitations period.” Id. In contrast, the Hendrickson formulation would not treat as determinative this aspect of the timing inquiry; so long as the plaintiff brought suit “within four years” of when its claim was discovered, the second element of the fraudulent concealment test would be satisfied. 840 F.2d at 1083.

Notwithstanding the Koch formulation, we have been clear that “when a ‘defendant fraudulently conceals the wrong, the time limit of the statute of limitations does not begin running until the plaintiff discovers, or by the exercise of reasonable diligence should have discovered, the cause of action.’” Pinaud v. Cnty. of Suffolk, 52 F.3d 1139, 1157 (2d Cir. 1995) (alteration adopted) (citing Keating v. Carey, 706 F.2d 377, 382 (2d Cir. 1983)). Under this principle, we afford significance to the date of the plaintiff’s discovery, without regard to whether it occurred within the usual limitations period. The Hendrickson formulation does

25

the same, rendering it consistent with our case law and the purpose of the fraudulent concealment doctrine. For that reason, Hendrickson should be applied as the governing articulation of the second element of the fraudulent concealment test. 7 III. Fraudulent Concealment In This Case Turning to the crux of the parties’ dispute on appeal, we conclude that Phhhoto has sufficiently alleged fraudulent concealment with respect to its antitrust claim to survive a motion to dismiss. The district court erred in concluding otherwise.

A. Phhhoto Has Adequately Pled Concealment.

To prevail on the first element of the fraudulent concealment test, the plaintiff must establish that “the defendant concealed from [the plaintiff] the existence of [its] cause of action.” Hendrickson, 840 F.2d at 1083. To plead concealment, the plaintiff can either: (1) identify “affirmative steps” taken by the

7 Hendrickson involved, as here, a Sherman Act claim with a four-year statute of limitations. 840 F.2d at 1069. In Hendrickson, we therefore stated that the plaintiff must have remained in ignorance “until some point within four years” of commencing the action. Id. at 1083. Of course, in a case in which the claim did not arise under the Sherman Act, the “four years” in Hendrickson’s second element would need to be substituted with the appropriate limitations period for the claim.

26

defendant to conceal the plaintiff’s claim; or (2) show that the defendant’s misconduct was inherently “self-concealing.” Id. In this case, Phhhoto argues that it has done both, pointing to certain of Meta’s statements as affirmative acts of concealment and the algorithmic feed as inherently self-concealing.

We conclude that Phhhoto has sufficiently pled that Meta “took affirmative steps,” through its public statements about the algorithm, “to prevent [Phhhoto]’s discovery of [its] claim.” Id. The statements on which Phhhoto relies, and to which the district court alluded in its concealment analysis, include: (1) that portion of Meta’s press release, issued in March 2016, indicating that the new “order of photos and videos in [the user’s] feed will be based on the likelihood [the user will] be interested in the content, [the user’s] relationship with the person posting and the timeliness of the post,” A-114–15, ¶ 86; and (2) a statement by a Meta representative at a 2018 press conference “disclaim[ing] that [Meta] either hid posts in its newsfeed, engaged in shadowbanning . . . or favored a format (photo or video), except to the extent an individual was more likely to engage with a particular format,” A-122, ¶ 112 (emphasis omitted). We focus on the first of these statements—the only one preceding Phhhoto’s discovery of its antitrust

27

claim—in evaluating whether Phhhoto sufficiently alleged that Meta affirmatively concealed from Phhhoto its Sherman Act claim.

The district court answered this question in the negative for two primary reasons. First, the district court faulted Phhhoto for failing to explain in its amended complaint “why Meta’s algorithm,” if implemented as Meta described in the 2016 press release, “would have optimized” rather than penalized Phhhoto’s content on Instagram. A-199. Second, the district court found it unreasonable for Phhhoto to have relied on Meta’s public statements, which it viewed as mere puffery. Each of these rationales is unpersuasive.

As to its first rationale, the district court posited that because Meta never guaranteed success to any particular content creator, Meta did not conceal the possibility that Phhhoto’s metrics could decline under the new algorithm. But Phhhoto’s allegations of concealment relate to the design, not the ultimate effects, of the algorithmic feed. In its amended complaint, Phhhoto alleges that Meta’s 2016 press release was “misleading” because it failed to list suppression of competitive third-party content as a factor driving the new algorithm. A-90–91, ¶ 10.

Meta argues that, even if its 2016 press release failed to explain “how . . . the algorithm would treat content posted from other apps,” this omission is

28

tantamount to silence and thus cannot constitute concealment. Appellee’s Br. at 26. We disagree. Meta is correct that the 2016 press release did not explicitly disclaim the possibility that aspects of the algorithmic feed were disfavoring content from competitors. But the absence of an express denial does not necessarily equate to silence. Rather, in listing three specific criteria on which the algorithm was based, Meta implied that other factors, such as the posting account’s status as a competitor, were not in play. See In re Polyurethane Foam Antitrust Litig., 152 F. Supp. 3d 968, 1006 (N.D. Ohio 2015) (defendants’ proffered reasons for a price increase were “affirmative acts of concealment,” despite their accuracy, because “a portion of the announced price increase” was due to undisclosed “collusion with competitors”). This is Phhhoto’s argument: by announcing the algorithm’s criteria, but only its neutral criteria, Meta downplayed the significance of—and allayed any possible anticompetitive suspicion about— Instagram’s shift to an algorithmic feed. Far from constituting silence, the 2016 press release was an “affirmative step[],” Hendrickson, 840 F.2d at 1083—“intended to exclude suspicion and prevent inquiry,” Wood v. Carpenter, 101 U.S. 135, 143 (1879)—in Meta’s alleged anticompetitive scheme to suppress third-party content and eliminate Phhhoto as a competitor.

29

Nor was Meta’s 2016 press release “inactionable ‘puffery.’” 8 City of Pontiac Policemen’s & Firemen’s Ret. Sys. v. UBS AG, 752 F.3d 173, 183 (2d Cir. 2014). Our precedent is clear that “general statements about reputation, integrity, and compliance with ethical norms are inactionable ‘puffery.’” Id. Some courts in this Circuit have gone further, as the district court did below, concluding that “communications to the community at large will not generally support a finding of fraudulent concealment.” A-200 (citing In re Merrill, Bofa, & Morgan Stanley Spoofing Litig., No. 19-cv-6002, 2021 WL 827190, at *11 (S.D.N.Y. Mar. 4, 2021) (“In re Merrill”), aff’d sub nom. Gamma Traders - I LLC v. Merrill Lynch Commodities, Inc., 41 F.4th 71 (2d Cir. 2022)); see also Litovich v. Bank of Am. Corp., 568 F. Supp. 3d 398, 440 (S.D.N.Y. 2021).

Even accepting that it was a “communication[] to the community at large,”

In re Merrill, 2021 WL 827190, at *11, Meta’s 2016 press release does not constitute puffery. The contrasting facts of City of Pontiac and In re Merrill—cases to which the district court cited in its puffery analysis and in which the relevant statements

8 To be sure, the district court’s puffery analysis only expressly pertained to Meta’s 2018 statement about shadowbanning. The district court reasoned that the 2018’s statement public and “general” nature made it puffery. A-200 (citation omitted). Because this broad reasoning would be equally applicable to the 2016 press release, we discuss it here.

30

were considered puffery—make this conclusion clear. There, the statements in question only vaguely referenced the company’s ethical and legal obligations. In City of Pontiac, the statements were about “compliance, reputation, and integrity,” 752 F.3d at 183; in In re Merrill, they assured investors that the defendants “maintain[ed] established procedures that ensure[d] compliance with all applicable laws and regulations,” 2021 WL 827190, at *9 (citation omitted). The specificity of Meta’s 2016 press release distinguishes it from these statements. Instead of merely signaling the company’s compliance with general legal obligations, Meta included in the 2016 press release a factual representation about the criteria on which the algorithmic feed operated. As such, the 2016 press release was neither “general,” nor broadly “about reputation, integrity, and compliance with ethical norms.” City of Pontiac, 752 F.3d at 183. In these circumstances, it was reasonable for Phhhoto to “rely on the statement[] to allay [any] concern” about Meta’s shift to an algorithmic feed. LC Cap. Partners, LP v. Frontier Ins. Grp., Inc., 318 F.3d 148, 155 (2d Cir. 2003). 9

9 Having concluded that Meta’s 2016 press release constitutes an affirmative act of concealment sufficient to support equitable tolling at the motion to dismiss stage, we need not now address Phhhoto’s alternative argument that the algorithmic feed is self- concealing.

31

B. Phhhoto Has Adequately Pled Lack of Notice.

As the preceding discussion makes clear, the second element of the fraudulent concealment test requires the antitrust plaintiff to establish that it “remained in ignorance of [its] cause of action until some point within four years of the commencement of [its] action.” Hendrickson, 840 F.2d at 1083. This element is not met, and the plaintiff’s claim of fraudulent concealment is defeated, if the plaintiff—more than four years prior to the commencement of its action— had either: (1) “actual notice,” or (2) “inquiry notice” of the facts giving rise to its claim, and thereafter failed to inquire. Dodds v. Cigna Secs., Inc., 12 F.3d 346, 350 (2d Cir. 1993). The standard for inquiry notice is objective, imputing knowledge to the plaintiff where “storm warnings” would cause “a person of ordinary intelligence [to] consider it ‘probable’” that the defendant had engaged in wrongdoing. Koch, 699 F.3d at 151 (citations omitted). To find inquiry notice “as a matter of law” on a motion to dismiss, there must be “uncontroverted evidence clearly demonstrat[ing] when the plaintiff should have discovered the [challenged] conduct.” Staehr v. Hartford Fin. Servs. Grp., Inc., 547 F.3d 406, 427 (2d Cir. 2008).

32

The district court identified four “storm warnings” that, in its view, should have made Phhhoto aware of Meta’s allegedly exclusionary behavior as early as April 2016. These purported “storm warnings” parallel the components of Meta’s alleged anticompetitive scheme: (1) the withdrawal of access to Instagram’s Find Friends API; (2) the abandonment of its proposal to integrate Phhhoto into the Facebook newsfeed; (3) the removal of the pre-populated hashtag capability; and (4) the introduction of Boomerang. The district court determined that these acts, each of which occurred prior to the algorithm’s launch, would have “aroused the suspicions of a reasonable business in Phhhoto’s situation,” triggering a duty to investigate whether Meta’s conduct was anticompetitive. A-205.

But none of Meta’s prior acts—whether viewed individually or collectively—made it “probable” that Meta was engaged in an anticompetitive scheme to destroy Phhhoto’s business. Koch, 699 F.3d at 151 (citations omitted). Until at least June 2015, during the period of the purported “storm warnings,” Meta appeared to continue to collaborate with Phhhoto, dispelling any suspicion that Meta had an ulterior motive to exclude Phhhoto from the market. After withdrawing Phhhoto’s access to the Find Friends API in March 2015, for example, Hurren exchanged emails with Bennett until early June 2015 about the Facebook

33

newsfeed integration project. Based on these communications, Phhhoto reasonably could have believed that Meta had a sincere interest in the companies’ continued collaboration, making it unlikely—and certainly not “probable”—that Meta’s true intent was anticompetitive. Id. (citations omitted).

Meta also offered plausible, non-exclusionary justifications for some of its adverse acts, making it seem improbable that they were anticompetitive. To start, Meta explained its decision to withdraw the pre-populated hashtag capability as based on “feedback that the pre-filled captions . . . often feel spammy.” A-111, ¶ 71. As for the Facebook newsfeed integration project, Meta repeatedly justified its delays by citing a need for further technical specifications and potential legal concerns. From a reasonable business’s perspective, these explanations lowered the probability that Meta’s acts were anticompetitive, suggesting instead that Meta was encountering legitimate roadblocks in its attempts to collaborate with Phhhoto.

Our prior cases finding a duty to inquire based on “storm warnings”

highlight the improbability that Meta’s pre-algorithm conduct would be seen as anticompetitive. For example, in Koch v. Christie’s International PLC, we held that certain “storm warnings” should have alerted the plaintiff, by at least October

34

2000, to the “probability” that the defendant had fraudulently attributed certain bottles of wine to Thomas Jefferson’s collection. Koch, 699 F.3d at 153. These “storm warnings” included that: (1) the plaintiff was aware of numerous articles in the early 1990s casting doubt on the authenticity of the wine he had purchased; (2) during the same period, he discovered a lawsuit against the seller alleging that the wine was counterfeit; and (3) in October 2000, after he submitted samples of the wine for radiocarbon testing, the testing lab reported a greater than 90% probability that the wine was from a period other than that engraved on the bottle. Id. at 146–47, 153. Given these “storm warnings,” we concluded that the plaintiff was on inquiry notice “[a]t least by October 16, 2000,” when the lab’s authenticity report was issued. Id. at 153.

None of the Koch “storm warnings” has an analogue in the present record.

Most significantly, Phhhoto did not receive—and presumably had no way to obtain, as Meta’s elusive algorithm is not discoverable through laboratory testing—a quantitative assessment of the likelihood that Instagram’s new algorithmic feed was anticompetitive. Nor did information suggesting that Meta was engaged in an exclusionary scheme emerge publicly until 2018. Phhhoto alleges that it “did not become aware of Meta’s overall campaign against

35

competitors . . . until the release of internal Meta documents by the UK Parliament in December of 2018.” A-91, ¶ 11. Phhhoto’s access to news articles on the subject came even later. Phhhoto alleges that the first news report documenting “Meta’s manipulative and deceptive conduct with respect to users’ newsfeeds” was published in September 2021—nearly four years after Phhhoto had already discovered this information for itself. A-123, ¶ 114. Thus, unlike the plaintiff in Koch, Phhhoto’s co-founders did not have at their fingertips public reports implicating Meta in an anticompetitive scheme until the algorithmic feed had been in place for nearly two years and its exclusionary nature had ultimately been discerned. See A-90–91, ¶ 10 (“[O]nly []after [October 2017] did further information emerge to reveal that Meta . . . had in fact been purposely suppressing Instagram users’ posts that contained Phhhoto content.”); cf. Staehr, 547 F.3d at 408, 417–18, 421 (concluding that multiple storm warnings, including four prior lawsuits alleging the same contingent commission scheme, widespread reporting of those allegations in mainstream media, and publication of an article in an industry journal highlighting the defendant’s potential involvement in this scheme, were insufficient to give rise to inquiry notice).

36

Our holding as to the alleged “storm warnings”—that they fall short of establishing inquiry notice—does not end the analysis. In the district court’s view, Phhhoto should have been on notice of Meta’s exclusionary scheme for an additional reason: the strong correlation between Meta’s adoption of the algorithmic feed and Phhhoto’s decline in user metrics. This conclusion overstates the inferences Phhhoto could have drawn from even a “precipitous downturn” in its user metrics. A-119, ¶ 104.

To be sure, as soon as this decline occurred in April 2016, Phhhoto knew it had been injured. Phhhoto may even have had reason to believe that Meta’s new algorithm was to blame. But knowledge of an injury—and the person or entity responsible for it—is distinct from knowledge that the injury was part of an anticompetitive or otherwise illegal scheme.

The district court did not recognize this distinction in determining that Phhhoto was on inquiry notice in April 2016. Instead, the district court faulted Phhhoto for failing to investigate “whether Meta’s new algorithm[] was injuring rather than benefitting Phhhoto” at that time. A-207. But even if Phhhoto had done so, and concluded that its business was indeed suffering due to the algorithmic feed, it could not have attributed this causal relationship to a probable

37

anticompetitive scheme on Meta’s part. Rather, Phhhoto could have identified only the likely cause of its metrics decline—or, in other words, discovered the fact of its injury and the entity responsible for it. From this information, Phhhoto neither could have “consider[ed] it ‘probable’” that the nexus between its failing business and the algorithmic feed was anticompetitive in nature, Koch, 699 F.3d at 151 (citations omitted), nor pled an antitrust claim “with sufficient detail and particularity to survive a 12(b)(6) motion to dismiss,” Charles Schwab Corp. v. Bank of Am. Corp., 883 F.3d 68, 95 (2d Cir. 2018) (internal quotation marks and citation omitted). To the contrary, Meta’s press release would have suggested more strongly to Phhhoto that the algorithmic feed—apparently tailored to users’ interests and in line with industry trends—was implemented for reasons of “efficiency and consumer satisfaction.” Trans Sport, Inc. v. Starter Sportswear, Inc., 964 F.2d 186, 188–89 (2d Cir. 1992) (citations omitted).

In concluding otherwise, the district court failed to follow our precedent requiring a “probability” of wrongdoing to trigger inquiry notice, see, e.g., Koch, 699 F.3d at 151 (citation omitted); Dodds, 12 F.3d at 350 (“[W]hen the circumstances would suggest to an investor of ordinary intelligence the probability that she has been defrauded, a duty of inquiry arises . . . .”), instead applying a standard more

38

akin to “reasonable suspicion,” see, e.g., A-208 (“Phhhoto’s reasonable suspicion regarding Meta’s algorithm should have been heightened . . . .”) (emphasis added); id. (“[T]he Court cannot reasonably infer from the Amended Complaint that until October 25, 2017 . . . Phhhoto had no reason to suspect that the . . . implementation of Meta’s new algorithm had likely affected its user engagement . . . .”) (emphasis added); A-205 (“[T]he Amended Complaint . . . includes alleged acts by Meta that should have aroused the suspicions of a reasonable business in Phhhoto’s situation . . . .”) (emphasis added). This was error. In the circumstances of this case, the distinction between finding Meta’s conduct suspicious—and considering it probable that Meta had engaged in wrongdoing—decides the matter.

We agree with the district court that the rapidity of Phhhoto’s business decline gave it reason to be suspicious of the new algorithm as the cause for the decline. But in April 2016, that suspicion did not amount to a probability that Meta’s algorithmic feed caused the decline through anticompetitive means. After all, anticompetitive and procompetitive conduct often produce similar effects.

Take Meta’s creation of Boomerang as an example. In the rapidly evolving technology industry, the release of Boomerang could have signified Meta’s effort

39

to develop a “superior product” that would boost competition in the market. Verizon Commc’ns Inc. v. Law Offs. of Curtis V. Trinko, LLP, 540 U.S. 398, 407 (2004) (citation omitted). Or, as Phhhoto now alleges, Meta’s creation of Boomerang could have been an exclusionary effort aimed at undermining Phhhoto’s business and monopolizing Meta’s power. Meta’s intent differs in each scenario, but the ultimate effect on Phhhoto—losing at least some users to a competing app—is the same. This example illustrates the difficulty of discerning whether a rival’s “development . . . of a superior product,” id. (citation omitted), is intended to boost competition in the market or to “caus[e] unreasonable exclusionary or anticompetitive effects,” Trans Sport, Inc., 964 F.2d at 188 (internal quotation marks and citation omitted).

Meta’s arguments ignore this feature of antitrust injuries. Instead of grappling with the ways in which Phhhoto could have distinguished between anticompetitive and procompetitive conduct, Meta urges us to conclude that Phhhoto’s business decline was sufficient to trigger inquiry notice in April 2016. In advancing this argument, Meta relies on our decision in SL-x IP S.à.r.l. v. Merrill Lynch, Pierce, Fenner & Smith, Inc. and the Fourth Circuit’s opinion in GO Computer, Inc. v. Microsoft Corporation; neither supports the conclusion Meta draws from it.

40

To start, Meta describes only part of our reasoning in SL-x IP S.à.r.l. v. Merrill Lynch, Pierce, Fenner & Smith, Inc., Nos. 21-2697, 21-2699, 2023 WL 2620041 (2d Cir. Mar. 24, 2023) (summary order). In that case, we affirmed the district court’s holding that the plaintiffs—whose claim was that the defendants “had colluded to engage in a group boycott of [the plaintiffs’ product], in violation of federal and state antitrust laws”—were on inquiry notice more than four years before their suit was filed. Id. at *2, *4.

This determination was predicated on: (1) statements that the defendants’

executives made directly to the plaintiffs’ representatives; and (2) “unusual” aspects of the defendants’ business dealings with the plaintiffs over the course of approximately one year. Id. at *4. Meta discusses only the latter on appeal, arguing that the defendants’ “dramatic[]” and “sudden[]” loss of interest in the plaintiffs’ product, id. at *4 n.3 (citation omitted), is akin to the “precipitous[]” decline in Phhhoto’s user metrics that occurred in April 2016, A-116, ¶ 91.

But this “strange reversal” in the defendants’ business relationship with the plaintiffs, SL-x, 2023 WL 2620041, at *4 n.3 (citation omitted), was not the only basis on which we affirmed the district court’s timeliness determination. Rather, included among the “markers that something was amiss,” id. at *4, were various

41

statements clearly referencing “a conspiracy to freeze [the plaintiffs’ product] out of the market by boycotting it,” SL-x IP S.à.r.l. v. Bank of Am. Corp., Nos. 18-cv- 10179, 19-cv-4885, 2021 WL 4523711, at *9 (S.D.N.Y. Sept. 30, 2021). For example, an executive of one broker defendant, Credit Suisse, told the plaintiffs at a February 2013 meeting that another defendant “was like ‘the mafia run by five crime families.’” Id. (citation omitted). The same executive also warned the plaintiffs, in July 2013, that Credit Suisse would maintain its interest in their product only “if ‘the Big Boys’ were committed to” it. Id. (citation omitted). We agreed with the district court that “[t]he warnings [contained] in these statements” would have prompted further inquiry “into the reasons why [d]efendants were rejecting” the plaintiffs’ product. Id. Thus, it was the parties’ deteriorating business relationship “together with” the defendants’ statements, implicating themselves in an anticompetitive scheme, that triggered inquiry notice. Id. (emphasis added).

Meta does not—and cannot—point to analogous statements in the present record. Far from alerting Phhhoto to the allegedly anticompetitive nature of the new algorithm, Meta cloaked it in innocuous terms, describing the algorithmic feed as based on neutral criteria such as user interest, user relationship with the

42

posting account, and timeliness of the post. Rather than communicating a “warning[]” that the new algorithm was anticompetitive, Meta’s 2016 press release allayed any possible “suspicio[n]” about its shift to an algorithmic feed. Id. In our independent review of the amended complaint, we have discerned no contrary statements, “made [by Meta] directly to [Phhhoto],” that “should have raised red flags” about the allegedly exclusionary nature of the new algorithm. Id.

Meta’s reliance on GO Computer, Inc. v. Microsoft Corp., 508 F.3d 170 (4th Cir.

2007), is also misplaced. There, the Fourth Circuit held that the plaintiffs’ antitrust claims were time-barred and not subject to equitable tolling, as the plaintiffs “were on inquiry notice of their claims [more than ten years before filing suit], when enough red flags had flown that a reasonably diligent person would have investigated and acted.” Id. at 172. Those “red flags” were threefold. First, when the Federal Trade Commission (“FTC”) was investigating Microsoft in 1991, an investigator told GO Computer’s co-founder, Jerry Kaplan, that “[t]his looks like a textbook case of abuse of monopoly power.” Id. at 178 (citation omitted). During his next meeting with an FTC investigator in 1992, Kaplan submitted a declaration describing “the array of obstacles Microsoft was allegedly

43

putting in GO’s way.” Id. Second, in a book he wrote in 1994, Kaplan “reported specific occasions, prior to his meeting with the FTC, where hardware manufacturers had told him about Microsoft’s restrictive licensing practices.” Id. Finally, Kaplan “became so suspicious” of Microsoft’s conduct that “he went to a law firm [in 1991] to discuss an intellectual property suit” against the company, which the law firm “thought was strong.” Id. In the Fourth Circuit’s view, there was “no question that this profusion of information was sufficient, as a matter of law, to spur a reasonably diligent person to investigate an antitrust claim.” Id.

Here, by contrast, there was no “multiplicity and specificity of . . . information,” id. at 179, that could have alerted Phhhoto to a probable anticompetitive scheme in April 2016. Despite its introduction of the algorithmic feed in March 2016, Meta was not publicly implicated in an “exclusionary scheme” until December 2018, when the U.K. Parliament released incriminating litigation documents. A-129, ¶ 133. It was not until December 9, 2020 that the FTC and various state attorneys general sued Meta for antitrust violations. See New York v. Meta Platforms, Inc., 66 F.4th 288, 295 (D.C. Cir. 2023) (“The States filed their Complaint on December 9, 2020, and the [FTC] filed a complaint on the same day.”).

44

In sum, while Phhhoto could suspect that Meta’s conduct was harming its business in April 2016, Phhhoto had no basis to suspect probable anticompetitive wrongdoing until October 25, 2017.

C. Phhhoto Has Adequately Pled Reasonable Diligence.

To satisfy the final element of the fraudulent concealment test, the plaintiff must establish that its “continuing ignorance” of the asserted claim “was not attributable to lack of diligence on [its] part.” Hendrickson, 840 F.2d at 1083. In this case, the district court held that “the Amended Complaint . . . directly undermine[s] Phhhoto’s argument that it acted with” reasonable diligence from April 2016 to October 2017, as Phhhoto failed to “investigat[e] its suspicion that Meta’s algorithm may have been suppressing third-party applications including Phhhoto.” A-211. We disagree.

As previously described, Phhhoto plausibly alleged that it did not have notice, during this period, that Meta’s algorithm was probably suppressing third- party applications. Moreover, even if Phhhoto suspected or should have suspected that Meta’s algorithm was at least in some way related to its decline, Phhhoto alleged that its founders “relied on Meta’s statement about the operation of the algorithm—that ‘the order of photos and videos in your feed will be based

45

on the likelihood you’ll be interested in the content, your relationship with the person posting and the timeliness of the post.’” A-116, ¶ 92. And while Phhhoto was certainly aware that its business was suffering, Phhhoto alleged that “[f]or months after the metrics dropped in April 2016, Phhhoto’s team worked tirelessly to figure out why Phhhoto was suddenly unpopular.” A-116, ¶ 92. Phhhoto’s founders “hypothesized various reasons the app may have suddenly dropped in popularity,” ruling out potential causes such as “competition from Boomerang and cyclical usage,” and “running analytics to determine if Phhhoto had bugs.” A-116–117, ¶¶ 92–93. Phhhoto’s reasonable diligence, in the circumstances alleged here, was adequately pled.

Meta’s argument to the contrary presumes that Phhhoto “would have discovered the alleged issue” with the algorithm as early as April 2016 if it had simply “looked at its own Instagram feed.” Appellee’s Br. at 30–31 (emphasis added). But Meta frames this hypothetical discovery in the singular, ignoring that the amended complaint alleges two “issue[s]” that were uncovered on October 25, 2017. Id. First, Phhhoto’s post “appeared to vanish from Bennett’s own Instagram feed.” A-120, ¶ 105 (emphasis omitted). Second, the video originating from Hypno, which had significantly fewer followers than Phhhoto,

46

received more views and “likes” relative to Phhhoto’s post. Even though the amended complaint pleads each of these discoveries—and Phhhoto emphasized both in its briefing before the district court—Meta’s argument that Phhhoto should have “simply . . . observ[ed] how its posts appeared in the Instagram app” is relevant only to the first one. Appellee’s Br. at 29.

We agree with Meta that Phhhoto presumably could have made the first discovery—in which Phhhoto’s post vanished from the Instagram feed—in the period between April 2016 and October 2017. To observe this phenomenon, Phhhoto only needed to access its own Instagram account, and the account of any individual, like Bennett, who followed Phhhoto. Phhhoto could have posted a photo or video to its page and then checked the placement of that post in the individual’s Instagram feed. But, even if Phhhoto had done so, the amended complaint provides no basis to conclude that Phhhoto should have discerned, from this observation alone, Meta’s probable involvement in an exclusionary scheme.

As alleged in the amended complaint, it was the second observation—

involving the metrics discrepancy between Phhhoto’s and Hypno’s posts— coupled with the first, that made exclusionary conduct the probable explanation

47

for Phhhoto’s decline after the adoption of the algorithmic feed. The metrics discrepancy directly undercut Meta’s prior description of the algorithm as “based on the likelihood [the user will] be interested in the content, [the user’s] relationship with the person posting and the timeliness of the post,” A-114–15, ¶ 86, suggesting instead that the algorithmic feed penalized posts from competitors’ accounts. That is, if the algorithm operated as Meta previously had represented, Phhhoto’s post likely would have outperformed Hypno’s. When the opposite occurred, one critical distinction between Phhhoto and Hypno became salient: the former competed with Meta, while the latter did not. Only at this point could Phhhoto attribute the algorithmic feed, and certain of Meta’s prior actions, to “anticompetitive behavior,” Affinity LLC v. GfK Mediamark Rsch. & Intel., LLC, 547 F. App’x 54, 57 (2d Cir. 2013) (summary order), rather than “efficiency and consumer satisfaction,” Trans Sport, Inc., 964 F.2d at 188–89 (citations omitted).

To be clear, Phhhoto alleges that its discovery of the metrics discrepancy was accidental. We express no view that the exercise of reasonable diligence invariably would require an antitrust plaintiff in Phhhoto’s position to undertake a comparison of the sort that occurred here by chance. But it is important to note that any comparative metrics analysis required a comparator Instagram account

48

to which Phhhoto could post content and against which it could measure the performance of its own posts. And, for the metrics comparison to be indicative of exclusionary conduct, the second account needed to belong to a company that did not compete with Meta, among other relevant criteria. The amended complaint suggests that access to such an account was not available during the “months” when Phhhoto’s team was “work[ing] tirelessly to figure out why Phhhoto was suddenly unpopular.” A-116, ¶ 92. It became available only after Phhhoto ceased operations and its co-founders rejoined Hypno. 10 Of course, Phhhoto’s allegations are not conclusive of its diligence during this period. As with the others in the amended complaint, the allegations concerning Phhhoto’s diligence can be disputed and potentially disproven in the discovery process. But, at the motion to dismiss stage, the district court reached

10 Contrary to the dissent’s suggestion, Phhhoto could not have “simply created a second account under a different username to compare metrics.” Dissent at 17. Rather, the second account had to share certain similarities with Phhhoto’s for it to provide an effective comparator. Meta had explained that its algorithm was “based on the likelihood [the user will] be interested in the content, [the user’s] relationship with the person posting [the content] and the timeliness of the post.” A-114–15, ¶ 86. To isolate and test for the impact of an additional, anticompetitive variable, Phhhoto thus would have needed a comparator account that could control for as many of these neutral factors as possible. It needed an account like Hypno’s—that of another business, operating in a similar industry, whose followers would likely be interested in similar content. Phhhoto’s co-founders could not simply have created such an account.

49

the wrong conclusion as to the diligence prong. Rather than reflecting “lack of diligence on [its] part,” Phhhoto’s “continuing ignorance” of its antitrust claim between April 2016 and October 2017 was at least plausibly attributable to the difficulty in discerning procompetitive from anticompetitive conduct when each results in business decline, as well as the subtle—and sometimes, practically undetectable—nature of algorithmic manipulation, Hendrickson Bros., 840 F.2d at 1083.

CONCLUSION

Having determined that Phhhoto adequately alleged equitable tolling based on fraudulent concealment, we hold that the district court erred by dismissing Phhhoto’s claim for unlawful monopolization under the Sherman Act as untimely at this stage. Accordingly, we VACATE the district court’s judgment and REMAND for further proceedings.

CHIN, Circuit Judge, dissenting:

Defendant-appellee Meta Platforms, Inc. ("Meta"), the social media company, operates Instagram, among other popular social networking applications ("apps"). According to plaintiff-appellant Phhhoto Inc. ("Phhhoto"), "[n]o other platforms rival Meta in the market for personal social networking services." J. App'x at 137 (Amended Complaint ("AC") ¶ 156).

Launched in 2014, Phhhoto was a social media app that allowed users to capture a burst of still photos and then loop them together into a video to create a moving image, known as a "phhhoto." At its peak, Phhhoto enjoyed "approximately 3.7 million [monthly average users]." Id. at 89 (AC ¶ 6). After just over three operative years, however, Phhhoto shut down operations on June 20, 2017, citing Meta's "fraudulent and anticompetitive conduct" as the reason. Id. at 127 (AC ¶ 128).

In this case, Phhhoto accuses Meta of violating section 2 of the Sherman Antitrust Act ("Section 2"), 15 U.S.C. § 2, based on Meta's allegedly anticompetitive behavior toward it. Its claims accrued no later than April of 2016. Yet, it did not file suit until November 4, 2021, well more than four years later. Phhhoto argues, however, that the district court erred when it granted

Meta's motion to dismiss on timeliness grounds, declining to equitably toll the statute of limitations under a theory of fraudulent concealment.

Applying a de novo standard of review, the majority concludes "that the district court erred at each step of the fraudulent concealment analysis." Maj. Op. at 17. While I agree that de novo review applies, I disagree that the district court erred. Phhhoto's allegations in the Amended Complaint demonstrate as a matter of law that it was on inquiry notice of Meta's anticompetitive conduct -- and thus had the requisite information to file suit -- well before October 25, 2017, the date on which Phhhoto claims it first began "to discover [Meta's] fraudulent and anticompetitive conduct." J. App'x at 50. Phhhoto's allegations further undermine its claims as to the other elements of its fraudulent concealment claim. I would affirm and, accordingly, I respectfully dissent.

I.

As Phhhoto alleges, "[n]ew entrants in the market for personal social networking services" face a challenge because they "must convince users that enough of their friends and family members will also engage with the social networking platform to make use of the platform worthwhile." Id. at 139 (AC ¶ 164). Throughout its lifespan as a social media app, Phhhoto relied on various

3

aspects of Meta's ubiquity in the social media universe to build and maintain its own user base. There was something in it for Meta, too, in this relationship: by offering Phhhoto access to key bits of its own infrastructure, Meta ensured that its users did not need to "exit the Meta ecosystem" even when those users explored other non-Meta platforms like Phhhoto. Id. at 140-41 (AC ¶ 169).

At the beginning of Phhhoto's relationship with Meta, the companies shared a "symbiotic relationship." Maj. Op. at 7. But over time, and through a distinct set of adverse acts alleged by Phhhoto, Meta dealt several blows to Phhhoto's stability as a platform, as it rescinded Phhhoto's privileges and access to critical components of Meta's social networking ecosystem.

The first integration feature that benefited Phhhoto was Meta's "Find Friends" Application Programming Interface ("API"). APIs are digital tools that facilitate data-sharing, functionality, and overall integration between two platforms. The Find Friends API allowed third-party apps like Phhhoto to access a user's Instagram friends list. When a user linked her Phhhoto account to her Instagram credentials, she could instantly access a list of Instagram friends who also had Phhhoto accounts. The Find Friends API thus allowed Phhhoto users to

"recreate their social graph" from Instagram, thereby alleviating what Phhhoto described as a barrier to entry for new platforms. J. App'x at 109 (AC ¶ 65).

The second feature was the "iPhone Hooks," which allowed users to post their phhhotos, created in the Phhhoto app, directly to Instagram. J. App'x at 163-64. Before a user publishes a post to Instagram, she has the option of adding a written caption that appears under the post. The caption can contain hashtags, which, when clicked, directs users to all other public posts whose caption features that hashtag. The iPhone Hooks pre-populated captions of phhhotos posted to Instagram with hashtags that attributed the posted content to Phhhoto. The benefit to Phhhoto was that, when a user posted her phhhotos directly to Instagram, the caption would contain a #phhhoto hashtag, which directed traffic to the app and attributed the content to Phhhoto.

On March 31, 2015, Meta abruptly revoked Phhhoto's access to the Find Friends API. Phhhoto alleges that Meta withdrew access "because [it] viewed Phhhoto as a potential competitive threat." Id. at 109 (AC ¶ 66). Phhhoto further alleged that the loss of the Find Friends API had a "negative[] impact [on] how potential investors perceived Phhhoto." Id. (AC ¶ 65). A few months later, on August 9, 2015, Meta also suspended the use of the iPhone Hooks for all third-

party apps, explaining that the pre-populated captions featuring the hashtags looked "spammy" on the Instagram feed. Id. at 111 (AC ¶ 71). Phhhoto alleges, however, that the "spammy" rationale "was pretextual." Id. (AC ¶ 72).

In addition to these integration features, Phhhoto and Meta nearly entered into a partnership. Earlier, on February 26, 2015, Meta's Strategic Partnerships Manager contacted one of Phhhoto's founders, Champ Bennett, to discuss integrating Phhhoto's moving photo technology into Facebook's newsfeed. The project was never memorialized in a contract, however, and after some delays related to "legal conversations," Meta apparently abandoned the integration project in June of 2015. J. App'x at 166-67.

On October 22, 2015, Phhhoto was set to launch for Android devices.

That morning, however, Meta issued its own announcement: it was launching Boomerang -- what Phhhoto describes as "a slavish clone of Phhhoto." Id. at 112 (AC ¶ 77).

On March 15, 2016, Meta issued a press release announcing a change to the order in which it sorted user's posts on Instagram. Until then, Instagram delivered content to users chronologically -- the most recent posts appeared at the top of the feed, and users scrolled down to see older content. Meta's switch

6

to the so-called "algorithmic newsfeed" discarded the chronological sorting in favor of an algorithm-driven shuffling that presented posts to users "based on the likelihood you'll be interested in the content, your relationship with the person posting and the timeliness of the post." Id. at 114-15 (AC ¶¶ 86-87).

Shortly thereafter, in April 2016, Phhhoto noticed that its "new user registrations declined precipitously." Id. at 116 (AC ¶ 91). Its ranking among photo and video platforms in "the Apple App Store dropped from 11th place to 41st place." Id. The change was abrupt and unprecedented; "Phhhoto had never before experienced such a significant decline in its ranking." Id. Prompted by this sudden decline in popularity, Phhhoto "worked tirelessly" to determine the reason behind the sharp decline. Id. (AC ¶ 92). Phhhoto claims it focused its efforts on problems with its own code, in reliance on Meta's press release statement: that "[t]he order of photos and videos in your feed will be based on the likelihood you'll be interested in the content, your relationship with the person posting and the timeliness of the post." Id. at 116-17 (AC ¶ 92).

On October 25, 2017, after Phhhoto shut down operations, Bennett "sought to connect Phhhoto's remaining Instagram followers to Hypno, a company that "had little social media presence." Id. at 119 (AC ¶ 104). Bennett

7

posted the same promotional video to Hypno's Instagram account and Phhhoto's Instagram account. Phhhoto claims that, after posting the same video from both accounts, "the post from the old Phhhoto account appeared to vanish from Bennett's own Instagram feed." Id. at 120 (AC ¶ 105) (emphasis in original). This, Phhhoto alleges, was the moment at which it discovered that Meta was engaging in anticompetitive conduct -- namely, through suppressing Phhhoto's content on the algorithmic feed.

II.

The district court did not err in concluding as a matter of law that the statute of limitations should not be subject to the extroardinary remedy of equitable tolling, as Phhhoto failed to sufficiently plead that it met the requisite elements of fraudulent concealment.

A. Elements of Fraudulent Concealment To succeed on a fraudulent concealment claim, a plaintiff must establish: "(1) that the defendant concealed from him the existence of his cause of action, (2) that he remained in ignorance of that cause of action until some point within four years of the commencement of his action, and (3) that his continuing

8

ignorance was not attributable to lack of diligence on his part." New York v. Hendrickson Bros., 840 F.2d 1065, 1083 (2d Cir. 1988).

B. Application

Phhhoto is not entitled to the judicial remedy of equitable tolling under a theory of fraudulent concealment because (1) it fails to plausibly allege that Meta's March 2016 press release was concealing; (2) Phhhoto's own allegations undermine its claim that it had no notice of a potential antitrust claim until October 25, 2017; and (3) Phhhoto fails to allege that it acted with reasonable diligence.

1. Concealment

To plead the first element of fraudulent concealment, Phhhoto must plausibly allege that Meta "concealed . . . the existence of [its] cause of action." Id. Phhhoto can meet this burden by showing either that Meta took affirmative steps to conceal its wrongdoing or that the wrongdoing was self-concealing. Id. But it must plead elements of fraudulent concealment with particularity, in accordance with the heightened pleading standards set forth in Rule 9(b) of the Federal Rules of Civil Procedure. See Armstrong v. McAlpin, 699 F.2d 79, 88 (2d Cir. 1983). Phhhoto contends that Meta engaged in fraudulent concealment when it issued a

9

press release in 2016 that announced a new algorithm for displaying posts. As the district court concluded, however, Phhhoto failed to plausibly allege that Meta's public announcement about the algorithm was fraudulent in any respect. Nor does it make sense, as Phhhoto asserts, that the change in algorithm was "self-concealing." Appellant's Br. at 33.

Fraudulent concealment requires proof of a "trick or contrivance intended to exclude suspicion and prevent inquiry." Wood v. Carpenter, 101 U.S. 135, 143 (1879). Phhhoto has not plausibly explained how the statements in Meta's 2016 press release constituted fraudulent concealment. This was a press release publicly announcing a new feed that would display posts based on metrics that measured user interest, relationship, and timeliness. Even considering the press release in the light most favorable to Phhhoto, Phhhoto fails to plausibly allege that there was anything in the press release that was false, fraudulent, or misleading. Phhhoto fails to explain how the announcement affirmatively concealed wrongdoing.

Apparently, Phhhoto contends that Meta concealed the "anticompetitive element in the algorithm's operation . . . by making false public statements representing how the algorithm worked." Appellant's Br. at 30.

Phhhoto contends, essentially, that Meta intended to harm Phhhoto and that the press release should have disclosed these facts. But this was a press release and Phhhoto -- and the public -- was being given notice of the change in the display procedures. There is nothing fraudulent about that notice. And Meta made no representation that Phhhoto would do better under the new algorithm.

Instead, Phhhoto itself acknowledges that "its own content might be promoted or demoted by Meta's algorithm." Phhhoto Inc. v. Meta Platforms, Inc., No. 21-cv-06159 (KAM) (LB), 2023 WL 2710177, at *16 (E.D.N.Y. Mar. 30, 2023). But the suggestion that Meta should have disclosed to Phhhoto and others that its new procedures might give it a competitive advantage makes no sense.

Moreover, Phhhoto fails to explain "why Meta's algorithm, 'if actually implemented as Meta had described,' would have optimized 'Phhhoto users' posts, rather than disfavored those posts." Phhhoto, 2023 WL 2710177, at *16 (citing J. App'x at 115-16 (AC ¶ 89)). As Meta points out, "Phhhoto does not allege that Meta said that Phhhoto's content was not being disfavored by the new algorithm." Appellee's Br. at 26. In my view, it is not enough that Phhhoto alleges that it was enjoying millions of users on a monthly basis -- it does not follow that, just because an app has many users, its posts would perform just as

11

well or better with Instagram's new algorithm. On this record, I disagree that Phhhoto has adequately pled fraudulent concealment.

2. Notice

Phhhoto fails to allege sufficient facts supporting the second prong --

that Phhhoto remained in ignorance of its cause of action until October 2017. Indeed, Phhhoto's own allegations demonstrate that it had notice of its claim by no later than April 2016. Once the "plaintiff ha[s] notice of th[e] possibility" of its claim, the "plaintiff is charged with whatever knowledge an inquiry would have revealed." Stone v. Williams, 970 F.2d 1043, 1049 (2d Cir. 1992). I agree with the majority that, to find inquiry notice as a matter of law on a motion to dismiss, there must be "uncontroverted evidence clearly demonstrat[ing] when the plaintiff should have discovered the [challenged] conduct." Staehr v. Hartford Fin. Servs. Grp., 547 F.3d 406, 427 (2d Cir. 2008). Unlike the majority, however, I think there are such facts alleged in Phhhoto's Amended Complaint itself.

For instance, Phhhoto alleged a "precipitous" "decline" in users, a "never before experienced" reduction in its App Store rankings, and "sudden[] unpopular[ity]" shortly after the algorithmic feed was rolled out. J. App'x at 116-

17 (AC ¶¶ 90-92). This steep decline in business occurred in the context of a litany of Meta's other adverse actions, including:

• March 31, 2015: Meta abruptly revokes Phhhoto's access to Instagram's Find Friends API;

• June 8, 2015: Meta stops responding to Phhhoto about the integration project;

• August 9, 2015: Meta withdraws access to the iPhone Hooks for third-

party applications;

• October 22, 2015: Meta announces its new app Boomerang, a clone of Phhhoto;

• March 15, 2016: Meta announces change to the Instagram algorithm;

and • April 2016: Phhhoto notices steep drop in new user registrations.

These facts surely put Phhhoto on inquiry notice that Meta might be taking

action against it.

Phhhoto states in a conclusory fashion that it relied on the representations in the March 2016 press release to rule out anticompetitive behavior stemming from the algorithmic feed. I agree with the district court that Phhhoto presents "[n]o plausible facts . . . [to] explain why the Phhhoto founders would rule out [the aforementioned] external possibilities and ignore its duty to inquire." Phhhoto, 2023 WL 2710177, at *18.

The majority suggests that Meta's creation of Boomerang, in the context of the parties' relationship, did not contribute to placing Phhhoto on

13

notice of anticompetitive behavior. See Maj. Opp. at 39. As the majority recognizes, however, the law requires only a "'probability' of wrongdoing to trigger inquiry notice." Id. at 38. Phhhoto's business decline occurred only a few months after Meta rolled out a product nearly identical to Phhhoto's, and just one month after Meta changed its algorithm. Particularly because Phhhoto's business success was reliant on the underlying platform created by Meta, Meta's creation of Boomerang was another incident among many that reasonably placed Phhhoto on notice that Meta was engaging in anticompetitive behavior.

Phhhoto's other allegations about Meta's pattern of anticompetitive behavior further undermine its claim that it had no reason to believe Meta was engaged in anticompetitive conduct until Bennett's happenstance discovery in October 2017. Its Amended Complaint, which spans almost 70 pages, explains in detail the sudden and inexplicable actions Meta took -- rescinding access to social networking infrastructure that Phhhoto relied on, introducing a competitor on the day that Phhhoto planned on rolling out its Android capabilities, and letting a business relationship go cold over seemingly nothing. And Phhhoto alleges that, from its inception, it was reliant on critical support from Meta through APIs and other tools. Based on these factual allegations, I agree with the district court

14

that it is implausible that Phhhoto had no reason to suspect wrongdoing until October 2017.

The Fourth Circuit's reasoning is persuasive on this point: "[w]here a plaintiff knows of a pattern of particular actions that a defendant has taken against him," that plaintiff "is on inquiry notice of his claim" even if "the pattern's precise scope might be unclear and its exact legal ramifications uncertain." GO Computer, Inc. v. Microsoft Corp., 508 F.3d 170, 179 (4th Cir. 2007); see also Klein v. Bower, 421 F.2d 338, 343 (2d Cir. 1970) ("[T]he statutory period . . . [must] not await [plaintiff's] leisurely discovery of the full details of the alleged scheme.").

The majority reasons that Meta "offered plausible, non-exclusionary justifications" for some of its changes, and therefore Phhhoto cannot be charged with notice. Maj. Op. at 32. For example, the majority notes that Meta explained its decision to withdraw the iPhone Hooks capability by stating that the pre- populated hashtags felt "spammy." Id. (quoting J. App'x at 111 (AC ¶ 71)). But Phhhoto specifically alleged that Meta's "professed . . . rationale" for withdrawing the iPhone Hooks access was "pretextual." J. App'x at 111 (AC ¶ 72).

And perhaps most crucially, Phhhoto's own allegations show that the algorithm was not self-concealing. Phhhoto fails to support its contention that it could not discover or was not on notice of the algorithmic suppression despite noticing such a dramatic drop in user engagement shortly after the algorithm changed.

3. Reasonable Diligence Phhhoto argues that it plausibly alleged that it acted with reasonable diligence, which "is a prerequisite to the applicability of equitable tolling." Koch v. Christie's Int'l PLC, 699 F.3d 141, 157 (2d Cir. 2012). Again, Phhhoto's own allegations belie this point. As noted above, Meta abruptly revoked Phhhoto’s access to the API; Meta stopped responding to Phhhoto about a possible joint project; Meta withdrew access to the iPhone Hooks; Meta launched Boomerang -- a purported clone of Phhhoto's product; Meta announced the change to its algorithm; and Phhhoto's new registrations dropped precipitously. All of these adverse actions occurred within a span of thirteen months; Phhhoto surely should have investigated.

Phhhoto "hypothesize[s] various reasons the app may have suddenly dropped in popularity," but fails to allege why it did not investigate the

16

algorithm. J. App'x at 117 (AC ¶ 93). Given Phhhoto's heavy reliance on Meta's infrastructure to conduct its own platform, Phhhoto has failed to plausibly allege that it acted with the requisite diligence.

The majority agrees that "Phhhoto presumably could have made the first discovery -- in which Phhhoto's post vanished from the Instagram feed -- in the period between April 2016 and October 2017." Maj. Opp. at 46. The majority then reasons that had Phhhoto performed the experiment earlier, it would still not have been put on notice of Meta's anticompetitive behavior. I disagree. This discovery, along with the many prior anticompetitive incidents by Meta, in my view, should have led Phhhoto to discern Meta's probable involvement in an exclusionary scheme. By not performing the simple experiment of observing a Phhhoto post from a user account earlier, Phhhoto failed to exercise reasonable diligence.

As to Phhhoto's second "accidental" observation involving the metrics discrepancy, I disagree that this observation was the necessary catalyst for Phhhoto to be placed on notice of Meta's anticompetitive behavior, and that Phhhoto did not need to perform this experiment purposefully. First, as mentioned above, Meta's earlier conduct was sufficient to place Phhhoto on

17

notice of its claim. Second, although the majority "express[es] no view that the exercise of reasonable diligence invariably would require an antitrust plaintiff in Phhhoto's position to undertake a comparison of the sort that occurred here by chance," this reasoning undercuts the high standard equitable tolling requires and disregards the effortlessness of performing such a simple experiment. Id. at 47. Phhhoto did not need to create another company, or wait to create another company, to accomplish what it purportedly found by accident; Phhhoto could have simply created a second account under a different username to compare metrics. Bennett could have -- and should have -- posted a video to Phhhoto's Instagram account and a different Instagram account long before October 25, 2017. Performing this experiment strikes me as the bare minimum exercise of reasonable diligence, especially in consideration of the context of the parties' relationship.

Finally, I respectfully disagree with the majority's statement that "the allegations concerning Phhhoto's diligence can be disputed and potentially disproven in the discovery process." Id. at 48. There is no factual dispute here. Whether Phhhoto's alleged facts met the diligence prong is a legal inquiry, and I fail to see how further discovery could suddenly demonstrate that Phhhoto acted

18

with diligence. And certainly not under the high standard that requires Phhhoto to have pled particular facts demonstrating how it sufficiently met the diligence prong to warrant the district court's tolling of the statute of limitations. Accordingly, Phhhoto has not plausibly alleged that it acted with due diligence, or that discovery could lead to a different conclusion.

III.

The district court's judgment should be affirmed because it correctly dismissed Phhhoto's Amended Complaint as time-barred after concluding that the extraordinary measure of equitable tolling was unwarranted as a matter of law. This was the right result. I therefore dissent from the majority's decision to vacate and remand.