ICLE Amicus to the DC Circuit in FTC v Meta
INTRODUCTION AND SUMMARY OF ARGUMENT
The Federal Trade Commission (“FTC” or “Commission”) and its supporting amici describe an opinion the district court did not write. They say the court misused the Hypothetical Monopolist Test (“HMT”) by measuring substitution from conditions already degraded by monopoly. But that objection assumes the premise that the Commission bore the burden of proving: that Meta already possessed monopoly power and had reduced the quality of its apps. The court instead asked whether a hypothetical monopolist could make Meta’s apps significantly worse “than they would be in a competitive market,” Op. 38, and found that the Commission had not established either the competitive benchmark or monopoly power.
The Commission’s purported direct evidence did not establish monopoly power either. Increased advertising load does not by itself connote a supracompetitive quality-adjusted price because advertisements differ in relevance and burden, and product quality includes all the apps’ attributes, not just ad quality. The Commission supplied neither a competitive ad-load benchmark nor an analysis of the product as a whole.
Nor do Meta’s enterprise-wide returns establish monopoly power in the alleged user-side personal social networking (“PSN”) market without evidence attributing those returns to that market and distinguishing monopoly rents from returns to superior products, efficiency, or successful risk-taking. The Commission supplied neither analysis.
The Cellophane fallacy does not rescue the Commission’s case. It warns that switching observed at a monopoly price may overstate the extent of substitution; it does not establish that prevailing terms are monopolistic, shift the burden of proof, or explain why users would rank the available alternatives differently.
Nor does the Cellophane fallacy address the question the court actually considered: which products consumers considered substitutes for Meta’s. Much of the court’s evidence involved a product’s complete removal, which the court used not as an HMT but to identify where users went when an app became unavailable. A monopoly baseline might affect the amount of switching or the characteristics of the users observed, but it does not alter the ranking of alternative products on which the district court relied. The Commission identified no mechanism or evidence that would alter the consistent ranking of TikTok and YouTube as Meta’s closest alternatives.
The Commission also cannot convert limitations in individual pieces of evidence into affirmative proof of its proposed market. It offered no measurement of a small quality reduction, diversion analysis, or study of margins that quantified the strength of the alleged constraints. Meta, by contrast, offered a payment experiment, an eighteen-week user panel, outages, TikTok’s permanent removal from India and temporary shutdown in the United States, ordinary adoption data, and evidence of defensive investment. Those sources have different limitations, but they consistently identified TikTok and YouTube as the closest alternatives.
The evidence thus repeatedly showed users substituting across products that the Commission claims were not “for the same purposes.” FTC Br. 52. But the HMT asks whether enough users would divert to make degradation unprofitable; it does not require the alternatives to be identical.
The district court rigorously evaluated what the evidence permitted. It assessed varied and imperfect evidence relevant to questions of actual and likely substitution patterns among social media platforms and held that the Commission had not met its prima facie burden of proving the willful maintenance of monopoly power in a relevant market.
The Commission and its amici, by contrast, never supply a competitive counterfactual showing that Meta could worsen quality to users or raise prices to advertisers without losing enough demand on either side of the platform to make the strategy unprofitable. They instead assume that prevailing conditions reflect monopoly, use that assumption to justify discounting substitution evidence inconsistent with their proposed market, and then treat the resulting absence of evidence as support for that market. That is not an empirical demonstration of monopoly power, much less of its unlawful exploitation.
ARGUMENT
I. The district court correctly required the Commission to prove that Meta could profitably make its apps significantly worse than they would be under competition
The district court applied the same economically relevant question to the Commission’s market-definition theory and its purported direct evidence: whether Meta could profitably make its apps significantly worse than they would be under competition. The Commission bore the burden of proving monopoly power. It failed to do so.
The HMT ordinarily asks whether a hypothetical monopolist could profitably impose a meaningful price increase. See, e.g., U.S. Dep’t of Just. and Fed. Trade Comm’n, Merger Guidelines § 4.3.A (2023); U.S. Dep’t of Just. and Fed. Trade Comm’n, Horizontal Merger Guidelines § 4.1.1 (2010). Because Facebook and Instagram “have always charged users the same amount: nothing,” Op. 17, the equivalent inquiry concerns quality, and making a zero-price product materially worse is deemed to raise its quality-adjusted price. See Merger Guidelines §§ 4.3.A–B, at 41–42 (2023).
That was the district court’s approach—the same approach the FTC adopted at trial. It asked whether a hypothetical monopolist could profit by making Meta’s apps significantly worse “than they would be in a competitive market (say, by bloating them with ads).” Op. 38 (emphasis added).
That question requires a competitive benchmark, not the prevailing price. Yet the Commission says the court “assumed its conclusion” by treating “current market conditions” as competitive, FTC Br. 4, 17–18, while the economist amici say it measured substitution from “a post-conduct, supra-competitive equilibrium,” Econ. Br. 15. The opinion says otherwise: The court expressly used a competitive market as its benchmark. Disagreement with the court’s answer does not establish that it asked the wrong question.
The Commission’s two asserted alternatives do not adequately provide the missing proof. Increased ad load measures only one aspect of product quality, and an enterprise-wide return without attribution says nothing about power in the alleged user-side market.
A. Increased advertising load alone does not establish a supracompetitive quality-adjusted price
The Commission’s asserted proof that Meta’s pricing exceeded the competitive baseline is that advertising load increased and that more advertisements mean a higher quality-adjusted price. FTC Br. 66–67. Consumers generally prefer fewer ads, all else equal. But ad count does not measure advertising burden: A relevant, unobtrusive advertisement imposes a different cost than a disruptive, irrelevant one. Op. 27–28. In fact, as the FTC itself has observed, it may not impose a cost at all. See Yan Lau, A Brief Primer on the Economics of Targeted Advertising 5–6 (Fed. Trade Comm’n Bureau of Econ. 2020) (“Consumers receiving targeted ads will on average find them more ‘relevant’ compared to untargeted ones. . . . Thus, targeting benefits the consumer because it effectively reduces their search costs.”).
The FTC’s inference also assumes that every other product attribute remained constant. The record shows otherwise: Meta added multiple features to Facebook (and to Instagram), ads became more relevant, and users acted on them at an increasing rate. Meta’s internal planning rule, moreover, permitted additional advertisements only as ad quality improved. Op. 27–29.
Those facts do not prove that users preferred more ads, of course. But they do show that advertising impressions were not homogeneous across time and that count alone could not measure net advertising burden, much less overall product quality. Thus, even assuming arguendo that the earlier quality-adjusted price was competitive, the district court found that Meta held that price constant by improving ad quality and product features while increasing ad load. Id. at 29.
Economic logic supports the court’s treatment. On an advertising-supported two-sided platform, advertisers purchase access to users’ attention, whose value depends on user engagement rather than simply whether they see an advertisement. Engagement, in turn, depends on ad quality and relevance: “The better the ad, the more consumers will interact with it.” Id. at 18. See David S. Evans, Attention Rivalry Among Online Platforms, 9 J. Comp. L. & Econ. 313, 313–14, 316 (2013). More ads increase inventory but may reduce the time and attention the platform can sell; the profit-maximizing ad load balances those effects. Op. 17–19, 28–29, 40; see also Attila Ambrus, Emilio Calvano & Markus Reisinger, Either or Both Competition: A “Two-Sided” Theory of Advertising with Overlapping Viewerships, 8 Am. Econ. J.: Microecon. 189 (2016). That cross-side discipline is consistent with the economic logic of the Supreme Court’s Amex decision. See Ohio v. Am. Express Co., 585 U.S. 529, 545–47 (2018): Competition cannot be assessed accurately by looking at one side in isolation.
An increase in Meta’s ad load therefore shows only that Meta displays more ads than it once did, not that it displays more than competition would permit or charges a higher quality-adjusted price. Establishing that required evidence of the ad load a competitive platform would choose. The Commission offered none. FTC Br. 66–68. Its assertion that independent firms “may have” improved ad quality “without spiking their ad load,” id. at 67–68, does not establish those firms’ profit-maximizing ad loads or whether their products were comparable to Meta’s. The observed increase therefore neither directly proves monopoly power nor supports the Commission’s Cellophane argument. Both theories assume, without demonstrating, that Meta degraded its product below the competitive level.
B. The Commission’s enterprise-wide rate-of-return evidence does not establish monopoly power in the alleged user-side market
The Commission’s expert estimated a 36%–41.4% enterprise-wide internal rate of return (“IRR”) against a 9.8% weighted average cost of capital; the Commission describes the result as “economic profits nearly four times its cost of capital.” ; FTC Br. 64. The district court acknowledged that persistent profits may suggest monopoly power. Op. 24. But the question is whether that enterprise-wide return establishes monopoly power in the alleged PSN market. The Commission calls Meta’s profits a “strong indicator” and argues that those profits, together with its quality-degradation evidence, are “alone sufficient” to establish monopoly. FTC Br. 18, 63–68. It dismisses Meta’s technology, investments, and management as possible sources of the return, considering them, “if at all,” only under Grinnell’s second element. Id. at 64–65. That approach misreads Grinnell.
Grinnell distinguishes possession of monopoly power from its willful acquisition or maintenance by means other than superior products, business acumen, or historic accident. United States v. Grinnell Corp., 384 U.S. 563, 570–71 (1966). Evidence of superior products or business acumen bears on the second element once monopoly power has been shown. See Verizon Commc’ns Inc. v. Law Offices of Curtis v. Trinko, LLP, 540 U.S. 398, 407 (2004); United States v. Microsoft Corp., 253 F.3d 34, 50–51 (D.C. Cir. 2001) (en banc) (per curiam). It can also bear on the first element when profits are offered as evidence from which the plaintiff asks the court to infer monopoly power.
Those uses address different propositions, however. For Grinnell’s second element, the question is whether a firm with monopoly power acquired or maintained it through exclusionary conduct rather than competition on the merits. For the first, the question is what the observed returns establish about the firm’s power over price or quality. Courts accordingly distinguish profits attributable to monopoly power from returns attributable to efficiency, management, or quality, and the district court properly added successful risk-taking to the list. See Blue Cross & Blue Shield United of Wis. v. Marshfield Clinic, 65 F.3d 1406, 1412 (7th Cir. 1995); Op. 25. As Judge Posner cautioned in Marshfield, however, “it is always treacherous to try to infer monopoly power from a high rate of return. . . [T]here is not even a good economic theory that associates monopoly power with a high rate of return.” 65 F.3d at 1412.
The district court expressly “decide[d] this case on the first element alone” and considered Meta’s technology, investments, and management only in assessing the direct evidence of monopoly power. Op. 22, 24–26. It neither excused a proven monopoly as the Commission suggests, FTC Br. 65, nor shifted the burden of persuasion to Meta. Microsoft’s burden-shifting framework begins with the distinct question whether conduct by a firm already shown to possess monopoly power is exclusionary. 253 F.3d at 58–59. Here, the alternative explanations were reasons the enterprise-wide estimate did not establish power in the first place, not procompetitive justifications for proven power. The Commission retained the burden of persuasion, and its experts’ failure to assess the record-supported alternatives left no basis for attributing Meta’s enterprise-wide return to monopoly power rather than those alternatives. Op. 25–26.
The Commission’s purported direct proof of monopoly also failed to satisfy Microsoft’s requirement of evidence that a firm “can profitably raise prices substantially above the competitive level.” 253 F.3d at 51 (emphasis added). An IRR above the weighted average cost of capital may indicate positive economic returns under certain assumptions, but it does not identify the competitive price or quality in the alleged market, attribute the returns to pricing power rather than superior performance, or translate profitability into the degree of pricing power monopolization requires. If the existence of profits established monopoly, ordinary returns to differentiated products would make virtually the entire competitive economy presumptively monopolistic. Cf. U.S. Football League v. Nat’l Football League, 842 F.2d 1335, 1362 (2d Cir. 1988) (warning against evidentiary inferences that would let “profitability alone provide a basis for antitrust liability”).
Drawing the inference the Commission seeks requires accounting for product differentiation, returns to valuable intangible assets, innovation risk, and compensation for failed investments. See, e.g., Michael Cragg, Patrick Holder, David Hutchings & Bin Zhou, The Proper Measure of Profits for Assessing Market Power, 37 Antitrust 49, 50–52 (2023). It also requires connecting any residual return to power in the alleged market. The Commission did neither. Op. 25–26.
An analysis comparing cash flows with the cost of capital must also identify the investments producing the returns and assess appropriate comparators. The Supreme Court has declined to infer monopoly from liberal profits “without proof of lack of comparable profits during those years in other prosperous industries.” United States v. E.I. du Pont de Nemours & Co., 351 U.S. 377, 404 (1956). The Commission offered no such comparison.
The Commission responds that it “did not need” an industry comparison because competitive markets yield no sustained economic profit. FTC Br. 65–66. But the zero-profit result is not a general implication of competition. It follows from a stationary, long-run equilibrium with homogeneous products, free entry, and freely reproducible technologies. In their absence (as here), measured positive profits may instead represent returns to scarce productive inputs. See, e.g., Louis Makowski & Joseph M. Ostroy, Perfect Competition and the Creativity of the Market, 39 J. Econ. Lit. 479, 483–84 (2001). Persistent returns may suggest some market power, but without distinguishing returns to superior capabilities from monopoly rents and attributing the residual to power in the alleged PSN market, they do not establish monopoly power.
That attribution problem is especially acute for two-sided technology platforms. See, e.g., David S. Evans & Richard Schmalensee, The Antitrust Analysis of Multi-Sided Platform Businesses, in 1 The Oxford Handbook of Int’l Antitrust Econ. 404, 420 (Roger D. Blair & D. Daniel Sokol eds., 2015) (“These linkages across the multiple groups of customers and the products and services being offered to each group have to be accounted for in the analysis of the relevant antitrust market and the assessment of market power.”). The alleged market concerns services supplied to users at a nominal price of zero, while Meta’s returns come from advertisers. The district court found that Meta’s technology made advertisements more relevant and effective and that the “vast majority” of projected revenue growth would come from improving relevance. Op. 25. The returns may therefore reflect Meta’s appeal to advertisers rather than “a tight grip on the users who pay it nothing.” Id. An enterprise-wide return that does not identify its source cannot establish power in the alleged user-side market. See Bailey v. Allgas, Inc., 284 F.3d 1237, 1252 n.21, 1255 (11th Cir. 2002) (explaining that monopoly-profit analysis concerns power over “some particular product” and rejecting company-wide returns not attributable to the alleged market).
At most, the profits evidence warranted the “high alert” the Commission invokes. FTC Br. 63, 68. But a direction to remain alert to a possibility is not proof of a fact, and the profits evidence supplied neither a competitive quality-adjusted-price benchmark nor the requisite market-specific attribution.
II. The Cellophane fallacy neither shifts the Commission’s burden nor undermines the court’s substitution evidence
The Commission and the economist amici contend that the district court committed the Cellophane fallacy by crediting switching under conditions already degraded by monopoly. They thus argue that its market definition was flawed. FTC Br. 58–62; Econ. Br. 9–11, 15–17.
But the Cellophane fallacy does not undermine the court’s analysis. At issue was the Commission’s alleged market definition, which stood or fell on its claim that Snapchat competes with Facebook and Instagram while TikTok and YouTube do not. FTC Br. 40–42. It was, in other words, a question of the market’s composition, not the degree to which the market was or was not competitive. The fallacy warns that switching at a monopoly price may overstate the extent of substitution observed at a monopoly price. It does not explain why, when a product disappears, users would rank the remaining alternatives differently.
A. The Cellophane fallacy does not affirmatively establish a monopoly baseline or shift the burden of proof
The Cellophane fallacy is an error of inference from observed switching at the seller’s prevailing price. Customers’ willingness to abandon a product after a further price increase does not necessarily establish that the seller lacks power, because “at a high enough price, even poor substitutes look good to the consumer.” United States v. Eastman Kodak Co., 63 F.3d 95, 105 (2d Cir. 1995); see also William M. Landes & Richard A. Posner, Market Power in Antitrust Cases, 94 Harv. L. Rev. 937, 960–61 (1981) (There will always be “some substitution of other products for [a monopolist’s] own when it is maximizing profits, even if it has great market power.”).
The fallacy therefore means that consumer readiness to switch products at prevailing terms cannot disprove monopoly power. It does not mean that such evidence proves the existence of monopoly. Nor does it relieve the plaintiff of proving that prevailing terms exceed competitive levels. Significant substitution at current or higher prices “does not tell us whether the defendant already exercises significant market power.” Eastman Kodak Co. v. Image Technical Services, Inc., 504 U.S. 451, 471 (1992) (quoting Phillip Areeda & Louis Kaplow, Antitrust Analysis ¶ 340(b) (4th ed. 1988)) (emphases added and omitted). Evidence that does not answer the question is inconclusive, not probative, and the plaintiff bears the burden of proving monopoly power. Grinnell, 384 U.S. at 570–71; Microsoft, 253 F.3d at 51.
Nor can the mere possibility that prevailing conditions reflect monopoly leave the Commission’s narrow market standing by default. Indeed, if prevailing terms are actually competitive, observed switching may understate the substitution a true monopolist would face and may yield a market that is too narrow. See Luke M. Froeb & Gregory J. Werden, The Reverse Cellophane Fallacy in Market Delineation, 7 Rev. Indus. Org. 241, 241 (1992). Which risk exists depends on the baseline—an empirical fact the Commission bore the burden of proving.
B. The Cellophane fallacy’s marginal mechanism does not invalidate the removal evidence
To assess which applications compete with Meta’s, the court considered two kinds of substitution evidence. A modest amount was what economists call marginal evidence, which reflects how user behavior changes when there is a small change in the effective price of the target product. Professor List’s payment experiment was evidence of this sort. Op. 44–47. But most of the evidence—and the evidence the Commission most vigorously contests—involved the complete removal of a product rather than a small change in its price. The 2021 Meta outage, TikTok permanent removal from India and temporary shutdown in the United States, and the 2018 YouTube outage were examples of this “removal” evidence. Id. at 47–52.
The Commission and its amici correctly observe that removal is not a small price increase but is analogous to a “huge” or “infinite” one. FTC Br. 56 n.9; Econ. Br. 19; AAI Br. 16. The district court said the same: bans and outages “do not impose small but significant price increases; they get rid of a product entirely.” Op. 55. But that point does not make the Cellophane-fallacy problem stronger; it makes its marginal mechanism inapplicable to the removal evidence.
Assessing substitution following a product’s removal does not measure the response to a marginal quality reduction and therefore does not itself constitute an HMT—and the court did not use removal evidence for that purpose. It nevertheless captures economically relevant substitution: When Facebook is unavailable, the applications to which users redirect their attention reveal their next-best choices among the remaining alternatives. Complete removal measures those choices across all Facebook users, not just the marginal ones. See Christopher Conlon & Julie Holland Mortimer, Empirical Properties of Diversion Ratios, 52 RAND J. Econ. 693, 699, 701–02 (2021). The district court used the outage evidence in precisely that limited way: as non-HMT evidence informing its qualitative appraisal. Op. 55–56.
The district court therefore properly observed that the fallacy “would not affect the order of substitution.” Id. at 58 (emphasis added). The Commission calls that “simply wrong,” FTC Br. 61, but it identifies no mechanism by which, or evidence showing that, a supracompetitive advertising load on Facebook would cause users deprived of Facebook to choose TikTok and YouTube over Snapchat and MeWe. Monopoly pricing may affect the number or mix of users who switch products; it cannot reorder their choice among the alternatives.
Every source of evidence in the record produced the same ranking. When Facebook and Instagram became costly or unavailable, users turned first to TikTok and YouTube; when TikTok or YouTube disappeared, users turned to Facebook and Instagram. Snapchat and MeWe trailed in every measurement. The court called this “a consistent and unmistakable story,” Op. 52. That conclusion was a factual finding reviewed for clear error, and no brief on the other side even attempts to identify record evidence to the contrary.
C. The Commission identifies no baseline effect that would change the observed ranking
A supracompetitive advertising load could affect who remains in the observed user base and the weight assigned to different users’ choices. That possibility bears on the weighting of the removal evidence for assessing the extent of monopoly power; it does not eliminate its relevance. The Commission still had to show that the omitted users would rank Snapchat or MeWe above TikTok or YouTube, or that including them would reverse the observed user behavior. It identified no such evidence.
In fact, there is reason to think that the substitution evidence would be even more persuasive, assuming monopoly pricing. On the Commission’s theory, excessive advertising had already driven away the users most burdened by ads. So the removal evidence overrepresented the friends-and-family core around which it built its market definition. Yet even that group turned first to TikTok and YouTube, while Snapchat captured less than a third of TikTok’s gain. Op. 47. Correcting for the alleged monopoly baseline would add back users more sensitive to advertising, and the record provides no reason to think that their choices would reverse the preference for TikTok and YouTube.
The alternatives themselves might also look different in the but-for, competitive world. If the market had remained competitive, some rival might have been better funded, higher quality, and more attractive than it is today. That is theoretically possible. And the Commission’s own counterfactual identifies one specific difference: Instagram operating independently of Meta and competing against Facebook. FTC Br. 55.
But Instagram is already in the evidence, maintained as a distinct product (albeit owned by Meta). And in the payment experiment, users paid to reduce time on Facebook shifted toward Instagram more than any other product, and next most frequently to TikTok and YouTube. Op. 46. Neither the Commission nor its amici have identified any mechanism by which an independent Instagram (as opposed to one owned—and improved—by Facebook) would cause Snapchat to leapfrog TikTok and YouTube in users’ rankings.
As the one piece of evidence measuring genuinely marginal changes, the payment experiment is most directly exposed to the Cellophane fallacy. But the court used it to identify users’ alternative destinations, not as a direct application of the HMT, id. at 46, 55–56, and the Commission offers no evidence that a different baseline would change their ranking. The Cellophane fallacy is inapposite.
D. The district court confronted the Cellophane fallacy, and its reasoning withstands the Commission’s response
The district court devoted a section of its opinion to the Cellophane argument. Id. at 58–59. The economists’ brief never cites those pages. The Commission does respond, FTC Br. 61–62, but does not refute the court’s reasoning.
First, the court explained that the fallacy “would not affect the order of substitution, which shows that the closest substitutes for Meta’s apps are YouTube and TikTok.” Op. 58. As discussed above, the Commission’s objections do not undermine this conclusion.
Second, two events measured substitution toward Facebook when TikTok became unavailable. As the court noted, if Meta’s apps carried a supracompetitive quality-adjusted price, it “would bias [substitution] rates down,” id. at 58, and understate how closely the products compete. On the Commission’s premise, those events are biased against Meta, yet they support Meta’s market definition anyway.
Third, the record did not show an advertising distortion large enough to change the ranking. Removing advertisements in their entirety correlated with a 7% increase in users’ time on Facebook, while an 80% reduction in advertising load for teenagers was projected to increase usage by 3%. Id. at 17–18. Meanwhile, “Facebook users who downloaded TikTok used Meta’s app 17–26% less.” Id. at 59. As the court noted, “[i]t is hard to believe that nudging down the ad load to whatever the FTC considers the competitive level would make substitution rates to TikTok unimportant.” Id.
The economists answer that the 7% increase is consistent with a profit-maximizing monopoly price. Econ. Br. 4, 16. But profit maximization is not diagnostic of monopoly: Every profit-maximizing firm forgoes unprofitable price cuts, whether operating in a competitive market or a monopolized one. The observation is therefore equally consistent with competition and does not disturb the court’s narrower comparison: TikTok adoption reduced adopters’ Facebook use by about three times more than removing all ads increased it. Op. 59. Those relative magnitudes support the court’s finding.
Fourth, the court observed that the fallacy “is a risk only if Meta is in fact a monopoly.” Id. at 59 (citing PepsiCo Inc. v. Coca-Cola Co., 114 F. Supp. 2d 243, 257–58 (S.D.N.Y. 2000)). The Commission calls that circular. FTC Br. 62. But the court first considered the Commission’s evidence on profits, alleged quality degradation, and price discrimination, all of which operate independently of substitution, and it rejected each for reasons unrelated to the fallacy. Op. 24–36. It then declined to presume monopoly in order to discard evidence that undermined the Commission’s market definition. If anything deserves the label “circular,” it is the Commission’s alternative, in which its principal “proof” that prevailing conditions are monopolistic assumes the narrow market contours it alleged—the very market the discounted evidence would otherwise refute.
III. The Commission offered no measurement showing that TikTok and YouTube were too weak to constrain Meta
Whether imperfect substitute products constrain Meta turns on degree: Are the alternatives close enough, in sufficient volume, that a hypothetical monopolist of the proposed group could not profitably worsen its products or increase its quality-adjusted price? The court found no record evidence that answered that question directly, and it carefully assessed what light the available evidence did shed. Op. 41. For its part, the Commission offered no measurement of a small quality decrease, diversion analysis, or study of margins—indeed, “no empirical evidence of substitution whatsoever.” Id. at 56.
Meta, by contrast, offered several forms of evidence that consistently identified TikTok and YouTube as the closest alternatives. The Commission attacks each source in isolation and discounts those products because their functions differ—an objection that assumes the disputed market boundary. A plaintiff cannot omit the analysis its theory requires and obtain reversal because the defendant’s evidence did not perfectly disprove the case the plaintiff never made.
A. Taken together, Meta’s substitution evidence withstands the Commission’s objections
The district court observed that, “[i]n the real world, no evidence is perfect.” Op. 56. Even so, Meta’s evidence was remarkably strong, including Professor List’s randomized payment experiment, an eighteen-week panel of roughly fifty thousand users, the 2021 Meta outage, TikTok’s permanent removal from India and temporary shutdown in the United States, and the 2018 YouTube outage. Id. at 43–52. It also included Meta’s billions of dollars in defensive investment against TikTok and YouTube, id. at 54–55, evidence of perceived competitive pressure and dynamic competition.
The design criticisms leveled by the Commission against each identify a limitation of one source that another source answers. The objection that removals are “infinite” does not reach the marginal payment experiment. The objection that outages are temporary does not apply to India’s permanent ban, where substitution grew over nine months. Op. 48–50. The payment experiment’s purported brevity does not impugn ordinary TikTok adoption evidence, which saw adopters reduce Facebook usage by 17–26%. Id. at 43. And the objection that India differs from the United States is answered by the January 2025 American shutdown, which reproduced the same result domestically. Id. at 50–52.
These six sources, imperfect in different ways, all identified the same closest alternatives and the same relative ranking. Requiring each source alone to reproduce every feature of a textbook test would disregard their corroboration and demand a perfect experiment that does not exist. The court instead applied the sensible rule: “no evidence is perfect. Nor is any single piece dispositive here.” Id. at 55–56.
B. The Commission’s and amici’s evidentiary objections assume the disputed market boundary
In an effort to show that imperfect substitutes exert no constraint on Meta’s alleged monopoly, the Commission and its amici liken the evidence to supermarket customers buying pet food at PetSmart during a temporary closure, storm-driven moviegoers and candle buyers who are obviously not in “the electricity market,” and readers displaced by a power failure. FTC Br. 3–4; Econ. Br. 19–20; AAI Br. 17. Those analogies seem decisive only because they assume that the paired products serve different demands.
Each depends on an easy intuition—not a proven fact—about market boundaries, made plausible by the examples selected. It may be intuitive that a movie theater does not constrain home electricity, but whether an HMT separates, say, pet food sold at a supermarket from pet food sold at PetSmart in defining an antitrust-relevant market is an empirical question. The economists are right only that “[r]esponses to large price increases or outages “may not be the same as responses to a SSNIP.” Econ. Br. 19 (emphasis added). The possibility does not show that removal evidence is irrelevant.
Indeed, even imperfect substitution evidence should not be presumed irrelevant. Differentiated products routinely compete. See generally Jerry Hausman, Gregory Leonard & J. Douglas Zona, Competitive Analysis with Differentiated Products, 34 Annals of Econ. & Stat. 159 (1994). Understanding the HMT and the limitations of the varied record evidence, the court rejected the FTC’s proposed PSN product market, which depended on selected features of Facebook and other platforms rather than demonstrated limits on substitution.
The record also showed functional convergence and disproportionate diversion. Most time on Meta’s applications is now spent watching video; Reels was built to match TikTok; TikTok added friend-connection features; and users share on all four platforms primarily through private messages. Op. 8–9, 14–16, 66–68. Candles and electricity do not converge; these products did.
Second, the record evidence of indexed diversion ratios distinguished specific replacements from generic time-fillers by comparing redirected time with users’ ordinary time allocation. TikTok and YouTube absorbed two to three times their ordinary shares, while nearly everything else remained near parity. Id. at 46.
The economists object that much redirected time was recorded in web browsers, Econ. Br. 16, which are not in the same market. But a browser is a “portal to reach other uses,” not a use itself, id. at 45, and its indexed diversion ratio remained far below those of TikTok and YouTube. Id. at 45–46. The browser data therefore do not undermine the specific diversion to the leading social-media substitutes.
C. Products need not be identical to exert competitive pressure on each other
The Commission’s remaining objection is that this evidence measures time-shifting rather than substitution “for the same purposes.” See, e.g., FTC Br. 52. On that account, users deprived of Facebook may have replaced only its entertainment function, while its friends-and-family function went unserved and unmeasured. Id. at 47–54; see also AAI Br. 11–14.
But functional equivalence between substitutes is not required. As the district court noted, “the Supreme Court has warned against any rule that would make ‘only physically identical products . . . part of the market.’” Op. 39 (quoting du Pont, 351 U.S. at 394). Rather, the operative question is whether consumers can turn to alternatives that constrain pricing, “[b]ecause the ability of consumers to turn to other suppliers restrains a firm from raising prices above the competitive level.” Microsoft, 253 F.3d at 51–52 (quoting Rothery Storage & Van Co. v. Atlas Van Lines, Inc., 792 F.2d 210, 218 (D.C. Cir. 1986)). Microsoft’s reference to uses “for the same purposes,” id. at 52 (quoting du Pont, 351 U.S. at 395) distinguishes alternatives that exert a competitive constraint from unrelated products; it does not require—against Supreme Court precedent—that substitute uses must be identical.
For applications offered to consumers at zero monetary price, the relevant restraint on pricing occurs through diversion of advertising, which is primarily a function of where users devote their attention, regardless of whether the substitutes were physically or functionally identical. Op. 40, 82–83 (noting that “time spent is the best proxy for what drives these apps’ revenue: ads”). Functionally different applications can therefore constrain Meta if they compete for the same scarce attention. Digital firms “compete to obtain scarce time, and then they are competing to sell that time to advertisers and others who would like to have it. Importantly they are competing even if they are providing very different services to people.” Evans, Attention Rivalry, supra, at 356.
The evidence repeatedly showed substitution across products that the Commission claims were for different purposes. When American users lost TikTok in January 2025, they turned first to Facebook and Instagram, not to YouTube, even though the Commission characterizes YouTube as TikTok’s closest alternative. Op. 51. During the 2021 Meta outage—before Reels, when Facebook and Instagram remained largely friends-and-family products—users deprived of those products turned to TikTok and YouTube. Id. at 47. And experiments that withheld friend-driven Reels from Instagram saw users spending more time with “unconnected” videos. Id. at 60–61.
As the court concluded: “It is unclear whether that was because Meta’s apps are really entertainment apps, because TikTok is really a social-networking app, or because those artificial categories do not make sense. What is clear is that the FTC’s hypothesis about how people use these apps is consistently disproven by the data.” Id. at 51.
Economic analysis likewise asks whether buyers switch, not whether products share characteristics. For digital products especially, “there is a strong presumption that it is wrong to define antitrust markets by looking purely at functional substitution among products.” Evans, Attention Rivalry, supra, at 357. Indeed, an improved rival product necessarily differs from the product it displaces. It has long been understood that “[p]hysically very different products may be close substitutes for one another and thus compete heavily for the favor of the consumer. . . .” Fritz Machlup, The Political Economy of Monopoly: Business, Labor and Government Policies 8 (1952); see also Edward H. Chamberlin, The Theory of Monopolistic Competition: A Re-orientation of the Theory of Value 81 (7th ed. 1956). It is also consistent with the law. See United States v. Continental Can Co., 378 U.S. 441, 452–55 (1964) (competition is not limited “to competition between identical products” and includes “‘inter-industry competition’ . . . between products with distinctive characteristics.”).
CONCLUSION
In light of the foregoing, the opinion of the District Court should be affirmed.
[1] Hearle did not, in fact, find economic profit four times the cost of capital. He estimated an IRR, not a dollar amount of economic profit. His comparison of IRR to the weighted average cost of capital indicates a positive net present value under his assumptions, but it does not show that Meta’s economic profits were “four times” its cost of capital.