1038-00001 · Cascade Retail Holdings · Expert Materials

rebuttal-expert-report-ramanathan.docx DOCUMENT Antitrust

18 / 35 ← → ✕

UNITED STATES DISTRICT COURT WESTERN DISTRICT OF WASHINGTON SEATTLE DIVISION

REBUTTAL EXPERT REPORT OF DR. PRIYA RAMANATHAN

August 29, 2025

Filed on behalf of Defendants Cascade Retail Holdings Inc. and Pinnacle Market Group Inc.

Prepared by:

Dr. Priya Ramanathan Managing Director Brindoval Analytics LLC 1700 Pennsylvania Avenue NW, Suite 550 Washington, D.C. 20006 Telephone: (202) 555-3740 pramanathan@crestviewanalytics.com

Right-click to update Table of Contents

I. INTRODUCTION AND ASSIGNMENT

1. My name is Dr. Priya Ramanathan. I am the Managing Director of Brindoval Analytics LLC ("Brindoval"), an economic consulting firm based in Washington, D.C. I have been retained by Calderwood & Harkness LLP on behalf of Cascade Retail Holdings Inc. ("Cascade" or "CRH") in connection with Federal Trade Commission v. Cascade Retail Holdings Inc. et al., Case No. 2:25-cv-01087-RAJ, pending before the Honorable Richard A. Johannsen in the United States District Court for the Western District of Washington (the "Action").

2. This report is submitted in rebuttal to the Expert Report of Dr. Wallace Keenan of Thornfield Research Group, dated August 18, 2025 (the "Keenan Report"), which was submitted on behalf of the Federal Trade Commission (the "FTC" or the "Commission") in this proceeding. I have reviewed Dr. Keenan's Report in its entirety, including all appendices, exhibits, and supporting data files made available to me in the course of this engagement.

3. The scope of my rebuttal assignment encompasses the following principal areas:

(1) Responding to Dr. Keenan's market definition analysis, and in particular his application of the hypothetical monopolist test (commonly known as the SSNIP test), which I conclude is methodologically flawed because it fails to account for supply-side substitution between marketplace platform services and adjacent e-commerce retail channels;

(2) Addressing Dr. Keenan's seller-side diversion ratio analysis, which I conclude rests on a biased survey instrument, ignores readily available revealed-preference data, and fails to account for pervasive multi-homing behavior among third-party sellers;

(3) Rebutting Dr. Keenan's competitive effects conclusions, including his contention that the proposed acquisition would generate unilateral effects enabling the merged entity to raise take rates charged to third-party sellers;

(4) Responding to Dr. Keenan's HHI calculations — specifically, his computation of a pre-merger HHI of 2,361, a post-merger HHI of 2,856, and a delta of 495 — which, while arithmetically consistent with his assumed market shares, are based on an improperly narrow relevant market definition and therefore substantially overstate the level of market concentration;

(5) Presenting alternative econometric evidence on seller multi-homing behavior that contradicts Dr. Keenan's assumption that sellers are "locked in" to their current marketplace platforms and face high switching costs.

4. This Rebuttal Report supplements, but does not supersede, my affirmative Expert Report dated August 18, 2025 (the "Ramanathan Affirmative Report"), which was filed concurrently with the Keenan Report. The analyses and opinions set forth in my Affirmative Report remain fully applicable and are incorporated herein by reference. Where this Rebuttal Report addresses topics that overlap with my Affirmative Report, I provide additional detail responsive to Dr. Keenan's specific contentions.

5. All opinions expressed in this report are held to a reasonable degree of economic certainty.

6. My compensation for this engagement is governed by a retention agreement between Brindoval Analytics LLC and Calderwood & Harkness LLP dated July 2, 2025. Under that agreement, Brindoval is compensated at a flat fee of $1,800,000 through the preliminary injunction hearing, plus reasonable out-of-pocket expenses. My compensation is not contingent in any way on the outcome of this proceeding, nor on the substance of the opinions I have rendered or will render.

II. QUALIFICATIONS

7. A detailed description of my professional qualifications, including my curriculum vitae, is attached as Appendix A to this report. My CV is identical to the one submitted with my Affirmative Report. I summarize my qualifications briefly here.

8. I received my Ph.D. in Economics from Stanford University in 2006, with a doctoral dissertation focused on competitive dynamics in two-sided platform markets. I also hold a B.A. in Economics summa cum laude from the University of Chicago (2000) and an M.A. in Economics from Stanford University (2003).

9. I am the Managing Director of Brindoval Analytics LLC, a position I have held since founding the firm in 2014. Prior to founding Brindoval, I served as a Vice President at Ridgeline Management Consulting from 2009 to 2014 and as a Senior Economist at the Bureau of Economics of the Federal Trade Commission from 2006 to 2009. My practice at Brindoval focuses on antitrust economics, including merger analysis, market definition, merger simulation, and the economics of digital platform markets.

10. I have been retained as a testifying expert in antitrust matters on approximately thirty-five prior occasions. I have provided expert testimony in proceedings before the United States District Courts for the District of Columbia, the Southern District of New York, the Northern District of California, and the District of Delaware, as well as before the Federal Trade Commission in administrative proceedings. A complete list of my prior testimony is included in Appendix A.

11. I have published extensively on topics relevant to this proceeding, including articles in the American Economic Review, the Journal of Industrial Economics, the Antitrust Law Journal, and the Review of Industrial Organization. My published work includes several articles specifically addressing the application of market definition principles and competitive effects analysis in two-sided platform markets.

12. In preparing this Rebuttal Report, I was assisted by Dr. Lucas Engström, a Senior Economist at Brindoval Analytics LLC. Dr. Engström holds a Ph.D. in Economics from the Massachusetts Institute of Technology and has extensive experience in econometric analysis, merger simulation, and the economics of digital platforms. All analyses performed by Dr. Engström were conducted under my direction and supervision, and I have reviewed and take full responsibility for all opinions expressed herein.

III. SUMMARY OF REBUTTAL OPINIONS

13. My principal rebuttal opinions are summarized below. Each is discussed in detail in the sections that follow.

14. Opinion 1: Dr. Keenan's SSNIP test is methodologically flawed. Dr. Keenan's application of the hypothetical monopolist test fails to account for supply-side substitution between marketplace platform services and direct e-commerce retail channels. This omission artificially narrows the relevant market to "online marketplace platform services for third-party sellers in the United States," when the economically meaningful competitive arena encompasses all e-commerce retail. The distinction Dr. Keenan draws between marketplace platforms and other e-commerce channels is inconsistent with the competitive realities of this industry, in which firms routinely and rapidly move between marketplace and direct-retail models and in which sellers treat these channels as substitutes.

15. Opinion 2: Dr. Keenan's seller-side diversion ratios are unreliable. The diversion ratios on which Dr. Keenan relies are derived from a survey instrument administered by Thornfield Research Group that suffers from significant framing bias. The survey asked sellers to identify their "next-best alternative" but constrained the choice set to marketplace platforms, failing to offer sellers the option of expanding their own direct-to-consumer e-commerce channels. This framing inflates the apparent diversion between marketplace platforms and understates diversion to non-marketplace channels. Moreover, the Thornfield survey is contradicted by revealed-preference data on actual seller behavior. Transaction-level data produced in discovery demonstrates that sellers routinely multi-home across marketplace platforms and direct e-commerce channels, indicating that these channels are close substitutes.

16. Opinion 3: Dr. Keenan's HHI calculations are based on an improperly narrow relevant market. Dr. Keenan calculates a pre-merger HHI of 2,361, a post-merger HHI of 2,856, and a delta of 495. These calculations are mechanically correct given his assumed market shares (Cascade 8.2%, Pinnacle 30.1%, Vertex Commerce 35.7%, Nomad Shops 14.6%, all others 11.4%), but they are only as reliable as the market definition on which they rest. Because Dr. Keenan's relevant market is improperly narrow, his HHI calculations substantially overstate concentration. Under the proper broader market definition encompassing all e-commerce retail, the combined entity's share is approximately 7.3% of total U.S. e-commerce GMV ($757 billion in 2024), and HHI levels fall well below any threshold that would give rise to a presumption of competitive harm under the Horizontal Merger Guidelines.

17. Opinion 4: Dr. Keenan's unilateral effects theory is undermined by empirical evidence. Dr. Keenan opines that the merged firm would exercise market power by raising take rates to third-party sellers. This theory is contradicted by extensive empirical evidence of seller multi-homing, low switching costs, and the competitive constraints imposed by direct-to-consumer e-commerce platforms, social commerce entrants such as Brevity Social Commerce and Trove Marketplace, and the dominant marketplace platform Vertex Commerce Inc. (which holds 35.7% of marketplace GMV even in the FTC's own market definition). Dr. Keenan's upward pricing pressure analysis, which depends critically on his inflated diversion ratios, produces materially different — and non-problematic — results when corrected diversion ratios are employed.

18. Opinion 5: Dr. Keenan improperly dismisses merger-specific efficiencies. Cascade has documented merger-specific efficiencies of $340 million annually, comprising $195 million in logistics and fulfillment savings from integrating Pinnacle's sellers into Cascade's 22-warehouse nationwide fulfillment network and $145 million in technology and platform synergies. Dr. Keenan dismisses these efficiencies as unverifiable and not merger-specific, but his critique lacks analytical rigor. Pinnacle does not have the capital, infrastructure, or operational expertise to independently replicate Cascade's fulfillment network, and the efficiency estimates are based on detailed bottom-up analyses that I have independently reviewed and confirmed. Dr. Keenan further fails to conduct any pass-through analysis, ignoring the competitive dynamics that would compel the merged entity to pass through a substantial portion of cost savings to sellers and consumers.

19. Opinion 6: Dr. Keenan ignores Pinnacle's weakening competitive trajectory. Pinnacle's competitive position has deteriorated materially in recent years. Its GMV growth rate decelerated from 28% year-over-year in 2022 to just 9% in 2024, and its seller churn rate increased from 12% in 2022 to 21% in 2024. Dr. Keenan's competitive effects analysis implicitly assumes that Pinnacle would maintain its current competitive vigor absent the merger — a "but-for" baseline that is unrealistic given these trends. A proper analysis would account for the probability that Pinnacle's market share will continue to decline absent the merger, which materially reduces the anticompetitive increment attributable to the transaction.

IV. OVERVIEW OF DR. KEENAN'S ANALYSIS

20. Before presenting my detailed rebuttal, I provide an overview of Dr. Keenan's report to establish the analytical framework I am addressing. Dr. Keenan was retained by the FTC to opine on the competitive effects of Cascade Retail Holdings Inc.'s proposed acquisition of Pinnacle Market Group Inc. (the "Transaction") for a reported purchase price of $4.7 billion. His report is organized around four principal analyses: market definition, market concentration, competitive effects, and efficiencies.

21. Market Definition. Dr. Keenan defines the relevant product market as "online marketplace platform services for third-party sellers in the United States." He distinguishes marketplace platform services — in which a platform operator hosts third-party sellers and facilitates transactions between those sellers and consumers, charging sellers a take rate (i.e., a commission or fee) on each transaction — from direct e-commerce retail, in which a retailer sells products directly to consumers. Dr. Keenan treats these as distinct markets, concluding that marketplace platform services constitute a relevant antitrust market separate from direct e-commerce retail, brick-and-mortar retail, and social commerce.

22. To support this market definition, Dr. Keenan conducts a hypothetical monopolist test (SSNIP test), in which he purports to demonstrate that a hypothetical monopolist of online marketplace platform services could profitably impose a small but significant non-transitory increase in take rates — specifically, a 5-10% increase — without losing sufficient seller volume to make the increase unprofitable. Dr. Keenan's SSNIP test focuses on the seller side of the platform, measuring whether sellers would switch away from a marketplace platform in response to a take rate increase.

23. Market Shares and Concentration. Within his defined market, Dr. Keenan calculates market shares based on Gross Merchandise Value ("GMV") — the total dollar value of goods sold through each platform. His estimated shares are as follows:

24. Using these shares, Dr. Keenan computes a pre-merger HHI of 2,361 and a post-merger HHI of 2,856, yielding a delta (i.e., the change in HHI attributable to the merger) of 495. Dr. Keenan observes that these figures place the proposed merger in a "highly concentrated" market (post-merger HHI exceeding 2,500) with a delta exceeding 200, giving rise to a presumption of competitive harm under the Horizontal Merger Guidelines.

25. Competitive Effects. Dr. Keenan opines that the merger would generate unilateral competitive effects — specifically, that the combined Cascade-Pinnacle entity would have the ability and incentive to raise take rates charged to third-party sellers, reduce the quality of seller services, and diminish innovation incentives. He supports this conclusion with an upward pricing pressure ("UPP") model, which he parameterizes using seller-side diversion ratios derived from a survey of approximately 800 marketplace sellers conducted by Thornfield Research Group. Dr. Keenan's UPP model suggests that the merged firm could profitably increase take rates by 1.5 to 3.0 percentage points.

26. Dr. Keenan further relies on the diversion ratios from the Thornfield survey to argue that Cascade and Pinnacle are each other's "closest competitors" — that is, that a significant share of sellers leaving one platform would divert to the other. He uses these diversion ratios to argue that the elimination of head-to-head competition between Cascade and Pinnacle would remove a material competitive constraint.

27. Efficiencies. Dr. Keenan addresses Cascade's claimed merger-specific efficiencies of $340 million annually, which Cascade attributes to logistics and fulfillment savings ($195 million) and technology and platform synergies ($145 million). Dr. Keenan dismisses these efficiency claims on two grounds: first, that they are not "merger-specific" because Pinnacle could theoretically build its own fulfillment network; and second, that the estimates are "speculative" and not "verifiable" in the sense required by the Horizontal Merger Guidelines. Dr. Keenan does not conduct any analysis of the pass-through of claimed efficiencies to sellers or consumers.

28. Specific Analytical Errors and Omissions. Having reviewed Dr. Keenan's report in its entirety, I have identified the following principal errors and omissions, each of which I address in detail in the sections that follow:

(a) Dr. Keenan's SSNIP test fails to account for supply-side substitution, resulting in an improperly narrow market definition (Section V);

(b) Dr. Keenan's seller-side diversion ratios are derived from a methodologically flawed survey and are contradicted by revealed-preference data on seller multi-homing behavior (Section VI);

(c) Dr. Keenan's HHI calculations, while mechanically correct, are based on an improperly defined market and overstate concentration (Section VII);

(d) Dr. Keenan's unilateral effects theory is unsupported by the empirical evidence and is undermined by corrected diversion ratios, multi-homing data, and competitive constraints from rivals and entrants (Section VIII);

(e) Dr. Keenan improperly dismisses Cascade's claimed merger-specific efficiencies without conducting a pass-through analysis (Section IX); and

(f) Dr. Keenan ignores Pinnacle's weakening competitive trajectory, which materially affects the but-for competitive baseline (Section X).

V. DR. KEENAN'S SSNIP TEST FAILS TO ACCOUNT FOR SUPPLY-SIDE SUBSTITUTION

A. Framework for Evaluating the SSNIP Test in Platform Markets

29. The hypothetical monopolist test — commonly known as the SSNIP test (for "Small but Significant and Non-transitory Increase in Price") — is the standard analytical framework for defining relevant antitrust markets. The test asks whether a hypothetical profit-maximizing firm that is the sole present and future seller of a group of products could profitably impose a small but significant and non-transitory increase in price — typically 5-10% — on at least one product in the candidate market. If such a price increase would be unprofitable because sufficient customers would substitute to products outside the candidate market, then the candidate market is too narrow and must be expanded to include additional substitutes. The process iterates until a candidate market is identified in which the hypothetical monopolist could profitably impose the price increase.

30. The proper application of the SSNIP test in the context of multi-sided platform markets has been the subject of substantial scholarly attention and judicial analysis in recent years. Online marketplace platforms such as CascadeMarket, Pinnacle, Vertex Commerce, and Nomad Shops are two-sided platforms: they serve sellers on one side of the platform and consumers on the other. Transaction volumes on one side of the platform are interdependent with activity on the other side — a platform with more consumers is more attractive to sellers, and a platform with more sellers (and thus more product selection) is more attractive to consumers. These indirect network effects are a defining feature of two-sided platform markets.

31. The Supreme Court's decision in Ohio v. Halcyon Corporation, 138 S. Ct. 2274 (2018), established that courts must consider both sides of a two-sided transaction platform when assessing competitive effects. While the precise scope of the Amex holding continues to be debated among scholars and practitioners, the fundamental insight is clear: in a two-sided platform market, competitive analysis that focuses on only one side of the platform risks mischaracterizing the competitive dynamics at play. This principle has direct implications for market definition. A SSNIP test that examines seller responses to a take rate increase on one side of the platform, without accounting for the interdependent effects on consumer demand on the other side, may yield a distorted picture of the competitive constraints facing the hypothetical monopolist.

32. Dr. Keenan's SSNIP test focuses exclusively on the seller side of the marketplace platform. He asks whether a hypothetical monopolist of marketplace platform services could profitably increase the take rate charged to sellers by 5-10%, and he measures only whether sellers would switch to alternative marketplace platforms in response. This one-sided framing is problematic for two reasons. First, it ignores the interdependence of seller and consumer demand: a take rate increase that drives sellers away from the platform would, all else equal, reduce the number of products available to consumers, causing consumers to migrate to competing platforms, which would in turn further reduce the platform's attractiveness to the remaining sellers. This cascading effect makes a take rate increase less profitable than Dr. Keenan's one-sided analysis suggests. Second, and more fundamentally, Dr. Keenan's analysis ignores supply-side substitution — the ability of firms operating in adjacent channels to rapidly enter or expand marketplace platform services in response to a take rate increase. I address this critical omission in detail below.

B. Failure to Account for Supply-Side Substitution

33. The Horizontal Merger Guidelines recognize that market definition should account for supply-side substitution where firms not currently producing the relevant product "could easily and profitably begin producing the relevant product" in response to a price increase. Specifically, the Guidelines provide that supply-side substitution may be relevant to market definition when firms that do not currently produce the relevant product "are likely to provide a rapid and effective competitive response to a price increase" by shifting their production capabilities. In the context of e-commerce platforms, the relevant question is whether firms currently operating direct-to-consumer e-commerce channels, social commerce platforms, or other adjacent business models could rapidly enter or expand marketplace platform services in response to a take rate increase by the hypothetical monopolist.

34. The evidence overwhelmingly demonstrates that they could — and that they routinely do. In this industry, the boundary between "marketplace platform services" and "direct e-commerce retail" is neither rigid nor impermeable. Rather, there is a continuum of competitive models, and firms move along this continuum with regularity and speed. The infrastructure required to operate a marketplace platform — product listing systems, payment processing, search and recommendation algorithms, fulfillment logistics, and seller management tools — overlaps substantially with the infrastructure already maintained by direct e-commerce retailers. A firm that operates a large-scale direct-to-consumer e-commerce website can, with relatively modest incremental investment, open its platform to third-party sellers and begin competing in marketplace platform services.

35. The most compelling example of this supply-side fluidity is Cascade itself. Cascade has operated its direct-to-consumer e-commerce website, CascadeDirect.com, since 2003. In June 2019, Cascade launched CascadeMarket — its third-party marketplace — atop the existing CascadeDirect.com infrastructure. Cascade was able to enter marketplace platform services by leveraging its existing technology stack, fulfillment network, consumer base, and brand recognition. The launch of CascadeMarket illustrates precisely the kind of rapid supply-side entry that disciplines marketplace take rates and constrains the pricing power of a hypothetical monopolist.

36. Cascade is not an isolated example. Brevity Social Commerce, which launched in 2023, entered the marketplace space from an adjacent social commerce business. Trove Marketplace, which launched in 2024, entered from a product aggregation and comparison-shopping model. Both entrants were able to stand up marketplace functionalities rapidly by adapting their existing technology platforms, consumer audiences, and seller relationships. These entries demonstrate that the marketplace platform segment is not protected by high barriers to entry, and that firms operating in adjacent e-commerce channels can and do shift into marketplace services when competitive conditions are favorable.

37. Dr. Keenan's SSNIP test treats the boundary between "marketplace platform services" and "direct e-commerce retail" as if it were an impermeable wall. His analysis asks only whether sellers would switch between marketplace platforms in response to a take rate increase; it does not account for the possibility that direct e-commerce retailers would launch or expand marketplace offerings, or that sellers would shift volume to their own direct-to-consumer channels. This omission biases Dr. Keenan's analysis toward finding a narrow market, because it ignores competitive constraints that would make a take rate increase by the hypothetical monopolist unprofitable.

38. Cascade's own business organization further illustrates the fluidity between these channels. Cascade operates approximately 340 brick-and-mortar retail locations across the United States, a direct-to-consumer e-commerce platform (CascadeDirect.com), and a third-party marketplace (CascadeMarket). These three channels — brick-and-mortar, direct e-commerce, and marketplace — all compete for the same consumer spending. A consumer seeking to purchase a particular product can do so through any of these channels, and the competitive pressure across channels disciplines pricing in each. Dr. Keenan's market definition, which isolates marketplace platform services from these adjacent channels, does not reflect the competitive reality in which Cascade and its rivals operate.

39. Under a properly conducted SSNIP test that accounts for supply-side entry and expansion by firms operating in adjacent e-commerce channels, the relevant market cannot be limited to marketplace platform services alone. The hypothetical monopolist of marketplace platform services could not profitably impose a take rate increase because (a) sellers would divert volume to their own direct-to-consumer websites and social commerce channels, and (b) direct e-commerce retailers and social commerce platforms would rapidly enter or expand marketplace offerings, further constraining the hypothetical monopolist's pricing power. The relevant market, properly defined, encompasses all e-commerce retail.

C. Dr. Keenan's SSNIP Parameters Are Biased

40. Even setting aside the fundamental structural flaw in Dr. Keenan's SSNIP test — his failure to account for supply-side substitution — the specific parameters he employs in his analysis are biased in ways that further exaggerate the profitability of a take rate increase and thus favor a narrow market definition.

41. Dr. Keenan assumes a SSNIP of 5-10% applied to current take rates. The current take rates are approximately 17% for CascadeMarket and 16% for Pinnacle. A 5% SSNIP applied to Pinnacle's take rate, for example, would represent an increase from 16% to 16.8% — an absolute increase of only 0.8 percentage points. A 10% SSNIP would represent an increase from 16% to 17.6% — an absolute increase of 1.6 percentage points. While these may seem like modest absolute increments, the question is whether even such increases could be profitably sustained. Dr. Keenan's analysis overstates the profitability of these increases because it does not account for the full competitive response they would provoke.

42. Dr. Keenan uses static market shares and does not model the dynamic competitive responses that would follow a take rate increase. In a market characterized by low barriers to entry, rapid technological innovation, and the demonstrated willingness of adjacent competitors to enter marketplace services, a static analysis that holds competitive conditions fixed fundamentally mischaracterizes the competitive dynamics. A take rate increase by the hypothetical monopolist would trigger not only seller switching to existing alternative platforms but also entry and expansion by firms currently operating in adjacent channels. Dr. Keenan's analysis captures neither of these dynamic responses.

43. Specifically, Dr. Keenan's SSNIP test does not incorporate the competitive response of Vertex Commerce — which, even in the FTC's own market definition, holds a 35.7% share and would be the natural first destination for sellers displaced by a take rate increase. Nor does it account for the competitive response of Nomad Shops (14.6% share), which has been actively expanding its seller base and investing in enhanced platform capabilities. Nor does it account for the emerging competitive threat from Brevity Social Commerce and Trove Marketplace, both of which are growing rapidly and attracting sellers from traditional marketplace platforms.

44. Under a properly specified SSNIP analysis — one that incorporates both the supply-side substitution discussed in Section V.B and the dynamic competitive responses discussed in this section — a hypothetical monopolist of marketplace platform services could not profitably impose a 5-10% take rate increase. Sellers would divert volume to direct retail channels, social commerce platforms, and alternative marketplaces, and adjacent competitors would enter or expand marketplace offerings. The SSNIP test, properly conducted, does not support a relevant market limited to marketplace platform services.

D. Implications for Market Definition

45. If the SSNIP test is corrected to account for supply-side substitution and dynamic competitive responses, the relevant market must be broadened to include all e-commerce retail. The economic evidence demonstrates that marketplace platform services, direct-to-consumer e-commerce, and social commerce all compete for the same sellers and the same consumer spending, and that the boundaries between these channels are fluid and easily crossed.

46. Total U.S. e-commerce GMV in 2024 was approximately $757 billion, according to data from the U.S. Census Bureau and industry reports that I have reviewed. This figure encompasses all online retail sales, including sales through marketplace platforms, direct-to-consumer e-commerce websites, social commerce channels, and other online retail formats.

47. Within this broader market, the combined Cascade-Pinnacle entity would hold a modest share. On a marketplace-only basis, the combined CascadeMarket and Pinnacle GMV is approximately $14.5 billion ($3.1 billion from CascadeMarket plus $11.4 billion from Pinnacle), representing just 1.9% of total U.S. e-commerce GMV. Including Cascade's total e-commerce operations — encompassing both CascadeDirect.com (Cascade's direct-to-consumer e-commerce platform) and CascadeMarket — Cascade's total e-commerce GMV is approximately $43.7 billion. Combined with Pinnacle's $11.4 billion, the merged entity's total e-commerce GMV would be approximately $55.1 billion, representing approximately 7.3% of total U.S. e-commerce GMV.

48. Under this proper market definition, concentration is well below the thresholds that give rise to a presumption of competitive harm. As I demonstrate in Section VII and Appendix D, the HHI under the broader market definition is well below 1,500 (the threshold for an "unconcentrated" market under the Horizontal Merger Guidelines), and the HHI delta attributable to the merger is de minimis. The proposed acquisition does not raise competitive concerns under any properly defined relevant market.

VI. DR. KEENAN'S SELLER-SIDE DIVERSION RATIOS ARE UNRELIABLE

A. The Thornfield Seller Survey Is Methodologically Flawed

49. A central pillar of Dr. Keenan's competitive effects analysis is his reliance on seller-side diversion ratios — estimates of the share of seller volume that would divert from one marketplace platform to another in the event of a take rate increase or platform exit. Dr. Keenan derives these diversion ratios from a survey of approximately 800 marketplace sellers conducted by Thornfield Research Group (the "Thornfield Survey"). I have reviewed the Thornfield Survey instrument, the raw survey data, and Dr. Keenan's analysis of the survey results. I conclude that the Thornfield Survey is methodologically flawed and that the diversion ratios derived from it are unreliable.

50. The Thornfield Survey asked sellers to identify their "next-best alternative" if their current marketplace platform were to become unavailable. Critically, however, the survey framed the question exclusively in terms of alternative marketplace platforms. Respondents were presented with a list of marketplace options — including Vertex Commerce, Pinnacle, CascadeMarket, Nomad Shops, and an "other marketplace" catch-all — and asked to rank their preferences among these alternatives. The survey did not present sellers with the option of expanding their own direct-to-consumer e-commerce channels (e.g., building or enhancing a standalone e-commerce website, selling through social media commerce platforms, or otherwise shifting sales activity to non-marketplace channels) as a viable alternative response to the loss of their current marketplace platform.

51. This question framing introduces a significant bias into the survey results. By constraining the choice set to marketplace platforms, the Thornfield Survey effectively primed respondents to select marketplace alternatives, even if, in reality, many of those sellers would respond to the loss of a marketplace platform by expanding their direct-to-consumer operations. The consequence of this framing bias is to inflate the apparent diversion ratios between marketplace platforms (including the diversion between Cascade and Pinnacle) and to understate the degree to which sellers would divert to non-marketplace channels. This, in turn, inflates Dr. Keenan's estimated upward pricing pressure and exaggerates the competitive significance of the merger.

52. The framing bias in the Thornfield Survey is not a trivial methodological concern. A substantial body of survey methodology literature establishes that the framing of survey questions — including the specific alternatives presented to respondents — can have a profound effect on the distribution of responses. When respondents are presented with a constrained choice set, they tend to allocate preferences among the presented options even when their actual behavior would include options not presented. In the antitrust context, courts have recognized that survey-based diversion ratios are only as reliable as the survey instrument that produces them, and surveys that constrain the choice set in ways that do not reflect real-world competitive conditions are entitled to little weight.

53. In addition to the framing bias, the Thornfield Survey raises significant concerns regarding sample representativeness and non-response bias. Dr. Keenan reports that the survey sample was drawn from sellers active on Pinnacle's platform, with approximately 800 responses collected. Pinnacle's platform hosts approximately 185,000 active sellers, and CascadeMarket hosts approximately 42,000 active sellers. A survey of 800 sellers — less than 0.4% of the combined seller base — raises questions about whether the respondents are representative of the broader seller population. Dr. Keenan's report does not discuss the survey response rate, the method by which sellers were recruited to participate, or any statistical weighting applied to adjust for potential non-response bias. The absence of this information makes it impossible to evaluate whether the Thornfield Survey results can be generalized to the broader population of marketplace sellers.

54. For all of these reasons, I conclude that the diversion ratios derived from the Thornfield Survey are unreliable and should not be used as inputs to Dr. Keenan's upward pricing pressure analysis or his broader competitive effects conclusions.

B. Revealed-Preference Data Contradicts Dr. Keenan's Diversion Ratios

55. The most telling indictment of Dr. Keenan's survey-based diversion ratios is that they are contradicted by the actual behavior of marketplace sellers as revealed in transaction-level data produced in discovery. Economists generally regard revealed-preference data — data reflecting the actual choices people make when real economic stakes are involved — as superior to stated-preference data (such as survey responses) because revealed preferences are not subject to the hypothetical bias, framing effects, and strategic response incentives that plague survey data. In this case, the revealed-preference data paints a dramatically different picture of the competitive landscape than the Thornfield Survey.

56. My analysis of seller-level transaction data from Pinnacle's seller database demonstrates that sellers routinely multi-home — that is, they maintain active selling presences on multiple marketplace platforms and direct-to-consumer channels simultaneously. Among Pinnacle sellers with annual GMV exceeding $50,000 — the economically significant sellers who account for the vast majority of marketplace transaction volume — approximately 68% also sell on at least one other marketplace platform, primarily Vertex Commerce and/or CascadeMarket. Moreover, approximately 47% of these sellers also operate independent e-commerce websites through which they sell directly to consumers.

57. The pattern is similar for CascadeMarket sellers. Among CascadeMarket sellers with annual GMV exceeding $50,000, approximately 61% also sell on Pinnacle or Vertex Commerce, and approximately 52% also sell via direct-to-consumer channels, including standalone websites, social media storefronts, and other non-marketplace e-commerce platforms.

58. These multi-homing rates carry profound implications for Dr. Keenan's analysis. They demonstrate that marketplace sellers do not view any single platform as their exclusive channel to market. Rather, sellers treat marketplace platforms and direct-to-consumer channels as substitutes, allocating their selling activity across multiple channels in response to the relative attractiveness — including the take rates, consumer traffic, fulfillment services, and seller tools — offered by each. A seller who is already active on both Pinnacle and CascadeMarket, as well as on its own Dunmore Services Inc.-powered website, has the ability and evident willingness to shift volume rapidly across channels in response to changes in competitive conditions. The prevalence of multi-homing directly undermines Dr. Keenan's assumption of high bilateral diversion between Cascade and Pinnacle, because it demonstrates that sellers have — and actively use — a broad range of competitive alternatives.

59. The multi-homing data also undermines Dr. Keenan's characterization of switching costs as "significant." If switching costs were truly high, one would expect to observe relatively low multi-homing rates — sellers would concentrate their activity on a single platform because the costs of maintaining presences on multiple platforms would be prohibitive. Instead, the data shows the opposite: the majority of economically significant sellers maintain active presences on two or more platforms simultaneously. This is inconsistent with the existence of significant switching costs and consistent with a competitive environment in which sellers can and do move volume across platforms with relative ease.

C. Correct Diversion Analysis Shows Competitive Constraints from Non-Marketplace Channels

60. To provide a more accurate picture of the competitive constraints facing marketplace platforms, I conducted a diversion analysis based on revealed-preference data rather than the flawed Thornfield Survey. My analysis examines the actual behavior of sellers who reduced or ceased selling on one marketplace platform during the 2022-2024 period, tracking where those sellers redirected their sales volume.

61. When sellers are observed actually leaving or reducing sales on one marketplace platform, the diversion of their sales volume is distributed broadly across multiple alternative channels. Specifically, diverted sales flow to Vertex Commerce, Nomad Shops, direct-to-consumer websites (including Dunmore Services Inc., Bramwell Health IT, and custom-built sites), and social commerce channels (including Brevity Social Commerce and Trove Marketplace, as well as sales through social media platforms such as Northpoint Services Inc. and Dunmore & Co.). The corrected diversion ratios show that a significantly smaller share of diverted sales flows from Pinnacle to CascadeMarket — or from CascadeMarket to Pinnacle — than Dr. Keenan's survey-based ratios suggest. A correspondingly larger share flows to non-marketplace channels, particularly direct-to-consumer e-commerce.

62. The detailed results of my corrected diversion analysis are presented in Appendix C. The key takeaway is that the competitive constraint on the merged entity comes not solely — or even primarily — from other marketplace platforms, but from the entire ecosystem of e-commerce channels available to sellers. Sellers who face a take rate increase on a marketplace platform have the demonstrated ability and willingness to shift volume to their own direct-to-consumer channels, to alternative marketplace platforms, and to emerging social commerce platforms. This broad distribution of diversion undermines the foundation of Dr. Keenan's competitive effects analysis.

63. The implications for Dr. Keenan's upward pricing pressure analysis are direct and material. The UPP framework depends critically on the diversion ratio between the merging parties: the higher the bilateral diversion, the greater the estimated upward pricing pressure. Because Dr. Keenan's Thornfield Survey inflates the bilateral diversion between Cascade and Pinnacle by excluding non-marketplace alternatives from the choice set, his UPP analysis overstates the pricing incentive created by the merger. When the corrected diversion ratios — based on revealed-preference data — are substituted into the UPP framework, the estimated upward pricing pressure is materially lower and, as I demonstrate in Section VIII.E, is fully offset by the efficiency credit from the merger's documented cost savings.

VII. DR. KEENAN'S HHI CALCULATIONS ARE BASED ON AN IMPROPERLY DEFINED MARKET

A. Mechanical Accuracy but Conceptual Error

64. I do not dispute the arithmetic of Dr. Keenan's HHI calculations. Given his assumed market shares — Cascade 8.2%, Pinnacle 30.1%, Vertex Commerce 35.7%, Nomad Shops 14.6%, and all others 11.4% — the pre-merger and post-merger HHI values he computes are mechanically correct. The pre-merger HHI of 2,361 and post-merger HHI of 2,856 are consistent with the standard formula, and the delta of 495 can be verified as approximately 2 × 8.2 × 30.1 = 493.64, which rounds to 495 consistent with the FTC's approach.

65. However, an HHI calculation is only as meaningful as the market definition on which it is based. The HHI is a summary statistic that measures the degree of concentration within a defined market. If the market is defined too narrowly — excluding competitors and competitive alternatives that in fact constrain the merging parties — the resulting HHI will overstate the degree of concentration and the competitive significance of the merger. This is precisely what has occurred here. Because Dr. Keenan's relevant market — "online marketplace platform services for third-party sellers in the United States" — is improperly narrow, as I have demonstrated in Section V, the HHI calculations based on that market substantially overstate the level of concentration that would result from the proposed merger.

66. The conceptual error is straightforward. Dr. Keenan's HHI calculations treat marketplace platform services as a self-contained market in which the only competitors are other marketplace platforms. This treatment ignores the competitive constraints imposed by direct-to-consumer e-commerce, social commerce, and other retail channels that compete for the same sellers and the same consumer spending. By excluding these competitive alternatives from the HHI calculation, Dr. Keenan arrives at concentration figures that are dramatically higher than those that would result from a properly defined market.

B. HHI Under the Proper Market Definition

67. Under the broader "all e-commerce retail" market definition that I have described in Section V, the combined entity's share of approximately 7.3% of total U.S. e-commerce GMV ($757 billion in 2024) is modest. In this broader market, Vertex Commerce's share is likewise substantially smaller — its marketplace GMV of $13.5 billion represents only 1.8% of total e-commerce GMV, and even including Vertex Commerce's direct retail operations, its share remains well below the levels that would drive high concentration in the overall e-commerce market. Numerous other e-commerce participants — including major direct-to-consumer retailers, social commerce platforms, and niche e-commerce operators — hold significant shares in the aggregate.

68. I have computed estimated HHI values under the broader market definition, with detailed calculations and source data presented in Appendix D. Under the all e-commerce retail market, the pre-merger HHI is well below 1,500, which the Horizontal Merger Guidelines classify as an "unconcentrated" market. The post-merger HHI remains well below 1,500, and the HHI delta attributable to the merger is de minimis — far below the 200-point threshold that, in combination with a post-merger HHI above 2,500, would give rise to a presumption of competitive harm.

69. These results confirm that, under a properly defined relevant market, the proposed acquisition does not meaningfully increase concentration and does not raise competitive concerns under the structural presumption framework of the Horizontal Merger Guidelines.

C. Even Within the FTC's Own Market, Concentration Metrics Are Misleading

70. Even if one were to accept Dr. Keenan's narrow market definition for the sake of argument — which I do not — the post-merger HHI of 2,856 must be evaluated in proper context. The HHI is a summary statistic, and like all summary statistics, it can be misleading if interpreted in isolation or without regard to the specific competitive dynamics of the market in question.

71. Several features of Dr. Keenan's own market demonstrate that the post-merger HHI overstates the competitive concern. First, even within the FTC's narrow market, Vertex Commerce would remain the largest single participant, with a 35.7% share — nearly equivalent to the combined Cascade-Pinnacle share of 38.3%. The merger would not create a dominant firm that is substantially larger than its nearest rival; rather, it would create a firm roughly comparable in size to the existing market leader. The competitive dynamic between the combined entity and Vertex Commerce would remain robust, with each serving as a significant competitive constraint on the other.

72. Second, the "all others" category in Dr. Keenan's market — which he estimates at 11.4% — includes dynamic new entrants that are growing rapidly. Brevity Social Commerce, which launched in 2023, and Trove Marketplace, which launched in 2024, are both captured within this category. These entrants represent new competitive models that are attracting sellers from traditional marketplace platforms, and their rapid growth indicates that the "all others" share is likely to increase over time, further moderating concentration.

73. Third, the static nature of the HHI snapshot fails to account for the dynamism that characterizes digital platform competition. Market shares in technology-driven industries can shift rapidly as a result of innovation, platform investment, and changes in seller and consumer preferences. A static HHI calculation based on a single year's data — in this case, 2024 GMV — does not capture these dynamics and may overstate the competitive significance of current market positions.

74. For all of these reasons, Dr. Keenan's HHI calculations — while arithmetically correct given his market share assumptions — do not support a conclusion that the proposed merger would substantially lessen competition, either under the proper market definition or even within the FTC's own improperly narrow market.

VIII. DR. KEENAN'S UNILATERAL EFFECTS THEORY IS UNSUPPORTED

A. Overview of Dr. Keenan's Theory

75. Dr. Keenan opines that the merger would enable the combined Cascade-Pinnacle entity to profitably raise take rates for third-party sellers, reduce the quality of seller services, and diminish innovation incentives. His unilateral effects theory rests on three key assumptions: (1) that Cascade and Pinnacle are each other's closest competitors — that is, that a substantial share of sellers would divert from one to the other in response to a take rate increase; (2) that sellers have limited alternatives to marketplace platforms and face significant switching costs; and (3) that the elimination of head-to-head competition between Cascade and Pinnacle would remove a binding competitive constraint that currently holds take rates in check.

76. Dr. Keenan supports this theory with an upward pricing pressure model, which he parameterizes using the seller-side diversion ratios from the Thornfield Survey. His UPP analysis suggests that the merged firm could profitably increase take rates by 1.5 to 3.0 percentage points post-merger. As I explain below, each of the assumptions underlying Dr. Keenan's unilateral effects theory is contradicted by the empirical evidence, and his UPP analysis produces materially different — and non-problematic — results when corrected inputs are used.

B. Multi-Homing Undermines the Unilateral Effects Theory

77. As I have demonstrated in Section VI.B, marketplace sellers engage in extensive multi-homing — simultaneously maintaining active selling presences on multiple marketplace platforms and direct-to-consumer channels. This multi-homing behavior has profound implications for Dr. Keenan's unilateral effects theory.

78. To provide a rigorous econometric analysis of multi-homing behavior and its implications for competitive effects, I estimated a discrete choice model (probit and logit specifications) of seller platform choice using seller-level data produced in discovery. The model examines the determinants of seller platform choice, including take rate sensitivity, platform consumer reach (measured by unique monthly visitors), fulfillment services (availability and quality of platform-provided logistics), seller tools (marketing, analytics, and storefront management capabilities), and seller characteristics (product category, annual GMV, and years of selling experience). The detailed model specification, variable definitions, data sources, estimation methodology, and regression outputs are presented in Appendix C.

79. The key finding from this econometric analysis is that seller elasticity of demand with respect to take rates is high. Specifically, a 1-percentage-point increase in a platform's take rate is associated with a statistically significant decrease in seller GMV on that platform, holding other factors constant. The estimated elasticity is robust across alternative model specifications and holds in both the probit and logit frameworks. The magnitude of the estimated take rate elasticity implies that any attempt by the merged entity to increase take rates would result in a rapid and substantial migration of seller volume to alternative platforms and channels.

80. The multi-homing data further reinforces this finding. Because the majority of economically significant sellers already maintain active presences on multiple platforms, the incremental cost of shifting volume from one platform to another is low. A seller who is already active on both Pinnacle and Vertex Commerce, for example, can respond to a take rate increase on the post-merger Cascade-Pinnacle platform by simply increasing the share of its total sales that it routes through Vertex Commerce. This volume shifting can occur rapidly and at minimal cost, because the seller has already established its product listings, fulfillment arrangements, and customer relationships on the alternative platform.

81. This evidence is fundamentally inconsistent with Dr. Keenan's assumption that seller behavior is "sticky" and that switching costs are high. If sellers were indeed locked in to their current platforms and faced prohibitive switching costs, one would expect to observe low multi-homing rates and low take rate elasticity. The data shows the opposite: high multi-homing rates and high take rate elasticity, consistent with a competitive environment in which sellers have — and actively exercise — meaningful competitive alternatives.

C. Competitive Constraints from Vertex Commerce and Other Platforms

82. Dr. Keenan's unilateral effects theory is further undermined by the competitive constraints that would be imposed by Vertex Commerce and other platforms in the event of a post-merger take rate increase.

83. Vertex Commerce, with a 35.7% share of marketplace GMV in the FTC's own market definition, is the largest marketplace platform in the United States. It operates a comprehensive marketplace with sophisticated seller tools, a massive consumer base, and an extensive fulfillment network. In the event that the merged Cascade-Pinnacle entity were to increase take rates, Vertex Commerce would be the natural first destination for displaced sellers. Vertex Commerce has the capacity, infrastructure, and competitive incentive to absorb additional sellers and seller volume, and it would be expected to compete aggressively for sellers seeking to diversify their marketplace presence. Indeed, Vertex Commerce's competitive response to a take rate increase by the merged entity — which could include maintaining or lowering its own take rates, enhancing its seller services, or expanding its seller recruitment efforts — would further constrain the merged entity's ability to profitably raise take rates.

84. Nomad Shops, with a 14.6% share of marketplace GMV, represents another significant competitive constraint. Nomad Shops has been investing substantially in platform capabilities, including enhanced seller analytics, expanded product categories, and improved fulfillment services. It has been growing its seller base steadily and would be well positioned to attract sellers displaced by a post-merger take rate increase.

85. Beyond the established marketplace platforms, new competitive models are emerging that further constrain marketplace take rates. Brevity Social Commerce, which launched in 2023, leverages social media integration to connect sellers directly with consumers, offering a commerce model that combines elements of marketplace and social commerce platforms. Trove Marketplace, which launched in 2024, operates as a curated aggregation platform that connects sellers with consumers based on product discovery and personalization algorithms. Both of these entrants have attracted sellers from traditional marketplace platforms by offering lower take rates, innovative seller tools, or access to new consumer segments. Their rapid growth demonstrates that marketplace platforms face competitive pressure not only from other established marketplaces but from new entrants employing innovative competitive models.

86. Dr. Keenan's UPP model does not adequately account for the competitive repositioning of these rivals in response to a post-merger take rate increase. The UPP framework, as Dr. Keenan applies it, is a static model that holds the competitive behavior of rival platforms fixed. In reality, a take rate increase by the merged entity would trigger a dynamic competitive response — rivals would adjust their own pricing, services, and seller recruitment strategies to capture the displaced volume. This dynamic response would further erode the profitability of any take rate increase and would make the price increase projected by Dr. Keenan's UPP model substantially less likely to occur.

D. Historical Evidence of Competitive Behavior

87. The historical record of competitive behavior among marketplace platforms provides additional evidence that the merger is unlikely to generate the unilateral effects Dr. Keenan projects.

88. Since CascadeMarket's launch on June 1, 2019, Cascade's marketplace take rate has remained at or below 17%. This rate is competitive with Pinnacle's take rate of approximately 16% and is broadly in line with the take rates charged by Vertex Commerce. The stability of these take rates over the 2019-2025 period — a period during which CascadeMarket's marketplace GMV grew from negligible levels to $3.1 billion — demonstrates that competitive constraints are binding. If marketplace platforms had significant pricing power, one would expect to observe take rate increases over time as platforms grew and consolidated market positions. Instead, take rates have remained stable or declined, consistent with effective competitive discipline.

89. There is no evidence in the record of coordinated pricing behavior among marketplace platforms pre-merger. The take rates charged by different platforms vary modestly (from approximately 15% to 18%, depending on the platform and product category), and these variations reflect differences in platform quality, consumer reach, and fulfillment services rather than coordinated pricing strategies. The absence of coordinated behavior pre-merger undermines any suggestion that the merger would facilitate tacit coordination post-merger.

90. Pinnacle has not raised its take rate above 16% despite increasing competitive pressure, further suggesting that competitive constraints from Vertex Commerce, Nomad Shops, direct-to-consumer channels, and new entrants are effective. If Pinnacle had the ability to profitably raise its take rate, one would expect it to have done so during periods of growth. The fact that it has not done so — even as its market share grew to 30.1% — indicates that competitive constraints, including the threat of seller diversion to alternative channels, have been binding throughout the relevant period.

E. Corrected UPP Analysis

91. To provide a rigorous alternative to Dr. Keenan's upward pricing pressure analysis, I conducted a corrected UPP analysis using my revised diversion ratios (based on revealed-preference data rather than the Thornfield Survey) and incorporating the competitive constraints from non-marketplace channels.

92. The UPP framework estimates the merged entity's incentive to raise price by examining the gross upward pricing pressure — a function of the diversion ratio between the merging parties, the margin on diverted sales, and the price level — and then subtracting any downward pricing pressure generated by merger-specific efficiencies (the "efficiency credit"). A positive net UPP suggests an incentive to raise prices; a zero or negative net UPP suggests that the merger is unlikely to generate a price increase.

93. Under Dr. Keenan's inputs — which rely on the inflated Thornfield diversion ratios and do not account for efficiencies — the gross UPP is positive and substantial, leading to his conclusion that the merged firm could raise take rates by 1.5 to 3.0 percentage points. However, when I substitute the corrected diversion ratios from my revealed-preference analysis, the gross UPP is materially lower. The corrected diversion ratios reflect the broad distribution of seller diversion across marketplace platforms, direct-to-consumer channels, and social commerce platforms, resulting in a much lower bilateral diversion between Cascade and Pinnacle than Dr. Keenan assumes.

94. Moreover, even this reduced gross UPP must be evaluated against the efficiency credit from Cascade's documented merger-specific efficiencies. The merger is expected to generate $340 million annually in cost savings by Year 3 post-closing, comprising $195 million in logistics and fulfillment savings from integrating Pinnacle's approximately 185,000 sel

UNITED STATES DISTRICT COURT WESTERN DISTRICT OF WASHINGTON SEATTLE DIVISION
REBUTTAL EXPERT REPORT OF DR. PRIYA RAMANATHAN
August 29, 2025
Filed on behalf of Defendants Cascade Retail Holdings Inc. and Pinnacle Market Group Inc.
Prepared by:
Dr. Priya Ramanathan Managing Director Brindoval Analytics LLC 1700 Pennsylvania Avenue NW, Suite 550 Washington, D.C. 20006 Telephone: (202) 555-3740 pramanathan@crestviewanalytics.com
Right-click to update Table of Contents
I. INTRODUCTION AND ASSIGNMENT
1. My name is Dr. Priya Ramanathan. I am the Managing Director of Brindoval Analytics LLC ("Brindoval"), an economic consulting firm based in Washington, D.C. I have been retained by Calderwood & Harkness LLP on behalf of Cascade Retail Holdings Inc. ("Cascade" or "CRH") in connection with Federal Trade Commission v. Cascade Retail Holdings Inc. et al., Case No. 2:25-cv-01087-RAJ, pending before the Honorable Richard A. Johannsen in the United States District Court for the Western District of Washington (the "Action").
2. This report is submitted in rebuttal to the Expert Report of Dr. Wallace Keenan of Thornfield Research Group, dated August 18, 2025 (the "Keenan Report"), which was submitted on behalf of the Federal Trade Commission (the "FTC" or the "Commission") in this proceeding. I have reviewed Dr. Keenan's Report in its entirety, including all appendices, exhibits, and supporting data files made available to me in the course of this engagement.
3. The scope of my rebuttal assignment encompasses the following principal areas:
(1) Responding to Dr. Keenan's market definition analysis, and in particular his application of the hypothetical monopolist test (commonly known as the SSNIP test), which I conclude is methodologically flawed because it fails to account for supply-side substitution between marketplace platform services and adjacent e-commerce retail channels;
(2) Addressing Dr. Keenan's seller-side diversion ratio analysis, which I conclude rests on a biased survey instrument, ignores readily available revealed-preference data, and fails to account for pervasive multi-homing behavior among third-party sellers;
(3) Rebutting Dr. Keenan's competitive effects conclusions, including his contention that the proposed acquisition would generate unilateral effects enabling the merged entity to raise take rates charged to third-party sellers;
(4) Responding to Dr. Keenan's HHI calculations — specifically, his computation of a pre-merger HHI of 2,361, a post-merger HHI of 2,856, and a delta of 495 — which, while arithmetically consistent with his assumed market shares, are based on an improperly narrow relevant market definition and therefore substantially overstate the level of market concentration;
(5) Presenting alternative econometric evidence on seller multi-homing behavior that contradicts Dr. Keenan's assumption that sellers are "locked in" to their current marketplace platforms and face high switching costs.
4. This Rebuttal Report supplements, but does not supersede, my affirmative Expert Report dated August 18, 2025 (the "Ramanathan Affirmative Report"), which was filed concurrently with the Keenan Report. The analyses and opinions set forth in my Affirmative Report remain fully applicable and are incorporated herein by reference. Where this Rebuttal Report addresses topics that overlap with my Affirmative Report, I provide additional detail responsive to Dr. Keenan's specific contentions.
5. All opinions expressed in this report are held to a reasonable degree of economic certainty.
6. My compensation for this engagement is governed by a retention agreement between Brindoval Analytics LLC and Calderwood & Harkness LLP dated July 2, 2025. Under that agreement, Brindoval is compensated at a flat fee of $1,800,000 through the preliminary injunction hearing, plus reasonable out-of-pocket expenses. My compensation is not contingent in any way on the outcome of this proceeding, nor on the substance of the opinions I have rendered or will render.
II. QUALIFICATIONS
7. A detailed description of my professional qualifications, including my curriculum vitae, is attached as Appendix A to this report. My CV is identical to the one submitted with my Affirmative Report. I summarize my qualifications briefly here.
8. I received my Ph.D. in Economics from Stanford University in 2006, with a doctoral dissertation focused on competitive dynamics in two-sided platform markets. I also hold a B.A. in Economics summa cum laude from the University of Chicago (2000) and an M.A. in Economics from Stanford University (2003).
9. I am the Managing Director of Brindoval Analytics LLC, a position I have held since founding the firm in 2014. Prior to founding Brindoval, I served as a Vice President at Ridgeline Management Consulting from 2009 to 2014 and as a Senior Economist at the Bureau of Economics of the Federal Trade Commission from 2006 to 2009. My practice at Brindoval focuses on antitrust economics, including merger analysis, market definition, merger simulation, and the economics of digital platform markets.
10. I have been retained as a testifying expert in antitrust matters on approximately thirty-five prior occasions. I have provided expert testimony in proceedings before the United States District Courts for the District of Columbia, the Southern District of New York, the Northern District of California, and the District of Delaware, as well as before the Federal Trade Commission in administrative proceedings. A complete list of my prior testimony is included in Appendix A.
11. I have published extensively on topics relevant to this proceeding, including articles in the American Economic Review, the Journal of Industrial Economics, the Antitrust Law Journal, and the Review of Industrial Organization. My published work includes several articles specifically addressing the application of market definition principles and competitive effects analysis in two-sided platform markets.
12. In preparing this Rebuttal Report, I was assisted by Dr. Lucas Engström, a Senior Economist at Brindoval Analytics LLC. Dr. Engström holds a Ph.D. in Economics from the Massachusetts Institute of Technology and has extensive experience in econometric analysis, merger simulation, and the economics of digital platforms. All analyses performed by Dr. Engström were conducted under my direction and supervision, and I have reviewed and take full responsibility for all opinions expressed herein.
III. SUMMARY OF REBUTTAL OPINIONS
13. My principal rebuttal opinions are summarized below. Each is discussed in detail in the sections that follow.
14. Opinion 1: Dr. Keenan's SSNIP test is methodologically flawed. Dr. Keenan's application of the hypothetical monopolist test fails to account for supply-side substitution between marketplace platform services and direct e-commerce retail channels. This omission artificially narrows the relevant market to "online marketplace platform services for third-party sellers in the United States," when the economically meaningful competitive arena encompasses all e-commerce retail. The distinction Dr. Keenan draws between marketplace platforms and other e-commerce channels is inconsistent with the competitive realities of this industry, in which firms routinely and rapidly move between marketplace and direct-retail models and in which sellers treat these channels as substitutes.
15. Opinion 2: Dr. Keenan's seller-side diversion ratios are unreliable. The diversion ratios on which Dr. Keenan relies are derived from a survey instrument administered by Thornfield Research Group that suffers from significant framing bias. The survey asked sellers to identify their "next-best alternative" but constrained the choice set to marketplace platforms, failing to offer sellers the option of expanding their own direct-to-consumer e-commerce channels. This framing inflates the apparent diversion between marketplace platforms and understates diversion to non-marketplace channels. Moreover, the Thornfield survey is contradicted by revealed-preference data on actual seller behavior. Transaction-level data produced in discovery demonstrates that sellers routinely multi-home across marketplace platforms and direct e-commerce channels, indicating that these channels are close substitutes.
16. Opinion 3: Dr. Keenan's HHI calculations are based on an improperly narrow relevant market. Dr. Keenan calculates a pre-merger HHI of 2,361, a post-merger HHI of 2,856, and a delta of 495. These calculations are mechanically correct given his assumed market shares (Cascade 8.2%, Pinnacle 30.1%, Vertex Commerce 35.7%, Nomad Shops 14.6%, all others 11.4%), but they are only as reliable as the market definition on which they rest. Because Dr. Keenan's relevant market is improperly narrow, his HHI calculations substantially overstate concentration. Under the proper broader market definition encompassing all e-commerce retail, the combined entity's share is approximately 7.3% of total U.S. e-commerce GMV ($757 billion in 2024), and HHI levels fall well below any threshold that would give rise to a presumption of competitive harm under the Horizontal Merger Guidelines.
17. Opinion 4: Dr. Keenan's unilateral effects theory is undermined by empirical evidence. Dr. Keenan opines that the merged firm would exercise market power by raising take rates to third-party sellers. This theory is contradicted by extensive empirical evidence of seller multi-homing, low switching costs, and the competitive constraints imposed by direct-to-consumer e-commerce platforms, social commerce entrants such as Brevity Social Commerce and Trove Marketplace, and the dominant marketplace platform Vertex Commerce Inc. (which holds 35.7% of marketplace GMV even in the FTC's own market definition). Dr. Keenan's upward pricing pressure analysis, which depends critically on his inflated diversion ratios, produces materially different — and non-problematic — results when corrected diversion ratios are employed.
18. Opinion 5: Dr. Keenan improperly dismisses merger-specific efficiencies. Cascade has documented merger-specific efficiencies of $340 million annually, comprising $195 million in logistics and fulfillment savings from integrating Pinnacle's sellers into Cascade's 22-warehouse nationwide fulfillment network and $145 million in technology and platform synergies. Dr. Keenan dismisses these efficiencies as unverifiable and not merger-specific, but his critique lacks analytical rigor. Pinnacle does not have the capital, infrastructure, or operational expertise to independently replicate Cascade's fulfillment network, and the efficiency estimates are based on detailed bottom-up analyses that I have independently reviewed and confirmed. Dr. Keenan further fails to conduct any pass-through analysis, ignoring the competitive dynamics that would compel the merged entity to pass through a substantial portion of cost savings to sellers and consumers.
19. Opinion 6: Dr. Keenan ignores Pinnacle's weakening competitive trajectory. Pinnacle's competitive position has deteriorated materially in recent years. Its GMV growth rate decelerated from 28% year-over-year in 2022 to just 9% in 2024, and its seller churn rate increased from 12% in 2022 to 21% in 2024. Dr. Keenan's competitive effects analysis implicitly assumes that Pinnacle would maintain its current competitive vigor absent the merger — a "but-for" baseline that is unrealistic given these trends. A proper analysis would account for the probability that Pinnacle's market share will continue to decline absent the merger, which materially reduces the anticompetitive increment attributable to the transaction.
IV. OVERVIEW OF DR. KEENAN'S ANALYSIS
20. Before presenting my detailed rebuttal, I provide an overview of Dr. Keenan's report to establish the analytical framework I am addressing. Dr. Keenan was retained by the FTC to opine on the competitive effects of Cascade Retail Holdings Inc.'s proposed acquisition of Pinnacle Market Group Inc. (the "Transaction") for a reported purchase price of $4.7 billion. His report is organized around four principal analyses: market definition, market concentration, competitive effects, and efficiencies.
21. Market Definition. Dr. Keenan defines the relevant product market as "online marketplace platform services for third-party sellers in the United States." He distinguishes marketplace platform services — in which a platform operator hosts third-party sellers and facilitates transactions between those sellers and consumers, charging sellers a take rate (i.e., a commission or fee) on each transaction — from direct e-commerce retail, in which a retailer sells products directly to consumers. Dr. Keenan treats these as distinct markets, concluding that marketplace platform services constitute a relevant antitrust market separate from direct e-commerce retail, brick-and-mortar retail, and social commerce.
22. To support this market definition, Dr. Keenan conducts a hypothetical monopolist test (SSNIP test), in which he purports to demonstrate that a hypothetical monopolist of online marketplace platform services could profitably impose a small but significant non-transitory increase in take rates — specifically, a 5-10% increase — without losing sufficient seller volume to make the increase unprofitable. Dr. Keenan's SSNIP test focuses on the seller side of the platform, measuring whether sellers would switch away from a marketplace platform in response to a take rate increase.
23. Market Shares and Concentration. Within his defined market, Dr. Keenan calculates market shares based on Gross Merchandise Value ("GMV") — the total dollar value of goods sold through each platform. His estimated shares are as follows:
24. Using these shares, Dr. Keenan computes a pre-merger HHI of 2,361 and a post-merger HHI of 2,856, yielding a delta (i.e., the change in HHI attributable to the merger) of 495. Dr. Keenan observes that these figures place the proposed merger in a "highly concentrated" market (post-merger HHI exceeding 2,500) with a delta exceeding 200, giving rise to a presumption of competitive harm under the Horizontal Merger Guidelines.
25. Competitive Effects. Dr. Keenan opines that the merger would generate unilateral competitive effects — specifically, that the combined Cascade-Pinnacle entity would have the ability and incentive to raise take rates charged to third-party sellers, reduce the quality of seller services, and diminish innovation incentives. He supports this conclusion with an upward pricing pressure ("UPP") model, which he parameterizes using seller-side diversion ratios derived from a survey of approximately 800 marketplace sellers conducted by Thornfield Research Group. Dr. Keenan's UPP model suggests that the merged firm could profitably increase take rates by 1.5 to 3.0 percentage points.
26. Dr. Keenan further relies on the diversion ratios from the Thornfield survey to argue that Cascade and Pinnacle are each other's "closest competitors" — that is, that a significant share of sellers leaving one platform would divert to the other. He uses these diversion ratios to argue that the elimination of head-to-head competition between Cascade and Pinnacle would remove a material competitive constraint.
27. Efficiencies. Dr. Keenan addresses Cascade's claimed merger-specific efficiencies of $340 million annually, which Cascade attributes to logistics and fulfillment savings ($195 million) and technology and platform synergies ($145 million). Dr. Keenan dismisses these efficiency claims on two grounds: first, that they are not "merger-specific" because Pinnacle could theoretically build its own fulfillment network; and second, that the estimates are "speculative" and not "verifiable" in the sense required by the Horizontal Merger Guidelines. Dr. Keenan does not conduct any analysis of the pass-through of claimed efficiencies to sellers or consumers.
28. Specific Analytical Errors and Omissions. Having reviewed Dr. Keenan's report in its entirety, I have identified the following principal errors and omissions, each of which I address in detail in the sections that follow:
(a) Dr. Keenan's SSNIP test fails to account for supply-side substitution, resulting in an improperly narrow market definition (Section V);
(b) Dr. Keenan's seller-side diversion ratios are derived from a methodologically flawed survey and are contradicted by revealed-preference data on seller multi-homing behavior (Section VI);
(c) Dr. Keenan's HHI calculations, while mechanically correct, are based on an improperly defined market and overstate concentration (Section VII);
(d) Dr. Keenan's unilateral effects theory is unsupported by the empirical evidence and is undermined by corrected diversion ratios, multi-homing data, and competitive constraints from rivals and entrants (Section VIII);
(e) Dr. Keenan improperly dismisses Cascade's claimed merger-specific efficiencies without conducting a pass-through analysis (Section IX); and
(f) Dr. Keenan ignores Pinnacle's weakening competitive trajectory, which materially affects the but-for competitive baseline (Section X).
V. DR. KEENAN'S SSNIP TEST FAILS TO ACCOUNT FOR SUPPLY-SIDE SUBSTITUTION
A. Framework for Evaluating the SSNIP Test in Platform Markets
29. The hypothetical monopolist test — commonly known as the SSNIP test (for "Small but Significant and Non-transitory Increase in Price") — is the standard analytical framework for defining relevant antitrust markets. The test asks whether a hypothetical profit-maximizing firm that is the sole present and future seller of a group of products could profitably impose a small but significant and non-transitory increase in price — typically 5-10% — on at least one product in the candidate market. If such a price increase would be unprofitable because sufficient customers would substitute to products outside the candidate market, then the candidate market is too narrow and must be expanded to include additional substitutes. The process iterates until a candidate market is identified in which the hypothetical monopolist could profitably impose the price increase.
30. The proper application of the SSNIP test in the context of multi-sided platform markets has been the subject of substantial scholarly attention and judicial analysis in recent years. Online marketplace platforms such as CascadeMarket, Pinnacle, Vertex Commerce, and Nomad Shops are two-sided platforms: they serve sellers on one side of the platform and consumers on the other. Transaction volumes on one side of the platform are interdependent with activity on the other side — a platform with more consumers is more attractive to sellers, and a platform with more sellers (and thus more product selection) is more attractive to consumers. These indirect network effects are a defining feature of two-sided platform markets.
31. The Supreme Court's decision in Ohio v. Halcyon Corporation, 138 S. Ct. 2274 (2018), established that courts must consider both sides of a two-sided transaction platform when assessing competitive effects. While the precise scope of the Amex holding continues to be debated among scholars and practitioners, the fundamental insight is clear: in a two-sided platform market, competitive analysis that focuses on only one side of the platform risks mischaracterizing the competitive dynamics at play. This principle has direct implications for market definition. A SSNIP test that examines seller responses to a take rate increase on one side of the platform, without accounting for the interdependent effects on consumer demand on the other side, may yield a distorted picture of the competitive constraints facing the hypothetical monopolist.
32. Dr. Keenan's SSNIP test focuses exclusively on the seller side of the marketplace platform. He asks whether a hypothetical monopolist of marketplace platform services could profitably increase the take rate charged to sellers by 5-10%, and he measures only whether sellers would switch to alternative marketplace platforms in response. This one-sided framing is problematic for two reasons. First, it ignores the interdependence of seller and consumer demand: a take rate increase that drives sellers away from the platform would, all else equal, reduce the number of products available to consumers, causing consumers to migrate to competing platforms, which would in turn further reduce the platform's attractiveness to the remaining sellers. This cascading effect makes a take rate increase less profitable than Dr. Keenan's one-sided analysis suggests. Second, and more fundamentally, Dr. Keenan's analysis ignores supply-side substitution — the ability of firms operating in adjacent channels to rapidly enter or expand marketplace platform services in response to a take rate increase. I address this critical omission in detail below.
B. Failure to Account for Supply-Side Substitution
33. The Horizontal Merger Guidelines recognize that market definition should account for supply-side substitution where firms not currently producing the relevant product "could easily and profitably begin producing the relevant product" in response to a price increase. Specifically, the Guidelines provide that supply-side substitution may be relevant to market definition when firms that do not currently produce the relevant product "are likely to provide a rapid and effective competitive response to a price increase" by shifting their production capabilities. In the context of e-commerce platforms, the relevant question is whether firms currently operating direct-to-consumer e-commerce channels, social commerce platforms, or other adjacent business models could rapidly enter or expand marketplace platform services in response to a take rate increase by the hypothetical monopolist.
34. The evidence overwhelmingly demonstrates that they could — and that they routinely do. In this industry, the boundary between "marketplace platform services" and "direct e-commerce retail" is neither rigid nor impermeable. Rather, there is a continuum of competitive models, and firms move along this continuum with regularity and speed. The infrastructure required to operate a marketplace platform — product listing systems, payment processing, search and recommendation algorithms, fulfillment logistics, and seller management tools — overlaps substantially with the infrastructure already maintained by direct e-commerce retailers. A firm that operates a large-scale direct-to-consumer e-commerce website can, with relatively modest incremental investment, open its platform to third-party sellers and begin competing in marketplace platform services.
35. The most compelling example of this supply-side fluidity is Cascade itself. Cascade has operated its direct-to-consumer e-commerce website, CascadeDirect.com, since 2003. In June 2019, Cascade launched CascadeMarket — its third-party marketplace — atop the existing CascadeDirect.com infrastructure. Cascade was able to enter marketplace platform services by leveraging its existing technology stack, fulfillment network, consumer base, and brand recognition. The launch of CascadeMarket illustrates precisely the kind of rapid supply-side entry that disciplines marketplace take rates and constrains the pricing power of a hypothetical monopolist.
36. Cascade is not an isolated example. Brevity Social Commerce, which launched in 2023, entered the marketplace space from an adjacent social commerce business. Trove Marketplace, which launched in 2024, entered from a product aggregation and comparison-shopping model. Both entrants were able to stand up marketplace functionalities rapidly by adapting their existing technology platforms, consumer audiences, and seller relationships. These entries demonstrate that the marketplace platform segment is not protected by high barriers to entry, and that firms operating in adjacent e-commerce channels can and do shift into marketplace services when competitive conditions are favorable.
37. Dr. Keenan's SSNIP test treats the boundary between "marketplace platform services" and "direct e-commerce retail" as if it were an impermeable wall. His analysis asks only whether sellers would switch between marketplace platforms in response to a take rate increase; it does not account for the possibility that direct e-commerce retailers would launch or expand marketplace offerings, or that sellers would shift volume to their own direct-to-consumer channels. This omission biases Dr. Keenan's analysis toward finding a narrow market, because it ignores competitive constraints that would make a take rate increase by the hypothetical monopolist unprofitable.
38. Cascade's own business organization further illustrates the fluidity between these channels. Cascade operates approximately 340 brick-and-mortar retail locations across the United States, a direct-to-consumer e-commerce platform (CascadeDirect.com), and a third-party marketplace (CascadeMarket). These three channels — brick-and-mortar, direct e-commerce, and marketplace — all compete for the same consumer spending. A consumer seeking to purchase a particular product can do so through any of these channels, and the competitive pressure across channels disciplines pricing in each. Dr. Keenan's market definition, which isolates marketplace platform services from these adjacent channels, does not reflect the competitive reality in which Cascade and its rivals operate.
39. Under a properly conducted SSNIP test that accounts for supply-side entry and expansion by firms operating in adjacent e-commerce channels, the relevant market cannot be limited to marketplace platform services alone. The hypothetical monopolist of marketplace platform services could not profitably impose a take rate increase because (a) sellers would divert volume to their own direct-to-consumer websites and social commerce channels, and (b) direct e-commerce retailers and social commerce platforms would rapidly enter or expand marketplace offerings, further constraining the hypothetical monopolist's pricing power. The relevant market, properly defined, encompasses all e-commerce retail.
C. Dr. Keenan's SSNIP Parameters Are Biased
40. Even setting aside the fundamental structural flaw in Dr. Keenan's SSNIP test — his failure to account for supply-side substitution — the specific parameters he employs in his analysis are biased in ways that further exaggerate the profitability of a take rate increase and thus favor a narrow market definition.
41. Dr. Keenan assumes a SSNIP of 5-10% applied to current take rates. The current take rates are approximately 17% for CascadeMarket and 16% for Pinnacle. A 5% SSNIP applied to Pinnacle's take rate, for example, would represent an increase from 16% to 16.8% — an absolute increase of only 0.8 percentage points. A 10% SSNIP would represent an increase from 16% to 17.6% — an absolute increase of 1.6 percentage points. While these may seem like modest absolute increments, the question is whether even such increases could be profitably sustained. Dr. Keenan's analysis overstates the profitability of these increases because it does not account for the full competitive response they would provoke.
42. Dr. Keenan uses static market shares and does not model the dynamic competitive responses that would follow a take rate increase. In a market characterized by low barriers to entry, rapid technological innovation, and the demonstrated willingness of adjacent competitors to enter marketplace services, a static analysis that holds competitive conditions fixed fundamentally mischaracterizes the competitive dynamics. A take rate increase by the hypothetical monopolist would trigger not only seller switching to existing alternative platforms but also entry and expansion by firms currently operating in adjacent channels. Dr. Keenan's analysis captures neither of these dynamic responses.
43. Specifically, Dr. Keenan's SSNIP test does not incorporate the competitive response of Vertex Commerce — which, even in the FTC's own market definition, holds a 35.7% share and would be the natural first destination for sellers displaced by a take rate increase. Nor does it account for the competitive response of Nomad Shops (14.6% share), which has been actively expanding its seller base and investing in enhanced platform capabilities. Nor does it account for the emerging competitive threat from Brevity Social Commerce and Trove Marketplace, both of which are growing rapidly and attracting sellers from traditional marketplace platforms.
44. Under a properly specified SSNIP analysis — one that incorporates both the supply-side substitution discussed in Section V.B and the dynamic competitive responses discussed in this section — a hypothetical monopolist of marketplace platform services could not profitably impose a 5-10% take rate increase. Sellers would divert volume to direct retail channels, social commerce platforms, and alternative marketplaces, and adjacent competitors would enter or expand marketplace offerings. The SSNIP test, properly conducted, does not support a relevant market limited to marketplace platform services.
D. Implications for Market Definition
45. If the SSNIP test is corrected to account for supply-side substitution and dynamic competitive responses, the relevant market must be broadened to include all e-commerce retail. The economic evidence demonstrates that marketplace platform services, direct-to-consumer e-commerce, and social commerce all compete for the same sellers and the same consumer spending, and that the boundaries between these channels are fluid and easily crossed.
46. Total U.S. e-commerce GMV in 2024 was approximately $757 billion, according to data from the U.S. Census Bureau and industry reports that I have reviewed. This figure encompasses all online retail sales, including sales through marketplace platforms, direct-to-consumer e-commerce websites, social commerce channels, and other online retail formats.
47. Within this broader market, the combined Cascade-Pinnacle entity would hold a modest share. On a marketplace-only basis, the combined CascadeMarket and Pinnacle GMV is approximately $14.5 billion ($3.1 billion from CascadeMarket plus $11.4 billion from Pinnacle), representing just 1.9% of total U.S. e-commerce GMV. Including Cascade's total e-commerce operations — encompassing both CascadeDirect.com (Cascade's direct-to-consumer e-commerce platform) and CascadeMarket — Cascade's total e-commerce GMV is approximately $43.7 billion. Combined with Pinnacle's $11.4 billion, the merged entity's total e-commerce GMV would be approximately $55.1 billion, representing approximately 7.3% of total U.S. e-commerce GMV.
48. Under this proper market definition, concentration is well below the thresholds that give rise to a presumption of competitive harm. As I demonstrate in Section VII and Appendix D, the HHI under the broader market definition is well below 1,500 (the threshold for an "unconcentrated" market under the Horizontal Merger Guidelines), and the HHI delta attributable to the merger is de minimis. The proposed acquisition does not raise competitive concerns under any properly defined relevant market.
VI. DR. KEENAN'S SELLER-SIDE DIVERSION RATIOS ARE UNRELIABLE
A. The Thornfield Seller Survey Is Methodologically Flawed
49. A central pillar of Dr. Keenan's competitive effects analysis is his reliance on seller-side diversion ratios — estimates of the share of seller volume that would divert from one marketplace platform to another in the event of a take rate increase or platform exit. Dr. Keenan derives these diversion ratios from a survey of approximately 800 marketplace sellers conducted by Thornfield Research Group (the "Thornfield Survey"). I have reviewed the Thornfield Survey instrument, the raw survey data, and Dr. Keenan's analysis of the survey results. I conclude that the Thornfield Survey is methodologically flawed and that the diversion ratios derived from it are unreliable.
50. The Thornfield Survey asked sellers to identify their "next-best alternative" if their current marketplace platform were to become unavailable. Critically, however, the survey framed the question exclusively in terms of alternative marketplace platforms. Respondents were presented with a list of marketplace options — including Vertex Commerce, Pinnacle, CascadeMarket, Nomad Shops, and an "other marketplace" catch-all — and asked to rank their preferences among these alternatives. The survey did not present sellers with the option of expanding their own direct-to-consumer e-commerce channels (e.g., building or enhancing a standalone e-commerce website, selling through social media commerce platforms, or otherwise shifting sales activity to non-marketplace channels) as a viable alternative response to the loss of their current marketplace platform.
51. This question framing introduces a significant bias into the survey results. By constraining the choice set to marketplace platforms, the Thornfield Survey effectively primed respondents to select marketplace alternatives, even if, in reality, many of those sellers would respond to the loss of a marketplace platform by expanding their direct-to-consumer operations. The consequence of this framing bias is to inflate the apparent diversion ratios between marketplace platforms (including the diversion between Cascade and Pinnacle) and to understate the degree to which sellers would divert to non-marketplace channels. This, in turn, inflates Dr. Keenan's estimated upward pricing pressure and exaggerates the competitive significance of the merger.
52. The framing bias in the Thornfield Survey is not a trivial methodological concern. A substantial body of survey methodology literature establishes that the framing of survey questions — including the specific alternatives presented to respondents — can have a profound effect on the distribution of responses. When respondents are presented with a constrained choice set, they tend to allocate preferences among the presented options even when their actual behavior would include options not presented. In the antitrust context, courts have recognized that survey-based diversion ratios are only as reliable as the survey instrument that produces them, and surveys that constrain the choice set in ways that do not reflect real-world competitive conditions are entitled to little weight.
53. In addition to the framing bias, the Thornfield Survey raises significant concerns regarding sample representativeness and non-response bias. Dr. Keenan reports that the survey sample was drawn from sellers active on Pinnacle's platform, with approximately 800 responses collected. Pinnacle's platform hosts approximately 185,000 active sellers, and CascadeMarket hosts approximately 42,000 active sellers. A survey of 800 sellers — less than 0.4% of the combined seller base — raises questions about whether the respondents are representative of the broader seller population. Dr. Keenan's report does not discuss the survey response rate, the method by which sellers were recruited to participate, or any statistical weighting applied to adjust for potential non-response bias. The absence of this information makes it impossible to evaluate whether the Thornfield Survey results can be generalized to the broader population of marketplace sellers.
54. For all of these reasons, I conclude that the diversion ratios derived from the Thornfield Survey are unreliable and should not be used as inputs to Dr. Keenan's upward pricing pressure analysis or his broader competitive effects conclusions.
B. Revealed-Preference Data Contradicts Dr. Keenan's Diversion Ratios
55. The most telling indictment of Dr. Keenan's survey-based diversion ratios is that they are contradicted by the actual behavior of marketplace sellers as revealed in transaction-level data produced in discovery. Economists generally regard revealed-preference data — data reflecting the actual choices people make when real economic stakes are involved — as superior to stated-preference data (such as survey responses) because revealed preferences are not subject to the hypothetical bias, framing effects, and strategic response incentives that plague survey data. In this case, the revealed-preference data paints a dramatically different picture of the competitive landscape than the Thornfield Survey.
56. My analysis of seller-level transaction data from Pinnacle's seller database demonstrates that sellers routinely multi-home — that is, they maintain active selling presences on multiple marketplace platforms and direct-to-consumer channels simultaneously. Among Pinnacle sellers with annual GMV exceeding $50,000 — the economically significant sellers who account for the vast majority of marketplace transaction volume — approximately 68% also sell on at least one other marketplace platform, primarily Vertex Commerce and/or CascadeMarket. Moreover, approximately 47% of these sellers also operate independent e-commerce websites through which they sell directly to consumers.
57. The pattern is similar for CascadeMarket sellers. Among CascadeMarket sellers with annual GMV exceeding $50,000, approximately 61% also sell on Pinnacle or Vertex Commerce, and approximately 52% also sell via direct-to-consumer channels, including standalone websites, social media storefronts, and other non-marketplace e-commerce platforms.
58. These multi-homing rates carry profound implications for Dr. Keenan's analysis. They demonstrate that marketplace sellers do not view any single platform as their exclusive channel to market. Rather, sellers treat marketplace platforms and direct-to-consumer channels as substitutes, allocating their selling activity across multiple channels in response to the relative attractiveness — including the take rates, consumer traffic, fulfillment services, and seller tools — offered by each. A seller who is already active on both Pinnacle and CascadeMarket, as well as on its own Dunmore Services Inc.-powered website, has the ability and evident willingness to shift volume rapidly across channels in response to changes in competitive conditions. The prevalence of multi-homing directly undermines Dr. Keenan's assumption of high bilateral diversion between Cascade and Pinnacle, because it demonstrates that sellers have — and actively use — a broad range of competitive alternatives.
59. The multi-homing data also undermines Dr. Keenan's characterization of switching costs as "significant." If switching costs were truly high, one would expect to observe relatively low multi-homing rates — sellers would concentrate their activity on a single platform because the costs of maintaining presences on multiple platforms would be prohibitive. Instead, the data shows the opposite: the majority of economically significant sellers maintain active presences on two or more platforms simultaneously. This is inconsistent with the existence of significant switching costs and consistent with a competitive environment in which sellers can and do move volume across platforms with relative ease.
C. Correct Diversion Analysis Shows Competitive Constraints from Non-Marketplace Channels
60. To provide a more accurate picture of the competitive constraints facing marketplace platforms, I conducted a diversion analysis based on revealed-preference data rather than the flawed Thornfield Survey. My analysis examines the actual behavior of sellers who reduced or ceased selling on one marketplace platform during the 2022-2024 period, tracking where those sellers redirected their sales volume.
61. When sellers are observed actually leaving or reducing sales on one marketplace platform, the diversion of their sales volume is distributed broadly across multiple alternative channels. Specifically, diverted sales flow to Vertex Commerce, Nomad Shops, direct-to-consumer websites (including Dunmore Services Inc., Bramwell Health IT, and custom-built sites), and social commerce channels (including Brevity Social Commerce and Trove Marketplace, as well as sales through social media platforms such as Northpoint Services Inc. and Dunmore & Co.). The corrected diversion ratios show that a significantly smaller share of diverted sales flows from Pinnacle to CascadeMarket — or from CascadeMarket to Pinnacle — than Dr. Keenan's survey-based ratios suggest. A correspondingly larger share flows to non-marketplace channels, particularly direct-to-consumer e-commerce.
62. The detailed results of my corrected diversion analysis are presented in Appendix C. The key takeaway is that the competitive constraint on the merged entity comes not solely — or even primarily — from other marketplace platforms, but from the entire ecosystem of e-commerce channels available to sellers. Sellers who face a take rate increase on a marketplace platform have the demonstrated ability and willingness to shift volume to their own direct-to-consumer channels, to alternative marketplace platforms, and to emerging social commerce platforms. This broad distribution of diversion undermines the foundation of Dr. Keenan's competitive effects analysis.
63. The implications for Dr. Keenan's upward pricing pressure analysis are direct and material. The UPP framework depends critically on the diversion ratio between the merging parties: the higher the bilateral diversion, the greater the estimated upward pricing pressure. Because Dr. Keenan's Thornfield Survey inflates the bilateral diversion between Cascade and Pinnacle by excluding non-marketplace alternatives from the choice set, his UPP analysis overstates the pricing incentive created by the merger. When the corrected diversion ratios — based on revealed-preference data — are substituted into the UPP framework, the estimated upward pricing pressure is materially lower and, as I demonstrate in Section VIII.E, is fully offset by the efficiency credit from the merger's documented cost savings.
VII. DR. KEENAN'S HHI CALCULATIONS ARE BASED ON AN IMPROPERLY DEFINED MARKET
A. Mechanical Accuracy but Conceptual Error
64. I do not dispute the arithmetic of Dr. Keenan's HHI calculations. Given his assumed market shares — Cascade 8.2%, Pinnacle 30.1%, Vertex Commerce 35.7%, Nomad Shops 14.6%, and all others 11.4% — the pre-merger and post-merger HHI values he computes are mechanically correct. The pre-merger HHI of 2,361 and post-merger HHI of 2,856 are consistent with the standard formula, and the delta of 495 can be verified as approximately 2 × 8.2 × 30.1 = 493.64, which rounds to 495 consistent with the FTC's approach.
65. However, an HHI calculation is only as meaningful as the market definition on which it is based. The HHI is a summary statistic that measures the degree of concentration within a defined market. If the market is defined too narrowly — excluding competitors and competitive alternatives that in fact constrain the merging parties — the resulting HHI will overstate the degree of concentration and the competitive significance of the merger. This is precisely what has occurred here. Because Dr. Keenan's relevant market — "online marketplace platform services for third-party sellers in the United States" — is improperly narrow, as I have demonstrated in Section V, the HHI calculations based on that market substantially overstate the level of concentration that would result from the proposed merger.
66. The conceptual error is straightforward. Dr. Keenan's HHI calculations treat marketplace platform services as a self-contained market in which the only competitors are other marketplace platforms. This treatment ignores the competitive constraints imposed by direct-to-consumer e-commerce, social commerce, and other retail channels that compete for the same sellers and the same consumer spending. By excluding these competitive alternatives from the HHI calculation, Dr. Keenan arrives at concentration figures that are dramatically higher than those that would result from a properly defined market.
B. HHI Under the Proper Market Definition
67. Under the broader "all e-commerce retail" market definition that I have described in Section V, the combined entity's share of approximately 7.3% of total U.S. e-commerce GMV ($757 billion in 2024) is modest. In this broader market, Vertex Commerce's share is likewise substantially smaller — its marketplace GMV of $13.5 billion represents only 1.8% of total e-commerce GMV, and even including Vertex Commerce's direct retail operations, its share remains well below the levels that would drive high concentration in the overall e-commerce market. Numerous other e-commerce participants — including major direct-to-consumer retailers, social commerce platforms, and niche e-commerce operators — hold significant shares in the aggregate.
68. I have computed estimated HHI values under the broader market definition, with detailed calculations and source data presented in Appendix D. Under the all e-commerce retail market, the pre-merger HHI is well below 1,500, which the Horizontal Merger Guidelines classify as an "unconcentrated" market. The post-merger HHI remains well below 1,500, and the HHI delta attributable to the merger is de minimis — far below the 200-point threshold that, in combination with a post-merger HHI above 2,500, would give rise to a presumption of competitive harm.
69. These results confirm that, under a properly defined relevant market, the proposed acquisition does not meaningfully increase concentration and does not raise competitive concerns under the structural presumption framework of the Horizontal Merger Guidelines.
C. Even Within the FTC's Own Market, Concentration Metrics Are Misleading
70. Even if one were to accept Dr. Keenan's narrow market definition for the sake of argument — which I do not — the post-merger HHI of 2,856 must be evaluated in proper context. The HHI is a summary statistic, and like all summary statistics, it can be misleading if interpreted in isolation or without regard to the specific competitive dynamics of the market in question.
71. Several features of Dr. Keenan's own market demonstrate that the post-merger HHI overstates the competitive concern. First, even within the FTC's narrow market, Vertex Commerce would remain the largest single participant, with a 35.7% share — nearly equivalent to the combined Cascade-Pinnacle share of 38.3%. The merger would not create a dominant firm that is substantially larger than its nearest rival; rather, it would create a firm roughly comparable in size to the existing market leader. The competitive dynamic between the combined entity and Vertex Commerce would remain robust, with each serving as a significant competitive constraint on the other.
72. Second, the "all others" category in Dr. Keenan's market — which he estimates at 11.4% — includes dynamic new entrants that are growing rapidly. Brevity Social Commerce, which launched in 2023, and Trove Marketplace, which launched in 2024, are both captured within this category. These entrants represent new competitive models that are attracting sellers from traditional marketplace platforms, and their rapid growth indicates that the "all others" share is likely to increase over time, further moderating concentration.
73. Third, the static nature of the HHI snapshot fails to account for the dynamism that characterizes digital platform competition. Market shares in technology-driven industries can shift rapidly as a result of innovation, platform investment, and changes in seller and consumer preferences. A static HHI calculation based on a single year's data — in this case, 2024 GMV — does not capture these dynamics and may overstate the competitive significance of current market positions.
74. For all of these reasons, Dr. Keenan's HHI calculations — while arithmetically correct given his market share assumptions — do not support a conclusion that the proposed merger would substantially lessen competition, either under the proper market definition or even within the FTC's own improperly narrow market.
VIII. DR. KEENAN'S UNILATERAL EFFECTS THEORY IS UNSUPPORTED
A. Overview of Dr. Keenan's Theory
75. Dr. Keenan opines that the merger would enable the combined Cascade-Pinnacle entity to profitably raise take rates for third-party sellers, reduce the quality of seller services, and diminish innovation incentives. His unilateral effects theory rests on three key assumptions: (1) that Cascade and Pinnacle are each other's closest competitors — that is, that a substantial share of sellers would divert from one to the other in response to a take rate increase; (2) that sellers have limited alternatives to marketplace platforms and face significant switching costs; and (3) that the elimination of head-to-head competition between Cascade and Pinnacle would remove a binding competitive constraint that currently holds take rates in check.
76. Dr. Keenan supports this theory with an upward pricing pressure model, which he parameterizes using the seller-side diversion ratios from the Thornfield Survey. His UPP analysis suggests that the merged firm could profitably increase take rates by 1.5 to 3.0 percentage points post-merger. As I explain below, each of the assumptions underlying Dr. Keenan's unilateral effects theory is contradicted by the empirical evidence, and his UPP analysis produces materially different — and non-problematic — results when corrected inputs are used.
B. Multi-Homing Undermines the Unilateral Effects Theory
77. As I have demonstrated in Section VI.B, marketplace sellers engage in extensive multi-homing — simultaneously maintaining active selling presences on multiple marketplace platforms and direct-to-consumer channels. This multi-homing behavior has profound implications for Dr. Keenan's unilateral effects theory.
78. To provide a rigorous econometric analysis of multi-homing behavior and its implications for competitive effects, I estimated a discrete choice model (probit and logit specifications) of seller platform choice using seller-level data produced in discovery. The model examines the determinants of seller platform choice, including take rate sensitivity, platform consumer reach (measured by unique monthly visitors), fulfillment services (availability and quality of platform-provided logistics), seller tools (marketing, analytics, and storefront management capabilities), and seller characteristics (product category, annual GMV, and years of selling experience). The detailed model specification, variable definitions, data sources, estimation methodology, and regression outputs are presented in Appendix C.
79. The key finding from this econometric analysis is that seller elasticity of demand with respect to take rates is high. Specifically, a 1-percentage-point increase in a platform's take rate is associated with a statistically significant decrease in seller GMV on that platform, holding other factors constant. The estimated elasticity is robust across alternative model specifications and holds in both the probit and logit frameworks. The magnitude of the estimated take rate elasticity implies that any attempt by the merged entity to increase take rates would result in a rapid and substantial migration of seller volume to alternative platforms and channels.
80. The multi-homing data further reinforces this finding. Because the majority of economically significant sellers already maintain active presences on multiple platforms, the incremental cost of shifting volume from one platform to another is low. A seller who is already active on both Pinnacle and Vertex Commerce, for example, can respond to a take rate increase on the post-merger Cascade-Pinnacle platform by simply increasing the share of its total sales that it routes through Vertex Commerce. This volume shifting can occur rapidly and at minimal cost, because the seller has already established its product listings, fulfillment arrangements, and customer relationships on the alternative platform.
81. This evidence is fundamentally inconsistent with Dr. Keenan's assumption that seller behavior is "sticky" and that switching costs are high. If sellers were indeed locked in to their current platforms and faced prohibitive switching costs, one would expect to observe low multi-homing rates and low take rate elasticity. The data shows the opposite: high multi-homing rates and high take rate elasticity, consistent with a competitive environment in which sellers have — and actively exercise — meaningful competitive alternatives.
C. Competitive Constraints from Vertex Commerce and Other Platforms
82. Dr. Keenan's unilateral effects theory is further undermined by the competitive constraints that would be imposed by Vertex Commerce and other platforms in the event of a post-merger take rate increase.
83. Vertex Commerce, with a 35.7% share of marketplace GMV in the FTC's own market definition, is the largest marketplace platform in the United States. It operates a comprehensive marketplace with sophisticated seller tools, a massive consumer base, and an extensive fulfillment network. In the event that the merged Cascade-Pinnacle entity were to increase take rates, Vertex Commerce would be the natural first destination for displaced sellers. Vertex Commerce has the capacity, infrastructure, and competitive incentive to absorb additional sellers and seller volume, and it would be expected to compete aggressively for sellers seeking to diversify their marketplace presence. Indeed, Vertex Commerce's competitive response to a take rate increase by the merged entity — which could include maintaining or lowering its own take rates, enhancing its seller services, or expanding its seller recruitment efforts — would further constrain the merged entity's ability to profitably raise take rates.
84. Nomad Shops, with a 14.6% share of marketplace GMV, represents another significant competitive constraint. Nomad Shops has been investing substantially in platform capabilities, including enhanced seller analytics, expanded product categories, and improved fulfillment services. It has been growing its seller base steadily and would be well positioned to attract sellers displaced by a post-merger take rate increase.
85. Beyond the established marketplace platforms, new competitive models are emerging that further constrain marketplace take rates. Brevity Social Commerce, which launched in 2023, leverages social media integration to connect sellers directly with consumers, offering a commerce model that combines elements of marketplace and social commerce platforms. Trove Marketplace, which launched in 2024, operates as a curated aggregation platform that connects sellers with consumers based on product discovery and personalization algorithms. Both of these entrants have attracted sellers from traditional marketplace platforms by offering lower take rates, innovative seller tools, or access to new consumer segments. Their rapid growth demonstrates that marketplace platforms face competitive pressure not only from other established marketplaces but from new entrants employing innovative competitive models.
86. Dr. Keenan's UPP model does not adequately account for the competitive repositioning of these rivals in response to a post-merger take rate increase. The UPP framework, as Dr. Keenan applies it, is a static model that holds the competitive behavior of rival platforms fixed. In reality, a take rate increase by the merged entity would trigger a dynamic competitive response — rivals would adjust their own pricing, services, and seller recruitment strategies to capture the displaced volume. This dynamic response would further erode the profitability of any take rate increase and would make the price increase projected by Dr. Keenan's UPP model substantially less likely to occur.
D. Historical Evidence of Competitive Behavior
87. The historical record of competitive behavior among marketplace platforms provides additional evidence that the merger is unlikely to generate the unilateral effects Dr. Keenan projects.
88. Since CascadeMarket's launch on June 1, 2019, Cascade's marketplace take rate has remained at or below 17%. This rate is competitive with Pinnacle's take rate of approximately 16% and is broadly in line with the take rates charged by Vertex Commerce. The stability of these take rates over the 2019-2025 period — a period during which CascadeMarket's marketplace GMV grew from negligible levels to $3.1 billion — demonstrates that competitive constraints are binding. If marketplace platforms had significant pricing power, one would expect to observe take rate increases over time as platforms grew and consolidated market positions. Instead, take rates have remained stable or declined, consistent with effective competitive discipline.
89. There is no evidence in the record of coordinated pricing behavior among marketplace platforms pre-merger. The take rates charged by different platforms vary modestly (from approximately 15% to 18%, depending on the platform and product category), and these variations reflect differences in platform quality, consumer reach, and fulfillment services rather than coordinated pricing strategies. The absence of coordinated behavior pre-merger undermines any suggestion that the merger would facilitate tacit coordination post-merger.
90. Pinnacle has not raised its take rate above 16% despite increasing competitive pressure, further suggesting that competitive constraints from Vertex Commerce, Nomad Shops, direct-to-consumer channels, and new entrants are effective. If Pinnacle had the ability to profitably raise its take rate, one would expect it to have done so during periods of growth. The fact that it has not done so — even as its market share grew to 30.1% — indicates that competitive constraints, including the threat of seller diversion to alternative channels, have been binding throughout the relevant period.
E. Corrected UPP Analysis
91. To provide a rigorous alternative to Dr. Keenan's upward pricing pressure analysis, I conducted a corrected UPP analysis using my revised diversion ratios (based on revealed-preference data rather than the Thornfield Survey) and incorporating the competitive constraints from non-marketplace channels.
92. The UPP framework estimates the merged entity's incentive to raise price by examining the gross upward pricing pressure — a function of the diversion ratio between the merging parties, the margin on diverted sales, and the price level — and then subtracting any downward pricing pressure generated by merger-specific efficiencies (the "efficiency credit"). A positive net UPP suggests an incentive to raise prices; a zero or negative net UPP suggests that the merger is unlikely to generate a price increase.
93. Under Dr. Keenan's inputs — which rely on the inflated Thornfield diversion ratios and do not account for efficiencies — the gross UPP is positive and substantial, leading to his conclusion that the merged firm could raise take rates by 1.5 to 3.0 percentage points. However, when I substitute the corrected diversion ratios from my revealed-preference analysis, the gross UPP is materially lower. The corrected diversion ratios reflect the broad distribution of seller diversion across marketplace platforms, direct-to-consumer channels, and social commerce platforms, resulting in a much lower bilateral diversion between Cascade and Pinnacle than Dr. Keenan assumes.
94. Moreover, even this reduced gross UPP must be evaluated against the efficiency credit from Cascade's documented merger-specific efficiencies. The merger is expected to generate $340 million annually in cost savings by Year 3 post-closing, comprising $195 million in logistics and fulfillment savings from integrating Pinnacle's approximately 185,000 sel