Apple's Expert Written Testimony - Epic Games v. Apple

Case 4:20-cv-05640-YGR Document 490 Filed 04/27/21 Page 1 of 2
1 THEODORE J. BOUTROUS JR., SBN 132099 MARK A. PERRY, SBN 212532

tboutrous@gibsondunn.com mperry@gibsondunn.com
2 RICHARD J. DOREN, SBN 124666 CYNTHIA E. RICHMAN (D.C. Bar No.
rdoren@gibsondunn.com 492089; pro hac vice)
3 DANIEL G. SWANSON, SBN 116556 crichman@gibsondunn.com
dswanson@gibsondunn.com GIBSON, DUNN & CRUTCHER LLP
4 JAY P. SRINIVASAN, SBN 181471 1050 Connecticut Avenue, N.W.
jsrinivasan@gibsondunn.com Washington, DC 20036
5 GIBSON, DUNN & CRUTCHER LLP Telephone: 202.955.8500
333 South Grand Avenue Facsimile: 202.467.0539
6 Los Angeles, CA 90071
Telephone: 213.229.7000 ETHAN DETTMER, SBN 196046
7 Facsimile: 213.229.7520 edettmer@gibsondunn.com
ELI M. LAZARUS, SBN 284082
8 VERONICA S. MOYÉ (Texas Bar No. elazarus@gibsondunn.com
24000092; pro hac vice) GIBSON, DUNN & CRUTCHER LLP
9 vmoye@gibsondunn.com 555 Mission Street
GIBSON, DUNN & CRUTCHER LLP San Francisco, CA 94105
10 2100 McKinney Avenue, Suite 1100 Telephone: 415.393.8200
Dallas, TX 75201 Facsimile: 415.393.8306
11 Telephone: 214.698.3100
Facsimile: 214.571.2900 Attorneys for Defendant APPLE INC.
12
13
14
15
UNITED STATES DISTRICT COURT
16
FOR THE NORTHERN DISTRICT OF CALIFORNIA
17
OAKLAND DIVISION
18
19 EPIC GAMES, INC., Case No. 4:20-cv-05640-YGR-TSH
20 Plaintiff, Counter- NOTICE OF FILING OF DEFENDANT

defendant APPLE INC.’S EXPERT WRITTEN DIRECT
21 TESTIMONY
v.
22
APPLE INC.,
23
Defendant,
24 Counterclaimant.
25
26
27
28
Gibson, Dunn &

Crutcher LLP NOTICE OF FILING OF DEFENDANT APPLE INC.’S EXPERT WRITTEN DIRECT TESTIMONY, 4:20-cv-05640-
YGR
Case 4:20-cv-05640-YGR Document 490 Filed 04/27/21 Page 2 of 2
1 Apple Inc. respectfully submits the written direct testimony of the following experts:
2 1. Lorin M. Hitt, Ph.D
3 2. Francine Lafontaine, Ph.D
3. Richard Schmalensee, Ph.D
4 4. Daniel L. Rubinfeld
5 5. Dominique Hanssens, Ph.D
6. Aviel D. Rubin, Ph.D
6 7. James E. Malackowski
7
9 Dated: April 27, 2021 Respectfully submitted,

10 GIBSON, DUNN & CRUTCHER LLP
11
12 By: /s/ Rachel S. Brass

Rachel S. Brass
13
14 Attorney for Defendant Apple Inc.
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Gibson, Dunn & 1

Crutcher LLP
NOTICE OF FILING OF DEFENDANT APPLE INC.’S EXPERT WRITTEN DIRECT TESTIMONY, 4:20-CV-
05640-YGR
Case 4:20-cv-05640-YGR Document 490-1 Filed 04/27/21 Page 1 of 91
NORTHERN DISTRICT OF CALIFORNIA
OAKLAND DIVISION
EPIC GAMES, INC., No. 4:20-CV-05640-YGR-TSH

Plaintiff, Counter-defendant, WRITTEN DIRECT TESTIMONY OF
LORIN M. HITT, PH.D.
vs.
Trial Date: May 3, 2021
APPLE INC., Time: 8:00 a.m.
Courtroom: 1, 4th Floor
Defendant, Counterclaimant. Judge: Hon. Yvonne Gonzalez Rogers
Ex. Expert 6
April 23, 2021
Epic v. Apple, No. 4:20-CV- 1 Written Direct Testimony of

05640-YGR-TSH Lorin M. Hitt, Ph.D.
I. Summary of opinions
1. My analysis in this matter focused on assessing the economic evidence related to the questions
of market definition, Apple’s alleged market power or monopoly power, and the competitive
effects of the challenged conduct. Based on my extensive analysis of available data, documents,
and testimony, I have reached the following opinions:
Market Definition
2. Opinion 1. My opinion is that the relevant antitrust market to assess the impact of Apple’s
policies, pricing, and innovations is the market for digital game transactions. Because the App
Store operates in a two-sided transaction market, we need to consider substitution possibilities on
both sides of the market. From the perspectives of both developers and consumers, the App Store
is a substitute for, and competes with, other digital game transaction platforms. (§III, p. 6)
3. Opinion 2. This overarching conclusion is strongly supported by quantitative and qualitative

evidence showing that developers of many games make digital game transactions across a wide
variety of platforms (on mobile devices, PCs, Macs, laptops, tablets, consoles and other devices).
Developers can substitute between game transaction platforms for transactions with users,
including iOS users. (§§III.A-III.D, pp. 6-16)
4. Opinion 3. With respect to the other side of the platform, my empirical analyses also
demonstrate that consumers easily substitute across game transaction platforms. That is,
consumers can and do make digital game transactions—including Fortnite transactions—across
different devices through several game transaction platforms available on those devices.
Consumers own or have access to multiple devices on which they can make game transactions,
and consumers can and do substitute game transactions through the App Store with transactions
through platforms on other devices. To provide just one example, empirical evidence shows that
in response to Fortnite’s removal from the App Store, many Fortnite consumers switched their
purchases from iOS to other transaction platforms. (§§III.E-III.I, pp. 16-37)
5. Opinion 4. Additional qualitative evidence that I have reviewed confirms that multiple digital
game transaction platforms are alternatives to transacting on the App Store and are substitute
products. Internal documents and public statements demonstrate that industry participants,
including Apple, operators of other game transaction platforms, and Epic itself, consider other
game transaction platforms as competitors to the App Store for digital game transactions.
Policies imposed on developers by Nintendo, Sony, and Microsoft aimed at limiting competition
from alternative transaction platforms further underscore the reality of competition in the
relevant market. (§III.J, pp. 37-41)
6. Opinion 5. The evidence shows that game transactions face different competitive conditions
compared to non-game app transactions and therefore Epic’s approach, which clusters together
all types of apps into an undifferentiated all-app transaction (or “distribution”) market, is
incorrect. The set of competing transaction platforms, the market participants on each side of the
platform, and how developers typically monetize their apps all differ as between game apps and
non-game apps. (§III.K, pp. 41-47)

7. Opinion 6. Dr. Evans’ market definition analyses—and accordingly his conclusions regarding
monopoly power—suffer from numerous economic flaws. His focus on substitutability of
devices ignores that the App Store produces transactions between consumers and developers—
transactions in which consumers can freely engage across platforms without replacing one device
with another. Indeed, this issue does not arise for the vast majority of consumers who already
own multiple devices hosting platforms for digital game transactions. (§§V.A-V.C pp. 78-80).
He inexplicably carves tablets out of his market definition even though iOS apps are developed
for both iPhones and iPads. (§V.D pp. 80-81) His conclusion that apps on smartphones are not
substitutes for apps on other devices rests on faulty evidence. (§§V.E-V.G pp. 81-83) And his
“case study” of Fortnite user behavior is uninformative for the hypothetical monopolist test he
attempted to conduct yet nevertheless shows meaningful substitution across game transaction
platforms. (§§V.H-V.I pp. 83-85)
Market Power
8. Opinion 7. My market share calculations support the conclusion that Apple does not have
market or monopoly power in a properly defined market. Apple’s share of the digital game
transaction market lies between 23.3% and 37.5%. In light of my conservative approach, these
market share estimates, especially at the high end, are likely to overstate Apple’s true market
share and are, in any event, inconsistent with Apple having substantial market power. The entry
of new game transaction platforms is also inconsistent with Apple having market power. (§IV.A,
pp. 49-53)
9. Opinion 8. Market outcomes are inconsistent with Apple having substantial market power
regardless of how the market is defined. Apple’s commission rate on paid transactions is
competitive with other game transaction platforms and Apple has decreased its commission rate
over time for many transactions. Developers have also increasingly shifted to free-to-download
apps, which do not require the developer to pay a commission to Apple. Factoring in free
transactions, Apple’s average effective commission for digital game transactions is 8.1%. The
total value of digital game transactions has risen for the market overall. And both the number of
transactions and total developer revenues on the App Store have increased, growing at a faster
rate than the market overall and other competing game transaction platforms. There is simply no
evidence of any price increase or output restriction associated with the onset of monopoly power,
which Dr. Evans dates to 2010. Further, despite Dr. Evans’ claims that Apple has failed to
innovate, Apple has continuously invested in innovation to improve the quality of the iOS
ecosystem and App Store transactions. (§IV.B, pp. 53-69)
10. Opinion 9. Apple would be constrained from exercising market power even if the relevant
two-sided transaction market were defined (incorrectly) to consist of all iOS app transactions
(and only iOS transactions). Several market characteristics support this conclusion. For example,
developers are able to monetize their apps without paying a commission to Apple, including
allowing consumers to engage in transactions outside of the app that provide content or other
digital functionality that can be used within the app. Consumers can (and do in significant
numbers) switch between iOS devices and Android devices, further constraining any ability by
Apple to exercise market power. Further, the empirical evidence on prices and quantity is
consistent with Apple being constrained from exercising market power in an alleged market for
iOS app distribution. (§§IV.C-IV.D, pp. 70-77)

Competitive Effects
11. Opinion 10. All evidence suggests Apple’s App Store model has been profoundly
procompetitive. As described in connection with my market power analysis, prices are falling,
entry is constant, output and value are increasing, and nonstop innovation is driving quality ever
higher (which effectively lowers prices). Whether looking at just game transactions or all app
transactions, the digital marketplace is thriving. Epic provides no evidence to the contrary.
(§VI.C, pp. 89-91)
12. Opinion 11. Dr. Evans’ case studies, designed to show that Apple’s conduct harms
competition, in fact reveal that Apple’s model leads to lower prices, greater output, and higher
quality. His analysis of Android app distribution in China and distribution of games on PCs and
Macs suggests that developers and consumers would be worse off if Apple were forced to allow
distribution outside the App Store. (§VI.A, pp. 86-87)
13. Opinion 12. Professor Athey’s conceptual analysis of “middleware” and “multi-platform app
stores” is wholly speculative and proposes a drastic form of forced interoperability to solve a
problem that does not actually exist. (§VI.B, pp. 87-89)

II. Background and Qualifications
14. My name is Lorin Moultrie Hitt. I am the Zhang Jindong Professor of Operations,
Information and Decisions at the University of Pennsylvania, Wharton School.
15. As a member of the information strategy and economics group at Wharton, my research and
teaching focus on the economics of the information technology industry and related industries,
with a specific emphasis on firms engaged in or affected by Internet-based commerce. I have
conducted research on pricing and competition among online travel agents, on the effect of
product recommendation services on consumer choice and product price, on switching costs and
customer loyalty to online brokers (and online businesses more broadly), on consumer behavior
in online services such as banking and healthcare, and on the economic impact of social media
services such as Facebook and Twitter. More broadly, my research encompasses both theoretical
modeling and empirical analysis of pricing, marketing, competition, and consumer behavior in
online markets.
16. I have published more than 35 peer-reviewed articles in top-tier economics and management
journals such as the Quarterly Journal of Economics, the Review of Economics and Statistics, the
Journal of Economic Perspectives, Management Science, and Information Systems Research, as
well as more than two dozen other publications in books, trade journals, and other practice-
oriented outlets. I previously held senior editorial positions at three major research journals:
Management Science, Information Systems Research, and Journal of Management Information
Systems. I twice served on the conference committee for the Workshop on Information Systems
and Economics, the primary conference in my sub-discipline.
17. I have taught courses at the University of Pennsylvania and the Massachusetts Institute of
Technology on competition and customer pricing in a variety of commercial and consumer
markets, information systems management, electronic commerce, information economics, data
analysis, and methodologies used to understand the impact of information technology
investments and strategies on firms, consumers, and markets. I created one of the first courses on
the economics of electronic commerce (first offered in 1998), which I taught for over 16 years. I
continue to cover similar material in my Ph.D. seminar along with theoretical and empirical
methods used in economic research and their application to the study of online markets including
the markets for computers, smartphones, and software (such as apps).
18. I received my Ph.D. in Management from the Massachusetts Institute of Technology Sloan
School of Management in 1996 and my Sc.B. (1988) and Sc.M. (1989) degrees in Electrical
Engineering from Brown University. The majority of my Ph.D. coursework was in economics
and statistics. I took my Ph.D. qualifying exam in econometrics from Jerry Hausman (who
developed the theory underlying one of the most common methods of testing for econometric
identification), and my doctoral dissertation was supervised in part by Zvi Griliches (Harvard), a
former Chairman of the American Economic Association and a pioneer in methods for
understanding the relationship between prices and quality change in complex, differentiated
products.

III. Market Definition: The relevant antitrust market is the market for digital game
transactions
19. My market definition analysis builds on the economic frameworks set forth by Professor
Schmalensee and Professor Lafontaine. Based on Professor Schmalensee’s framework on two-
sided markets and Professor Lafontaine’s framework for defining antitrust markets, I have
conducted extensive qualitative and quantitative analyses regarding the relevant product market
in this case. I conclude that the evidence strongly supports that the relevant antitrust product
market is the market for digital game transactions.
20. Professor Lafontaine provides a clear framework that I adopt and apply in my analysis. As
she discusses, the starting point for assessing the economic effects of the conduct at issue is
market definition, which requires identifying relevant substitutes for the product provided by the
App Store.
21. Thus, I begin my analysis with the product that Apple jointly provides to Epic and Epic’s
customers through the App Store—“digital game transactions.” These are transactions that
facilitate the exchange of playable digital game content (including initial downloads, re-
downloads, and updates) and in-app purchases of game content between developers and
consumers.
22. The App Store is a two-sided transaction platform, serving both consumers and developers in
making transactions with each other. As described by Professor Schmalensee, it is therefore
necessary to consider substitutability for these transactions on both sides of the platform,
meaning for both consumers and developers, when conducting an analysis of the relevant
product market.
23. My empirical analyses and research described below support my finding that digital game
transaction platforms provide services that are strong substitutes, both for the same game and
across games, from the perspective of both consumers and developers of games. As discussed
below, other app transactions facilitated by Apple are not strong substitutes for digital game app
transactions. I therefore concluded that digital game transactions represent a distinct market.
Moreover, Apple competes in this market with other digital game transaction platforms found on
PCs, consoles, tablets, and other devices. Epic’s assertions that there is a single-brand market for
iOS app distribution in which the App Store has no competitors is therefore wrong.
A. Game developers have the option to make transactions through many platforms
besides the App Store
24. When considering the options for developers to make game transactions, I started by
identifying the transaction platforms that currently exist and in which consumers and developers
actively make game transactions. My analysis shows that, counter to Epic’s characterization of
the world, there are many transaction platforms that compete in the market for digital game
transactions and serve as substitutes to the App Store for developers.
25. In Figure 1 below I show a list of 14 prominent platforms that offer transaction services, by
which I mean the ability to transact with consumers, to digital game developers. Each of these

platforms serve the exact function as the App Store in that they connect developers to consumers
in order to conduct transactions for many different digital games.
Examples of digital game transaction platforms
Platform Compatible devices

1. Amazon Appstore Android, Fire OS, Blackberry
2. App Store iOS devices
3. Epic Games Store PC, Mac
4. GOG.com PC, Mac, Linux
5. Google Play Android, Chrome OS
Android, Chromecast Ultra, PC,
6. Google Stadia Mac, iOS devices (via Safari
browser)
7. itch.io PC, Mac, Linux, Android
8. Mac App Store Mac
9. Microsoft Store PC, Xbox One, Xbox Series
10. Nintendo eShop Switch, Nintendo 3DS, Wii U
11. Origin PC, Mac
12. PlayStation Store PS4, PS5
13. Samsung Galaxy Store Samsung Android
14. Steam PC, Mac, SteamOS + Linux
Source: DX-4758
26. Developers may choose to develop certain games for only one platform—for example, an
exclusive release of Gears of War 3 on Xbox. Developers may also choose to create different
games for different platforms. For example, Epic released Infinity Blade on iOS, Gears of War
on Xbox and PC, Gears of War 3 on Xbox, and Fortnite on PlayStation, Xbox, Switch, PC, iOS
and Android. Developers can also release different versions of a game for different platforms—
for example, the initial release of the Minecraft mobile version had limited features compared to
non-mobile versions. It is the developer’s choice where to offer its games, and hence where to
transact with potential customers.
B. Many game developers choose to use multiple transaction platforms to make game
transactions with consumers for a single game
27. Some developers may choose not to conduct transactions for a given game on multiple
platforms. But this is not a sign that the choice was not available to them to do so. The most
direct evidence that developers can transact on different game transaction platforms is the fact
that many developers indeed choose to make transactions for a single game on multiple
platforms. Developers commonly do this, and empirical evidence shows that a large number of
game app developers that transact through the App Store also transact through other platforms.

game transaction platforms into perspective. Epic’s data are also helpful in this case because they
allow me to examine how individual consumers of the same game transact across platforms.
37. In particular, I analyzed a dataset containing monthly playtime and spending for over 350
million Fortnite users from January 2017 to December 2020. I understand that these data capture
all Fortnite users across the globe.
38. These data show that Epic makes transactions with the majority of its users through game
transaction platforms other than the App Store (see Figure 6). Before the release of Fortnite on
iOS, Fortnite had already attracted 64 million users through other devices and transaction
platforms. The number of accounts that accessed the game on other platforms also continued to
grow after the launch of Fortnite on iOS. In fact, prior to the removal of Fortnite from the App
Store, the game had 355 million distinct user accounts, only about one-third of which had
accessed the game through iOS.
Worldwide Fortnite user accounts by platform (July 2017 – December 2020)
Source: DX-4761 (summarizing DX-5339, DX-4314, DX-3686)
39. My analysis overall shows that game developers in general have many choices of where and
how to conduct game transactions with consumers and that they actively choose between these
game transaction platforms. My analysis also shows that Epic in particular has been extremely
successful in conducting game transactions through these other game transaction platforms.

C. Developers, including Epic, treat the App Store and other game transaction
platforms as substitutes for making game transactions
40. Developers substitute across game transaction platforms in part by making decisions about
whether, when, and how much to invest in a given game for a particular transaction platform.
41. My research for this case shows that developers consider the relative benefits of different
game transaction platforms along multiple dimensions in deciding on what platforms to transact.
This includes the ability to generate revenue through the platform, the commission charged by
the platform, the technical sophistication of the devices on which the platform is available, the
developer tools available to aid the developer in creating the game, the number of consumers on
the platform, the amount those consumers might be expected to spend on the platform, and other
services provided directly by the platform operator to the developer. These are dimensions upon
which digital game transactions compete with one another for developers. And as my analysis
above shows, developers have the ability to substitute between these platforms depending on
these factors—this includes choosing which games, and which versions of games, to offer on
each game transaction platform.
42. My research also shows that many technical tools exist to provide developers the maximum
set of options in terms of choosing among platforms on which to make game transactions. For
example, the Unity game engine now supports development for 20 platforms, allowing
developers to create one software “build” that they can distribute to any of Unity’s supported
platforms. These tools, combined with the evidence above that developers have created games
for multiple platforms, make it clear that developers are not locked in to any given platform.
Instead, developers can choose between different transaction platforms based on those platforms’
economic merits.
43. Documentary evidence I have reviewed for this matter support this exact dynamic. Evidence
shows that Epic considered the App Store and other game transaction platforms as alternatives
and made a deliberate choice to focus mobile development efforts for Fortnite on iOS based on
an assessment of the merits of the iOS platform.
44. Fortnite was first launched on Xbox, PlayStation, and PCs. Epic then decided to “focus [its]
engineering efforts” for Fortnite on iOS as opposed to launching on both iOS and Android
simultaneously. 3 This demonstrates Epic’s willingness to reallocate resources across transaction
platforms depending on their relative advantages. Furthermore, Epic expected to receive “extra
support” from Apple by launching on the App Store rather than Google Play, which is consistent
with Epic’s understanding that Apple treats Google Play as a competing transaction platform. 4
45. Similarly, Epic’s strategy to initially forgo offering Fortnite on Google Play shows that Epic
considered alternative game transaction platforms for the game. Emails between Tim Sweeney
and Samsung indicate that Epic did not want to partner with Google and decided to instead offer
transactions for Fortnite for Android devices through the Samsung Galaxy Store instead of
3
DX-3732.002-003
4
DX-3732.002

Google Play, even using its competitive leverage to obtain a lower commission rate for Fortnite
with Samsung. 5
46. Developers of course do not always choose to offer every game on every platform, or even
every game on multiple platforms. But this does not mean that different transaction platforms are
not substitutes merely because the developer chose where to invest its efforts across games. Epic
has itself weighed the benefits of game transactions on different platforms against each other for
different games. For instance, while Epic developed and released the Infinity Blade trilogy only
on the App Store, its Gears of War 3 (released after Infinity Blade) is exclusive to Xbox. Fortnite
is available on many different game transaction platforms because Epic has chosen to invest in
making it so, while other developers, choosing fewer platforms on which to transact, have not.
D. Game developers who have decided to make games for iOS devices have multiple
options for making paid transactions with iOS users outside of the App Store, even for
content consumed on iOS apps
47. My analysis of the market for game transactions shows that beyond the choice on which of
the many substitute transaction platforms to offer any given game, developers also have
meaningful design choices available to them that impact where and how consumers will pay
them for their games, i.e., make game transactions.
48. In regard to iOS apps, developers can use competing game transactions platforms, as well as
websites, for paid transactions with consumers, even for in-game content that is used in an iOS
app. It is thus demonstrably wrong to claim, as Epic’s economists do, that game developers can
transact with Apple mobile device users only through the App Store and therefore other game
transaction platforms are not part of the relevant market.
49. Developers have many design choices that allow them to sell digital content on other
platforms that differ from where that content is consumed. I show that these design choices are
commonly used by developers for the top game apps in the App Store, facilitating the ability to
substitute between transaction platforms.
50. Many game developers link in-game content and game progression across a user’s devices
through a common user account (or “single sign-on”). This single sign-on can be offered by the
developer or through a third-party (such as Apple, Google or Facebook) and lets consumers
easily shift their play and in-game content across devices. My Figure 7 shows that the vast
majority of the highest grossing game apps on the App Store and a large portion of the most
downloaded game apps on the App Store offer single sign-on and allow users to port in-game
content and progression across devices.
5
DX-4457
Portability of FY2019 top game apps in the App Store
Offer full or
Offer single partial
App Category sign-on portability
Top 25 by revenue 92% 84%
Top 25 by downloads 32% 32%
Source: DX-4777 (summarizing DX-5338, DX-4324, online public materials)
51. Many game developers also allow consumers to purchase digital content that will be used
within iOS from alternative transaction platforms or through a website. Figure 8 shows that the
vast majority of the highest grossing game apps allow consumers to purchase content outside the
iOS app for use within the iOS app, and a full one-third offer direct purchases of content through
a web browser on the iOS device for content that can be accessed in an iOS app.
Monetization and portability of FY2019 top game apps in the App Store
For apps offering purchased digital content

Offer digital Content purchased Content purchased
content for outside of iOS app through web browser
App Category purchase accessible in iOS app accessible in iOS app
Top 25 by revenue 100% 84% 32%
Top 25 by downloads 100% 28% 8%
Source: DX-4798 (summarizing DX-5338, DX-4324, online public materials)
52. Epic is an excellent example of a game developer that has chosen to invest in making these
various features available in its games, allowing consumers to seamlessly substitute digital game
transactions for the same game across transaction platforms (when supported by the platform, as
Apple does).
Epic allows consumers to play Fortnite and to pay for V-bucks (Fortnite’s digital
currency) on different platforms.
Consumers’ ability to purchase V-Bucks and the ability to acquire in-game content
using V-Bucks is functionally identical across devices.
Customers have device-agnostic accounts, linked through a single sign-on, that allow
them to access their V-bucks on any device. (V-bucks purchased on the PlayStation
Store or the Nintendo eShop are an exception because of restrictions imposed by
those platforms.)

Once content is purchased in-game with V-Bucks, it is tied to the user’s account and
can be accessed and used across any device.
In-game progress is tied to a user’s Fortnite account. This progress persists across all
devices that offer Fortnite, allowing users to continue playing on one device where
they left off on another.
Consumers using different devices can play with each other simultaneously without
restrictions (known as “cross-play”). This cross-play functionality allows players to
switch devices without affecting their ability to play with others. This limits the
potential for network effects on a single device to discourage users from switching to
a different device.
53. In fact, as shown above, if a developer chooses, it can let consumers make transactions
through substitute platforms, or websites, even for content used in the iOS app. Epic has made it
possible for a Fortnite user on iOS to choose to never make a single paid transaction through the
App Store but still enjoy all the paid features of the game when playing on an iOS device. A
Fortnite user can simply use content acquired on other devices or through a website on their iOS
device, and can also spend V-bucks purchased on competing game transaction platforms for
digital content used when playing on iOS.
54. Overall, the evidence above clearly demonstrates that developers have the option to conduct
digital game transactions across platforms and can substitute between platforms for transacting
with consumers, both within and across games on different game transaction platforms, and even
for content consumed within iOS apps.
E. Empirical evidence shows that consumers treat game transactions on other

transaction platforms as substitutes for game transactions on the App Store
55. The fact that developers have so many options to offer game transactions to consumers is
mirrored by the fact that consumers have many options available to make digital game
transactions—this connection is a fundamental feature of two-sided platforms. I conducted a
number of analyses in order to assess whether consumers can, and do, substitute across game
transaction platforms.
First, I assessed the degree to which consumers use, own, or have access to multiple
devices on which game developers offer games—this establishes that consumers
multi-home across relevant devices for game transactions and have the ability to
switch among devices.
Second, I assessed the degree to which consumers play games across multiple
devices—this establishes additional support for the fact that consumers do indeed
multi-home when making game transactions across devices and across games.
Third, I assessed the degree to which consumers could multi-home within the most
popular games, for example playing or paying for content for the same game across
different devices. I analyzed this in detail through data provided by Epic.
Finally, I assessed real world examples that show how consumers actively substitute
across game transaction platforms for making game transactions.

These analyses together show that consumers can and do substitute digital game transactions
across different devices through the various game transaction platforms available on those
devices.
F. Consumers can make transactions across many platforms because they own or
regularly use different devices on which game transaction platforms are available
56. The first type of evidence I assessed was whether consumers regularly use, own, or have
access to other devices besides iOS devices. The evidence on this front was clear, and confirmed
by a number of different sources.
57. Multiple surveys show that consumers do in fact use, own, or have access to multiple devices
on which they can make game transactions. For instance, a 2019 study of all U.S. consumers
shows that 81% owned a smartphone, almost 75% owned a desktop or laptop computer, and
approximately 50% owned a tablet. 6 These general observations also hold for owners of iOS
devices. An Apple survey of iPhone owners conducted in July 2020 found that 44% owned a
Mac notebook or desktop, 55% owned a Windows PC laptop or desktop, 30% owned a gaming
console, and 21% owned a non-iPad tablet (such as a Samsung or Amazon Fire tablet). 7 These
findings are consistent with the opinion of Professor Athey who opined that consumers typically
have access to two to three general purpose devices.
58. Two surveys conducted for this case by Professor Hanssens also find that consumers use,
own, or have access to multiple devices besides their iOS device. As my Figure 9 shows below,
according to Professor Hanssens’ survey, 81% of App Store users regularly use another device
besides their iOS device. Professor Hanssens’ survey finds that an even greater amount regularly
use or have access to another device besides their iOS device: 86% regularly use or have access
to a laptop, 64% a desktop, 61% a console/handheld game device, 56% a non-iOS smartphone,
and 48% a non-iOS tablet. Overall, 95% of App Store users regularly use, or could use, a device
besides their iOS device. The results are similar when looking at the more specific group of
consumers who play Fortnite on an iOS device. Of this group almost all (i.e., 94%) regularly use
other devices besides their iOS device.
6
DX-4453
7
DX-3174.003-004
Other electronic devices used by iOS App Store users

Regularly Used This Device
in the Last 12 Months
Number of
Respondents Percentage of
Device (n=500) Respondents
Smartphones with Non-iOS Operating Systems 136 27%
Android 108 22%
Microsoft 66 13%
Tablets with Non-iOS Operating Systems 116 23%
Android 90 18%
Microsoft 57 11%
Laptops 357 71%
Apple 182 36%
Brands Other Than Apple 252 50%
Desktops 238 48%
Apple 111 22%
Brands Other Than Apple 178 36%
Gaming Consoles and/or Handheld Gaming Devices 206 41%
Nintendo Switch (including Nintendo Switch Lite) 104 21%
PlayStation (PS Series Console) 125 25%
Xbox 106 21%
Nintendo DS Series 52 10%
PlayStation Vita 28 6%
GPD XD Plus 6 1%
Other Electronic Devices 458 92%

Other Electronic Devices (Non-Apple) 405 81%
Source: DX-4661
59. The fact that owners of iOS devices use, own, or have access to multiple other devices that
provide game transactions means that these consumers can make game transactions on them.
G. Consumers can make transactions across many platforms because they play games
on the many devices they own
60. My analysis shows that consumers play games across multiple devices and thus are already
making game transactions through different platforms.
61. Survey data show that consumers play games and can make digital game transactions across
devices and transaction platforms. Professor Hanssens’ survey, for instance, found that 92% of
iOS Fortnite players also play games on non-iOS devices. Additionally, a 2018 industry survey
of game players showed that of U.S. gamers play games on more than one device and
on more than two devices. 8 This study found that of gamers play on a mobile device and “at
least one other platform,” and that “play on Mobile, PC & Console.” 9 These sources show
8
DX-4170.010
9
DX-4170.011-012
that consumers play games and can make digital game transactions across devices and
transaction platforms.
62. Substitution of game transactions by consumers is not of course limited to the same game
across game transaction platforms. Evidence indicates that cross-game substitution is common.
For example, Professor Hanssens’ survey found that 92% of iOS Fortnite players play games on
non-iOS devices. Epic’s own data indicate that 35.9% of iOS Fortnite players also play Fortnite
on non-iOS devices. 10 The difference in these two statistics indicates that many iOS Fortnite
players play other games on non-iOS devices, even if they do not play Fortnite on non-iOS
devices.
63. Even this high degree of substitution between devices may of course understate the
opportunities for consumers to play or transact on other devices if they chose or were
incentivized to do so. In other words, just because consumers do not substitute game transactions
across devices now does not mean they would not do so if faced with a price increase on one
particular platform.
H. Consumers can and do substitute game transactions through the App Store with
transactions through other game transaction platforms
64. As I have already described in my analysis of top game apps, many developers offer versions
of their games on multiple devices. Because of this, consumers can not only substitute game
transactions between platforms for different games but can also substitute game transactions for
the same game on different game transaction platforms.
65. My analysis also assesses actual consumer spending and use across game transaction
platforms to show that they are technical substitutes. While I have already established that
consumers own, access, and regularly use multiple devices on which to make game transactions
and that consumers play games on these same devices, the goal of this analysis is to assess
individual consumers and their actual game transactions across platforms to provide insights into
how consumers substitute across game transaction platforms.
66. The only data available to me that tracked individuals and their play and purchase decisions
across game transaction platforms are the Epic data. While these data are limited to just one
game (Fortnite), they are instructive because they exhibit both the scope of substitutes available
to developers and the consumer substitution behavior in regard to playing and making paid game
transactions.
67. My analysis of the Epic data led me to the strong conclusion that consumers can and do
substitute game transactions across devices and transaction platforms. Dr. Evans agrees with me
on this basic fact. The Epic data establish a number of important facts in support of this
conclusion, including that consumers often multi-home within Fortnite generally, that iOS
consumers in particular do, and that there is ample evidence that iOS users substitute their paid
transactions for transactions on other platforms.
10
DX-4767
68. I first assessed the Epic data to show that Epic’s customers generally play Fortnite on
multiple devices and make transactions through multiple game transaction platforms. For
example, about a third of all Fortnite users have an iOS account, as shown in Figure 10. Yet, as I
show in Figure 11 below, the Fortnite data also show that Fortnite users spent the vast majority
(95.6%) of their time playing Fortnite on non-iOS devices. For time played, iOS ranks fifth
overall, trailing Sony PlayStation, Microsoft Xbox, PC, and Nintendo Switch.
Worldwide share of Fortnite user accounts across platforms (March 2018 – July 2020)
User accounts
(in millions)
Platform Count Percent
1. PS4 104 30.2%
2. Xbox One 53 15.5%
3. PC 90 26.1%
4. Switch 31 9.0%
5. iOS 115 33.4%
6. Android 31 9.1%
7. Google 8 2.2%
8. Other 10 2.9%
9. Total 344 -
Source: DX-4765 (summarizing DX-5339)

Note: Percentages do not total to 100% because a user can have accounts on multiple devices.

Percent of worldwide Fortnite time played across platforms (March 2018 – July 2020)
69. Fortnite users also make the vast majority of their paid transactions outside of the App Store.
As I show in Figure 12, only 7% of revenue from Fortnite users is collected through the App
Store —again ranking fifth behind PlayStation, Xbox, PC, and Switch.

Percent of worldwide Fortnite revenue across platforms (March 2018 – July 2020)
70. I then assessed the extent to which individual Fortnite users on a given transaction platform
multi-home by playing Fortnite on other devices. These multi-homing users can easily shift game
transactions from one particular game transaction platform, such as the App Store, to another
platform, either for the same game or for other games.
71. I found that for each platform on which Fortnite is available, a large number of Fortnite users
already play Fortnite on at least one additional device. These users could therefore make game
transactions for Fortnite through those different game transaction platforms. My Figure 13 shows
more than a third of iOS Fortnite users, and more than 40% of Fortnite players on PC, have
played the game on at least one other device. This demonstrates that these consumers can easily
substitute game transactions through other platforms.

Worldwide user accounts that access Fortnite through multiple platforms (March 2018 – July 2020)
Total users Percent of users that are

Platform (in millions) multi-platform users
1. iOS 115 35.9%
2. PS4 104 32.1%
3. PC 90 42.4%
4. Xbox One 53 33.0%
5. Android 31 53.7%
6. Switch 31 34.7%
7. Hong Kong 10 5.9%
8. Google 8 54.4%
9. Other 0 99.8%
72. I then focused specifically on consumers who play Fortnite on iOS devices to confirm that
these users also play and transact on multiple devices and transaction platforms. I found that they
do. My analysis shows that Epic can and does use alternative game transaction platforms to
transact with consumers who access Fortnite through an iOS device.
73. Not only do most Fortnite users spend most of their time and money outside iOS, as I have
already shown, but my analysis in Figure 14 shows that even the specific subset of users who
accessed Fortnite on iOS spent the vast majority of their time and money with regard to Fortnite
on non-iOS devices and through game transaction platforms other than the App Store. This is
direct evidence that these iOS users have and use substitutes to the App Store for making game
transactions.
Percent of worldwide time played and revenue by user accounts that accessed Fortnite through iOS
(March 2018 – July 2020)
Hours played Revenue

(in millions) (in millions)
Platform Hours Percent Revenue Percent
1. iOS 2,752 10.2% $745 13.2%
2. Other platforms 24,219 89.8% $4,890 86.8%
3. Total 26,971 100.0% $5,635 100.0%
74. Data on where Fortnite customers choose to make in-game purchases show that transaction
platforms are substitutes for consumers to engage in game transactions. While the majority of
Fortnite’s iOS users do not make paid Fortnite transactions on any devices, of those that do, well

I. Evidence shows that developers and consumers substitute across game transaction
platforms, even in the absence of a relative price change incentivizing them to do so
77. Game transactions on the App Store are substitutes, rather than complements, for game
transactions on other game transaction platforms. This is true by the very nature of game
transactions: any digital game transaction made through an alternative transaction platform or a
website is a direct, technical substitute for a game transaction made through the App Store.
78. Consumers and developers of course substitute across game transaction platforms when a
consumer chooses to play a game app on a non-iOS device. But consumers and developers can
also substitute payment for content across transaction platforms, even if a consumer uses the
content in an iOS app. This is because consumers can pay for content on one transaction
platform and access that content on another transaction platform. In other words, any observed
paid transaction on one platform is a technical substitute for a paid transaction on an alternative
platform or through a developer’s website, even if a consumer is using the content in an iOS app.
79. Epic’s experts recognize that most consumers own multiple devices and can conduct game
transactions both through the App Store and through other platforms. But they incorrectly
characterize these other platforms as complements rather than substitutes. As one example,
Professor Athey simply asserts that devices are complements based on her casual observations
that consumers use devices in different situations (like when watching TV on the couch or while
working at home). Neither she, nor Epic’s other experts, actually test whether devices are
complements (i.e., goods for which a decline in price for one good leads to an increase in
demand for the other good) based on the standard economic approach of looking at what happens
to the quantity of one good when the price of another good changes.
80. Epic’s experts’ assertions are simply wrong. A digital game transaction on the Sony
PlayStation store is a substitute for a digital game transaction on the App Store even if a
consumer chooses to make transactions on both of those platforms (and even if a consumer
chooses to make transactions in a different game), just as a Honda and a Toyota are substitutes
even if a consumer purchases both.
81. Real world examples prove this point. Even in the absence of a change in price on the App
Store that would incentivize consumers to substitute to other transaction platforms, these
examples show that developers and consumers treat transactions on other game transaction
platforms as economic substitutes, not complements, for game transactions on the App Store. In
particular, I analyzed five real world examples, three involving game transactions and two
involving non-game apps, all of which show that transactions on other platforms are substitutes
for transactions on the App Store: (1) users who download a console or PC game transaction app,
(2) the launch of Fortnite on the Nintendo Switch, (3) the removal of Fortnite from the App Store
after the Hotfix, (4) Spotify’s decision to stop offering subscriptions for in-app purchase on iOS,
and (5) Netflix’s decision to stop offering subscriptions for in-app purchase on iOS.

i. Analysis of users who downloaded a console or PC game transaction platform

companion app
82. First, to assess whether users substitute between game transactions on the App Store for
transactions on PCs and consoles, I analyzed iOS device users who downloaded a free
companion app for console or PC game transaction platforms (e.g., the Nintendo Switch Online
companion app, the Xbox companion app, and the Steam Mobile companion app). This analysis
(and many of my other analyses) uses Apple transaction data containing all initial downloads
(but not updates or re-downloads) and in-app purchases since the launch of the App Store until
September 2019, allowing me to calculate total spending on game apps for all iOS device
owners.
83. The download of these companion apps serves as a proxy for whether a user owns and plays
games on a console or a PC. While I do not have data on which iOS users own a game console or
play PC games, this proxy variable approach will identify individuals who have likely started
playing games more frequently on a console or PC. This could be because the user has recently
purchased a console or gaming PC or just because the user has decided to start playing games
more on an existing console or PC. But either way, if transaction platforms on consoles and PCs
are substitutes for the App Store, I would expect to see a decline in spending on games after
users download a companion app. 11
84. My results find exactly that. iOS device users who downloaded a companion app in 2018
had less spending growth on iOS games between 2017 and 2019 compared to “control group”
users who did not download a companion app prior to or in FY2019. The difference in growth
rates between users who did and did not download such a companion app is 5.1 percentage
points (Figure 16). The decline in relative spending for users who downloaded a companion app
could reflect both shifting of transactions from the App Store to a competing platform for the
same game the user would have purchased through the App Store or the user’s choice to
purchase an entirely different game.
85. I checked other sensitivities of this analysis, and the result is robust: users who downloaded a
companion app for console or PC game transaction platforms shifted at least some of their game
transactions from the App Store to other transaction platforms, indicating that these other
platforms are substitutes for the App Store.
11
There may be a concern that the companion app proxy variable is imperfect if, for instance,
users who own a console or play games on a PC do not download a companion app. This would
only imply that my estimate is conservative. If many users who own a console or play games on
a PC do not download a companion app then my analysis would underestimate the decline in
spending for users who download a companion app relative to the control group.
Comparison of growth rates in iOS games spending between 2017 and 2019
ii. The launch of Fortnite on the Nintendo Switch
86. Another real world example that shows that game transactions on other platforms are
substitutes for game transactions on the App Store, even in the absence of a price increase, is my
analysis of the launch of Fortnite on the Nintendo Switch.
87. Fortnite launched on the Nintendo Switch in June 2018. To determine if the App Store is a
substitute for the Nintendo eShop (the game transaction platform available on the Nintendo
Switch), I looked at users who played Fortnite on both devices in the first month the game was
available on the Nintendo Switch to see if there was a relative decline in iOS playtime and
spending on Fortnite. The results of my analysis clearly show that these users shifted their
playtime and spending away from iOS following the launch of Fortnite on the Nintendo Switch.
88. To understand this analysis, it is first important to note that spending on Fortnite fell for most
months I analyzed compared to May 2018 (the month prior to the release of Fortnite on the
Nintendo Switch). For example, from May 2018 to March 2019, Fortnite spending fell by 18%
on the App Store and by 33% across all platforms.

89. Thus, to account for the decline in overall spending on Fortnite, it is necessary to compare
users who played Fortnite on both iOS and the Nintendo Switch in June 2018 to a “control
group.” In particular, I compare the playtime and spending on iOS for these users to a “control
group” of users who played the game on iOS but not the Nintendo Switch in June 2018.
90. Figure 17 and Figure 18 clearly illustrate the findings. Fortnite users who accessed the game on
both iOS and the Nintendo Switch in June 2018 played Fortnite on iOS devices less and spent
less in the App Store in June 2018 and subsequent months (particularly in the months
immediately after June 2018) compared to other iOS users. While this analysis cannot capture
competition for new users who may have chosen to start playing and spending on the Nintendo
Switch rather than on iOS, it nevertheless shows that existing iOS Fortnite users substituted
transactions between iOS and the Nintendo Switch after Fortnite became available on the
Nintendo Switch.
Worldwide Fortnite iOS time played by user accounts that accessed Fortnite on an iOS device in
June 2018 (March 2018 – March 2019)
Source: DX-4822 (summarizing DX-5339, Nintendo, “Fortnite”)

Worldwide Fortnite iOS revenue from user accounts that accessed Fortnite on an iOS device in
June 2018 (March 2018 – March 2019)
Source: DX-4823 (summarizing DX-5339, Nintendo, “Fortnite”)
91. Because of the decline in spending on Fortnite across platforms during this time, it is also
informative to look at whether the share of revenues shifted across platforms in June 2018 and
later months for users who accessed the game on both an iOS device and the Fortnite Switch in
June 2018. The trend in the share of spending across different platforms for these users, shown in
Figure 19, particularly in the months immediately after June 2018, tells a consistent story that
users substituted spending away from the App Store after Fortnite became available on the
Nintendo Switch. The App Store’s share of spending by users who accessed the game on both
iOS and the Nintendo Switch in June 2018 fell from 27% in June 2018 to 24% in August 2018.
Over the same period, the share of Nintendo Switch increased from 15% to 20%. By March
2019, the App Store’s share fell to 17% and Nintendo Switch’s share was 19%.

Worldwide share of revenue from user accounts that accessed Fortnite both on iOS and on
Nintendo Switch devices in June 2018 by platform (March 2018 – March 2019)
92. Epic’s user data similarly show that the App Store’s share of total time played by users who
accessed the game on both iOS and the Nintendo Switch in June 2018, fell from 23% to 9% from
May 2018 to September 2018, while the Nintendo Switch’s share of total time played rose to
31% by September 2018. 12
93. Thus, empirical evidence following the introduction of Fortnite on Nintendo Switch,
particularly immediately after the event, is consistent with consumers substituting between the
two platforms.
iii. Analysis of Fortnite and the removal of Fortnite from the App Store after the Hotfix
94. Fortnite exemplifies the substitutability of game transactions across game transaction
platforms. As already described, Fortnite users play Fortnite on several devices and make game
transactions on different platforms. This is reflected in the Fortnite data, which show that from
March 2018 to July 2020, the share of users that accessed Fortnite on a particular platform that
also accessed Fortnite on at least one other platform (i.e., multi-homers) ranged from 32% to
54%. Between January and July 2020, iOS multi-homers accounted for over 85% of Fortnite
revenue from iOS device users.
12
DX-4770

95. V-bucks embody the substitutability of transactions for Fortnite across platforms. With
limited exceptions, V-bucks purchases are fungible across devices, and V-bucks purchased on
one transaction platform are the same as V-bucks purchased on another transaction platform. 13 A
consumer can purchase V-bucks all on one transaction platform, all on another transaction
platform, or on both. However, combining together V-bucks purchased on multiple platforms
does not result in something different or better (unlike when complementary products, like salt
and pepper, are combined) compared to V-bucks purchased on only one transaction platform.
96. Epic and Epic’s experts recognize this substitutability. Epic CEO Tim Sweeney has stated
“that games come first, and that a great game will succeed wherever it’s sold. It proves that
developers have the real power in the industry, and that where developers go customers will go
with them.” 14 An internal Epic presentation also found that when Epic introduced its Season 4
Battle Pass on May 1, 2018, Fortnite’s revenue generated on PS4 experienced a sharp increase,
while the share of revenue on iOS devices fell. 15 And Dr. Evans states that he can assess how the
removal of Fortnite from the App Store after the Hotfix can be used to determine whether users
found substitutes. 16
97. I analyzed data on all of Epic’s Fortnite users following the Hotfix to see whether iOS
Fortnite users switched their spending on Fortnite to non-iOS and non-Google platforms. Indeed,
they did. My analysis shows that Epic retained the vast majority (81.1%–87.7%) of iOS Fortnite
users’ pre-Hotfix revenue (across all transaction platforms) in the four months post-Hotfix period
(see Figure 20). The rate of revenue retention was
40.4%–50.7% for iOS single-homers, who in response to the change, began playing
and spending money on new non-iOS, non-Google transaction platforms for the first
time;
86.3%–93.7% for iOS multi-homers, who in response to the change, began spending
more money on non-iOS, non-Google transaction platforms than they had previously.
July 2020 iOS user switching and retention, worldwide (September 2020 – December 2020)
Jul/Sep Jul/Oct Jul/Nov Jul/Dec
Percent of pre-“hotfix” total revenue retained across
1. 81.1% 84.9% 87.7% 86.2%
all “platforms” (all iOS users)
Increase in revenue on other “platforms” as a

fraction of revenue lost on iOS
2. All iOS users 22.4% 36.5% 51.4% 40.9%
3. iOS single-homers 26.5% 37.5% 30.6% 22.4%
4. iOS multi-homers 18.7% 35.3% 63.2% 49.4%
13
Sony and Nintendo impose contractual terms which require V-bucks purchased through their
platforms only to be used on their own consoles. Apple does not have any such restrictions.
14
DX-3199.002
15
DX-4133.009-012
16
Evans Direct Testimony, ¶ 127

98. Dr. Evans cannot disagree with the fact that consumers switched spending from the App
Store to alternative transaction platforms. Despite his flawed approach to analyzing the Fortnite
data after the Hotfix (which I explain in more detail below), Dr. Evans himself finds that iOS
Fortnite users shifted a meaningful amount of Fortnite transactions to other game transaction
platforms after the Hotfix, a finding which Dr. Evans characterizes as substitution. He even finds
that iOS Fortnite users who had never previously made a Fortnite transaction on a non-iOS
platform before the Hotfix substituted to other transaction platforms after the Hotfix. 17
99. In fact, Epic publicly encouraged users to switch transaction platforms following Fortnite’s
removal from the App Store (Figure 21). My analyses of the Fortnite data show that effort was
successful. Such an effort to switch consumers to alternative transaction platforms could only
happen, however, because game transactions on these alternative platforms are substitutes for
transactions on the App Store.
Epic’s #FreeFortnite Cup press release
Source: DX-3724.002
17
Source: DX-4796 (summarizing DX-4641; Dawn Chmielewski, “Apple says Spotify’s app
already violates App Store rules,” Vox, Jul 1, 2016)
105. Netflix’s decision to stop offering subscriptions for in-app purchase on iOS. The results
are also consistent with the Netflix revenue and initial download numbers on iOS—
each continued to grow meaningfully after Netflix stopped allowing users to purchase in-app
subscriptions within the Netflix iOS app. 18 Consumers therefore substituted transactions through
the App Store for transactions on other transaction platforms as well as through the Netflix
website.
106. These real-world examples provide consistent evidence that consumers do indeed substitute
across transaction platforms, providing strong evidence that transactions on digital game
transaction platforms are substitutes for transactions on the App Store.
107. Epic’s experts, however, ignore empirical evidence of real-world substitution between
transaction platforms. Instead they claim that transaction platforms are complements, rather than
substitutes, because consumers use transaction platforms in different situations. These opinions
are non-economic and unreliable.
108. Epic’s experts have not provided any reliable economic evidence that game transaction
platforms are complements. They also ignore that consumers substitute across games since
consumers can choose to make a transaction for one game on a particular transaction platform
18
Netflix saw revenue increase from $15.8 billion in 2018 to $20.2 billion in 2019. See DX-
4558.022
instead of a different game on either the same transaction platform or a different transaction
platform.
109. For instance, I understand that one of Epic’s experts, Dr. Cragg, has analyzed Fortnite
users’ play time across devices and comes to the unremarkable conclusion that Fortnite players
who are motivated to play more tend to play on a greater number of devices (and vice versa).
This would not be evidence that different devices are complements. And whatever this may show
about complementarity in use of devices, such analysis would not imply that game transactions
are complements. Even if Epic’s experts showed that mobile and console devices are
complements, this would still be entirely consistent with substitution in game transactions. In
other words, users may be more likely to spend more time playing a game overall when they
have more devices to play on, but also substitute in-game purchases across the many devices on
which they play.
J. Industry participants recognize that the App Store competes with other game
transaction platforms
110. I have also analyzed internal and public documents from industry participants, and these
documents are consistent with my empirical analyses showing that developers and consumers
substitute between game transaction platforms. All of these documents show that game
transaction platforms compete with the App Store.
111. I have reviewed several documents which show that Apple views game transaction
platforms on many types of devices, including platforms on consoles, PCs, and mobile devices,
as competitors to the App Store. For instance, an October 2009 email to Apple’s vice president
of Worldwide Developer Relations, Ron Okamoto, discussed the competitive pressures from
Sony’s PSP Go (Figure 26). Apple also summarized a competitive analysis of the Amazon
Appstore in March 2011 email, identifying the competitive threat as “very high.” 19 Further, in
2018, Apple employees tracked Google’s launch of its Subscription Center and its possible
impact on the App Store. 20 A 2018 internal Apple presentation also calculated Apple’s market
share as a total of all game transactions, including game transaction platforms on console, PC,
and mobile, reflecting how Apple considers these other game transaction platforms to be
competitors to the App Store (Figure 27).
19
DX-4447
20
DX-4131
120. My analyses show significant variation in the typical business models and monetization
strategies for game apps relative to other apps. This variation reflects differences in competitive
conditions between game apps and other apps.
121. For example, I show in Figure 30 the proportion of apps by app genre in the App Store that
were free-to-download and did not have in-app purchases in FY2019. I looked at the proportion
of apps by app genre that did not monetize through the App Store or pay a commission to Apple.
While most apps in FY2019 (over 80%) do not monetize through the App Store, game apps are
more likely to monetize through the App Store compared to most non-game apps. Developers of
game apps therefore differ from developers of non-game apps in their ability and willingness to
monetize through other methods (such as the sale of physical goods, in-app advertising, or
payment outside of the app) that avoid paying a commission to Apple.
Percent of apps that are free-to-download without in-app purchases by genre (October 1, 2018 –
September 30, 2019)
122. The contrast between game and non-game apps is also apparent by looking at how much
revenue from in-app purchases comes from subscriptions versus non-subscriptions by app genre.
The results in Figure 31 are striking. Nearly all in-app revenue—98%—for game apps came
from non-subscription in-app purchases. The overwhelming majority of in-app revenue for non-
game apps, on the other hand, is from subscription in-app purchases. Ultimately, game apps
differ dramatically from other apps in how they generate in-app revenue, and combining game
and non-game apps into a single market would ignore these differences in how game and non-
game app monetize.
Percent of App Store in-app purchase revenue from non-subscription in-app purchases by genre
(October 1, 2018 – September 30, 2019)
123. The average commission rate paid by developers by app genre highlights the differences in
competitive conditions between transactions for game and non-game apps. I show in Figure 32
and Figure 33 the average commission rate paid by developers for in-app purchases and initial
downloads, respectively, by app genre. These average commission rates reflect how developers
monetize, such as whether developers offer free-to-download apps or whether they offer
subscription or non-subscription in-app purchases. The figures show meaningful variation in
average commission rates across app genres, highlighting differences in competitive conditions
and market outcomes between games and non-game apps.

Average in-app purchase commission rate by genre (October 1, 2018 – September 30, 2019)

Average initial download commission rate by genre (October 1, 2018 – September 30, 2019)
124. I also observed considerable variation in the average transaction price between app genres,
and, in particular, between game apps and other apps. The average transaction price for game
apps is $9.65, while the averages for other app genres range between $7.11 for photo and video
apps and $14.10 for health and fitness apps. The average download price also differs
substantially.
125. To determine whether it is appropriate to consider game app transactions as in the same
relevant market as non-game app transactions, I also considered whether game app transactions
are engaged in by distinct populations of developers. I found that developers that engage in
digital game transactions are not the same as the developers that engage in transactions for non-
game apps. Game developers (such as Epic) typically specialize in game development, not the
development of apps in general. Figure 34 shows that the share of game developers’ App Store
revenue from game apps is 88.2% but is only 11.8% for non-game apps. This analysis makes
clear that game developers tend to focus on game apps, not other apps.

App Store revenue earned by game developers (October 1, 2018 – September 30, 2019)
126. Looking at industry practices also shows how game transactions and non-game transactions
are in distinct markets. I show in Figure 35 that consumers browse games through a distinct
“games” tab in the App Store Google Play and the Amazon Appstore also separate games from
non-game apps in their storefronts.

Games category in transaction platforms on mobile devices
Source: DX-4774
127. Apple’s organizational structure also reflects how Apple separates game and non-game
apps: Apple employs two separate heads of App Store business development, one for games and
one for all other non-gaming app. Apple’s behavior is thus consistent with my view that game
transactions are a distinct market from non-game app transactions.
128. These industry practices, along with my empirical analyses showing differences in
monetization and outcomes across app genres, make clear that game apps are in a separate
market from non-game apps. It is therefore incorrect to cluster many potential separate markets
for apps into a single market that encompasses all apps.
***
129. All of the evidence I have presented shows that the relevant product market in this matter is
the market for digital game transactions. This market includes all platforms on which developers
and consumers can conduct digital game transactions, and it should not be limited to just
transactions on iOS. The relevant product market in this matter must also focus on game
transactions and should not cluster together multiple potential markets for transactions of
different types of apps.

IV. Market power: Apple lacks monopoly power for digital game transactions
130. I turn now to the question of whether Apple possesses monopoly or market power in the
relevant market. Economists define market power as the “ability of firms to raise price[s] above
the competitive level for a sustained period.” 25 A firm can also be said to have market power if it
can profitably restrict output. Firms are sometimes said to have “monopoly power” when
possessing a high degree of market power.
131. Epic and Dr. Evans’ claim that Apple possesses monopoly power is premised on an
incorrect definition of the relevant product market—iOS app distribution. As I’ve already shown,
the proper relevant market in which to assess the effects of Apple’s conduct on competition is the
market for digital game transactions. This market is demonstrably broader than the App Store,
including a variety of competing game transaction platforms.
132. Professor Lafontaine explains that, to assess whether a firm possesses market power in a
relevant market, economists evaluate both structural factors such as market shares, market
concentration, and ease of entry, as well as market outcomes such as price, output, quality, and
innovation.
133. I have conducted quantitative and qualitative analyses of both the structure of the market for
digital game transactions as well as available evidence on price, output, and innovation, and have
reached the opinion that Apple does not possess market power in the two-sided market for digital
game transactions:
First, the structure of this market is inconsistent with Apple possessing substantial
market power. Apple has only a small share—between 23.3% and 37.5%—and the
constant influx of new, innovative competitors shows that entry barriers are low.
Market participants are constrained by competition from other businesses, such as
direct distribution of games.
Second, economic evidence on key market outcomes—namely price, output, and
quality—demonstrates that Apple lacks market power in the digital game transactions
market. Commissions charged by Apple and other participants have fallen while
output has consistently increased. Innovation has been continuous.
134. To assess Epic’s claim that Apple possesses monopoly power, I conducted additional
analyses assuming for the sake of argument that Dr. Evans is correct and the market is limited to
iOS app transactions. Even then, I conclude that Apple is constrained by many factors from
raising prices above competitive levels. These include the ability of developers to adopt business
models that do not require paying a commission to Apple (such as in-app advertising or selling
subscriptions and other content outside of the iOS app) and competitive pressure from Android
mobile devices. Moreover, the evidence regarding price and output is the opposite of what one
would expect to see in a monopolized market—just as with game transactions, prices for all iOS
app transactions have not risen and output has exploded.
25
Jonathan B. Baker and Timothy F. Bresnahan, “Economic Evidence in Antitrust: Defining
Markets and Measuring Market Power,” in Handbook of Antitrust Economics, ed. Paolo
Buccirossi, (Cambridge, MA:MIT Press, 2008), p. 15
135. My analyses to support these opinions follows.
A. The structure of the market for digital game transactions is inconsistent with Apple
possessing market power
i. Apple’s market share of game transactions is inconsistent with market power
136. I begin my analysis of market power by looking at the structure of the market for digital
game transactions. The first piece of evidence regarding market structure is Apple’s share of the
digital game transactions market. I find that Apple’s share in this market is inconsistent with
market power.
137. As Professor Schmalensee explains, in a two-sided transaction market such as the game
transaction market there are multiple measures of output that could be used to calculate market
shares. One such measure is the total number of game transactions; however, this measure would
treat each transaction equally, regardless of price or value. Data on the total number of game
transactions is also not available market-wide. So I instead look at the dollar value of game
transactions to estimate market share, as recommended by Professor Schmalensee.
138. To calculate Apple’s share of the market for digital game transactions in the U.S., I
measured the dollar value of game transactions using consumer spending on games across
platforms. I computed the total game revenue on the App Store from the Apple transaction data,
and I use 2018, the most recent complete year of data available to me. I use two third-party data
providers to estimate the total market size: Newzoo and the Entertainment Software Association,
or ESA. 26
139. Since the third-party data include different measures of consumer spending in their
estimates, I performed several sensitivities which included different sets of consumer spending in
the total market size. On the one end, I take a very conservative approach in which I only
consider spending that is definitely within the boundaries of the market. For other estimates I
include other measures of consumer spending.
140. Across all of these sensitivities, I find that Apple’s market share in 2018 (the most recent
complete year of data) lies between 23.3% (including subscriptions and boxed games as part of
the overall market) and 37.5% (limiting the market to digital game transactions only). These
results are reflected in Figure 36.
26
DX-3248, DX-4217.020
Apple share of US digital game transactions market (2018)
Share
100%
90%
80%
70%
60%
50%
40% 37.5%
33.2%
27.4% 28.2%
30% 23.3%
20%
10%
0%
Data Source Newzoo Newzoo Newzoo ESA/NPD ESA/NPD
Revenue Included
Digital
Transactions
Subscriptions
Boxed Games
Source: DX-4797 (summarizing DX-5338, DX-3248, DX-3814, DX-4217 )
141. Using a conservative definition of the market for digital game transactions, which likely
overstates the correct figure, I estimate Apple’s share of the relevant market to be 37.5%. As Dr.
Evans acknowledges, other forms of distribution, including single-sided distributors, serve as
competitive constraints and should be included in the market. Accounting for these other
methods gives Apple a share of 23.3% or 27.4% of game transaction revenues, depending on the
data source. These estimated shares are inconsistent with Apple having substantial market power.
ii. Entry of new game transaction platforms is inconsistent with market power, as at least
seven new game transaction platforms have entered since the launch of the App Store and
new technologies provide additional ways for consumers to play and pay for games
142. Many new digital game transactions platforms that compete with the App Store have
entered the market since its launch, including Google Play, Samsung Galaxy Store, Windows
Phone Store, Nintendo eShop, Amazon Appstore, EA’s Origin, and the Epic Games Store. The
entry of these competing game transaction platforms limits Apple’s market power and is
evidence that potential barriers to entry have not excluded new entrants from the market.

a console or PC, stating that “the graphics were very high resolution and all the – the details were
PC-like and felt like being on a more powerful platform than a mobile platform.” 30
149. Overall, the entry, growth, and development of game streaming services indicate low
barriers to entry, and such services provide an existing and increasing competitive constraint on
the App Store.
iii. Apple faces competition from other forms of game distribution such as direct
distribution by developers and traditional retail
150. It is not just competing digital game transaction platforms that limit Apple from having
market power. Apple is also constrained from exercising market power by alternative forms of
game distribution, including the direct distribution of games by developers (including through
web browsers available to all iOS users) and distribution through traditional retail channels.
151. Direct sales by developers within iOS and on different devices. Developers have the
option to forgo transacting through a game transaction platform by offering game apps and in-
app content directly to consumers through their websites. For example, Epic provides Fortnite
directly to consumers from its website for use on Android devices and through the Epic Games
Store for PCs, Microsoft tablets, and Macs.
152. Additionally, and separately, Epic and other developers offer paid transactions on their
websites for in-game content, such as V-bucks, that can be accessed and utilized on iOS. Thus,
even though Apple does not allow game apps to be downloaded through an iOS web browser, it
is still constrained by that fact that consumers can make purchases from developers through an
iOS web browser.
153. Other game developers, such as the developers of Roblox, similarly provide games and in-
game content directly to consumers in addition to transacting through game transaction
platforms. 31 From an economic standpoint, these direct downloads and direct sales of in-game
content compete with digital game transaction platforms as well and further constrain Apple
from obtaining or exercising market power in the market for digital game transactions.
154. Traditional retail channels. Apple is also constrained by the sale of games through
traditional retail channels. These retails sales include sales of games on physical disks (often
called boxed products) or sales of download codes inside a box.
155. For example, Epic distributed Fortnite: Save the World as a boxed product for consoles
through traditional retail channels. Epic has also distributed boxed content containing
redeemable codes for V-bucks and in-game items through retailers. 32
30
Deposition of Mark Rein, February 10, 2021, pp. 107–108
31
DX-3755; DX-4473; DX-3748; DX-4552 ; DX-4292; DX-3262; DX-4390
32
Epic Games, Inc.’s Responses and Objections to Apple Inc.’s Second Set of Interrogatories, at
11–12 (Jan. 29, 2021)

156. Physical format games remain a meaningful proportion of total game sales: one estimate
from 2018 by ESA/NPD found that 17% of all video games were physical format sales. 33 These
physical format games sales would further constrain Apple from obtaining or exercising market
power in the market for game transactions.
157. The evidence demonstrates that the structure of the digital game transaction market is
inconsistent with Apple possessing substantial market power. Apple has only a small share of the
market, emergence of new competitors shows that barriers to entry are low, and other methods of
game distribution constrain Apple from exercising market power.
B. Analysis of market outcomes show that Apple has not increased prices for game
transactions, quantity has risen, and quality has improved over time
158. Using the framework described by Professor Lafontaine and Professor Schmalensee, I next
analyze market outcomes—such as price, quantity, and quality—to determine whether Apple has
market power in the digital game transactions market. The evidence I analyzed does not support
a finding that Apple has market power in this market.
i. Evidence shows that Apple has not exercised market power to increase commission
rates or prices
159. Apple charges a commission on game transactions through the App Store that generate
revenue for the developer. When Apple first launched the App Store, it set a commission rate of
30% on all such transactions. While the 30% commission rate remains for many game
transactions on the App Store, Apple has lowered it to 15% for certain types of transactions.
Through its history, Apple has never raised the commission for game transactions (or any
transactions) performed on the App Store. The 30% commission was set before Apple could
have conceivably had market power, even according to Dr. Evans.
160. Apple’s commission rates for game transactions on the App Store are competitive with
other platforms offering game transactions. The App Store’s typical 30% commission rate on
game transactions is common among not only a variety of game transaction platforms but also
digital content providers more broadly.
161. When the App Store launched, Apple’s 30% commission rate was the same as rates that
developers were paying to other digital game transaction platforms at the time. In fact, Apple’s
30% commission rate provided developers and publishers with a larger share of revenue than
they typically earned from sales at traditional retailers like GameStop, Amazon, or BestBuy.
From these traditional retailers, developers and publishers typically received only 45% of sales
revenues, much smaller than the 70% share they receive from the App Store. 34
162. To evaluate how Apple’s commission model compares to other game transaction platforms,
I identified the largest digital game transaction platforms and determined the standard
33
DX-4217.020
34
DX-3120.011

Also in 2016, Apple lowered its commission rate to 15% for members of the Video
Partner Program that stream premium TV content.
Starting in 2021, Apple lowered its commission rate to 15% for all developers that
generated less than $1 million in proceeds (net commissions paid to Apple) during the
previous calendar year from transactions through the App Store.
167. AppFigures (a mobile analytics provider) estimates that about 98% of developers generated
less than $1 million in proceeds during the previous calendar year, and thus qualify for the lower
commission rate. 46 Thus, contrary to increasing prices, as would be expected if Apple had
market power, many developers now face lower commission rates for game transactions on the
App Store.
168. While Epic may argue that the introduction of in-app purchase functionality on the App
Store in 2009 constituted a price increase for developers, this is incorrect. Prior to the
introduction of in-app purchasing, developers could not offer such services to consumers through
Apple. Thus, the introduction of in-app purchases was a new feature of the App Store. By
charging developers a 30% commission on such purchases, Apple simply gave that new feature a
price.
169. Game developers have also taken advantage of monetization strategies that allow them to
pay zero commission to Apple. At launch, about 32% of apps on the App Store were free. As
Figure 38 shows, an increasing majority of games are free-to-download and offer no in-app
purchases, and thus are not subject to any Apple commission. In FY2019, approximately 66% of
game apps were free-to-download and offered no in-app purchases and about 25% more were
free-to-download with in-app purchases. In total, 76% of games and 83% of apps with at least
one initial download on the App Store in FY2019 were free and did not have in-app purchases so
did not generate any commissions to Apple.
46
Ariel Michaeli, “Apple Just Changed the Rules, And (Almost) Everyone’s Getting a Bonus,”
AppFigures, November 26, 2020, https://appfigures.com/resources/insights/apples-gift-to-
developers-is-awin- win
Percent of App Store game apps by monetization type (July 10, 2008 – September 30, 2019)
170. The actual average commission rates paid by developers for initial game downloads have
also decreased over time or remained constant.
171. As discussed by Professor Schmalensee, the price paid in the context of a two-sided market
is the total price that Apple charges for the transaction. For the App Store, the total price is often
zero, such as for the download of free-to-download games.
172. Thus, to capture the price actually paid per transaction, I calculated the average commission
rate charged to developers. I assigned app transactions with no associated revenue a commission
rate of 0% in order to include these transactions in the “average commission rate.”
173. My analysis shows that because of the increasing prevalence of free-to-download game
apps, the average actual commission rate on initial game downloads has declined over time
(Figure 39).

Average commission rate for App Store initial game downloads (July 10, 2008 – September 30, 2019)
174. I also looked at the average commission paid to Apple in dollar terms from in-game
purchases. I show the average commission paid per in-game purchase transaction as well as the
average in-game purchase price in Figure 40. As can be seen in this exhibit, while the
commission rate on game in-app purchases has remained relatively constant over time, the
average commission in dollar terms paid to Apple has risen over time due to increases in the
average price set by developers for in-app purchases for games.

Average commissions and prices from App Store in-app game purchases (July 1, 2009 – September
30, 2019)
175. I also graph the median price set by developers for game in-app purchases in Figure 41. The
trend in the median in Figure 41is similar to the trend in the average shown in Figure 40. The
average and median price of in-app purchases charged by Epic, which I graph in Figure 42
likewise show an increase over time, especially after the launch of Fortnite. The trends in the
price of in-game purchases for developers in general and for Epic in particular confirm that the
increase in the average commission in dollar terms paid to Apple for game in-app purchases is
driven by higher prices set by developers rather than an increase in Apple’s commission rate.

Median price of game in-app purchases (July 1, 2009 – September 30, 2019)

Average and median in-app purchase price for Epic apps through the App Store over time
(December 21, 2010 – September 30, 2019)
176. I also considered evidence of market-wide pricing, including commissions charged by other
game transaction platforms. I find that commission rates have also declined on these competing
game transaction platforms.
In January 2018, (following Apple) Google Play reduced its commission rate to 15%
for all subscription renewals after the first year. 47
In November 2018, Valve, which is Steam’s parent company, lowered commission
rates for the largest developers. Commissions were lowered to 25% for developers
with revenues through Steam between $10 million and $50 million annually and to
20% for developers with revenues through Steam exceeding $50 million annually. 48
Most recently, in March 2021, Google Play announced that starting July 1, 2021, it
would decrease its commission rate to 15% on the first $1 million in developer
revenue. 49
47
DX-3966
48
DX-3599
49
Chaim Gartenberg, “Google will reduce Play Store cut to 15 percent for a developer’s first
$1M in annual revenue,” March 16, 2021,
https://www.theverge.com/2021/3/16/22333777/google-play-store-fee-reduction-developers-1-
million-dollars
177. These declines in the commission rates charged by other game transaction platforms are
inconsistent with Apple possessing market power.
178. As with the App Store, other transaction platforms also offer free-to-download games that
do not incur commissions. Game transaction platforms such as Google Play offer many free-to-
download games. Free-to-download games are not limited only to game transaction platforms for
mobile devices; free-to-download games for consoles and PCs have also increased over time.
179. Dr. Evans purports to have calculated an average effective commission rate on the App
Store of 27.7% in 2019. However, he fails to account for free transactions when performing his
calculation. As I showed above in Figure 39, initial game app downloads that are paid have
declined to less than 1%, meaning over 99% of initial game app downloads were free and did not
pay a commission to Apple. Dr. Evans inexplicably fails to account for the overwhelming
majority of game app download transactions when calculating the average effective commission
rate. In fact, his average effective commission rate would be the same if the number of free
initial game app downloads were 99%, like it is, or 1%.
180. These free initial download transactions all had a price—$0—but Dr. Evans’ does not
attempt to consider this price when calculating his average commission rate. Dr. Evans’ failure to
consider free transactions leads him to overestimate the average commission rate by a factor of at
least three for all apps. For FY2019, applying a proper measure of transaction price in a two-
sided market, I find that developers paid Apple an average commission rate of 8.1% for games
and 4.7% for all apps. 50
ii. Evidence does not show that Apple has exercised market power to reduce output
181. The evidence also shows that Apple has not exercised market power to reduce output in the
digital game transaction market.
182. Output on the App Store has increased since its launch. In the beginning, the App Store’s
U.S. storefront offered 452 third-party apps (including 131 game apps) by 312 distinct
developers. In the FY2019, there were over 300,000 game apps available on the App Store.
183. Figure 43 shows the total number of game transactions—both free and paid initial
downloads and in-app purchase—since the beginning of the App Store. It shows that the total
number of game transactions on the App Store has increased many times over. Between the year
following the App Store's launch in 2008 and FY2019, game transactions grew by over 1,200%.
Figure 44 shows total developer revenues from game apps through the App Store since its
launch. It shows that total revenue earned by developers has risen dramatically over time:
between 2010 and 2018, developer revenues from game apps grew by 2,600%.
50
DX-5469
App Store initial game downloads and in-app purchases (July 10, 2008 – September 30, 2019)
Developer revenue from App Store initial game downloads and in-app purchases (July 10, 2008 –
September 30, 2019)

184. Further evidence that Apple has not restricted output is the growth in the market-wide value
of digital game transactions over time. My analyses show that U.S. consumers spent $29.7
billion on digital game transactions in 2018, up from just $5.4 billion in 2010 (see Figure 45).
The share of game spending that happens digitally has also risen over this time period, from 31
percent to 83 percent. These increases are not limited to the App Store: worldwide developer
revenue from game downloads and in-app purchases has also grown
significantly over time, as I show in Figure 46. Between calendar years 2013 and 2018, total
annual game revenue for developers increased by
Amount spent by U.S. consumers on game transactions (2010 – 2018)
Source: DX-4808

185. Growth on the App Store has outstripped market-wide growth, providing additional
evidence that Apple has not restricted output. Market-wide spending on digital game transactions
grew around 448% from 2010 to 2018, but developer revenue for digital game transactions on
the App Store grew by more than 2,600% over the same time period, nearly six times the
industry rate. Simply put, these facts are inconsistent with Apple having market power.
186. Sales of iOS devices that can be used for game transactions have also increased since the
launch of the iPhone and the iPad. Since the exercise of market power to increase prices or
reduce quality of game transactions could reduce demand for iOS devices, this growth in devices
is another indicator that Apple has not exerted market power in the game transaction market.
187. In Figure 47 I plot unit sales of Apple iPhones and iPads in the U.S. from Apple FY2009 to
FY2019. Unit sales of these iOS devices experienced significant growth after launch and annual
sales of iOS devices continue to remain high. For instance, from 2009 to 2019, the number of
iPhones sold in the U.S. increased from 9.7 million in 2009 to in 2019. Apple also
sold iPads in the U.S. in 2019.

U.S. iPhone users (2012 – 2018)
189. The growth in devices for accessing game transactions has not been limited to iOS devices
and is consistent with increasing ownership of smartphones and tablets in the U.S. The
proportion of the U.S. adults who own a smartphone has increased from 35% in May 2011 to
81% in February 2019. 51 Similarly, the proportion of U.S. adults who own a tablet has increased
from 3% in May 2010 to 52% in February 2019. 52 The continued growth in the iOS platform in
particular and the continued growth in smartphone use overall is inconsistent with restricting
output but consistent with increases in overall economic value created by the iOS ecosystem to
consumers and developers.
190. In addition to the growth in smartphones and tablets, consoles continue to be popular,
including the introduction of new consoles. The prior generation of major consoles started with
launch of the Nintendo Wii U in 2012, followed by the Xbox One and PS4 in 2013. The present
generation of major consoles began with the Nintendo Switch in 2017 followed by the launch of
the PS5 and XSX in November 2020. Such consoles have experienced success recently. The
Nintendo Switch’s 2020 sales as of November were “the second highest of any [game console]
51
DX-4453
52
DX-4453

platform in U.S. history.” 53 Furthermore, the recently launched PS5 and XSX consoles
experienced strong sales since launch, with retailers frequently out of stock and unable to meet
the demand.
iii. Evidence shows that Apple has not exercised market power to reduce quality
191. Despite the evidence that game transactions on the App Store are growing faster than a
close competitor, Dr. Evans claims that Apple has failed to innovate or invest in improving the
quality of the App Store. Setting aside that Dr. Evans based this opinion on developer complaints
from surveys that Apple conducted specifically to help them understand developer concerns and
improve the App Store, Dr. Evans also errs by ignoring that the quality and value of the App
Store platform—as well as the app transactions it facilitates—are improved by Apple’s
innovation across its software and hardware ecosystem.
192. As Apple executives will testify, Apple has made enormous investments and innovations
that impact the quality of games delivered through the App Store. Among the long list of such
investments, Apple has invested in developing high performance graphics processing units and
the APIs and programming tools that enable developers, especially game developers, to use these
capabilities (such as the Metal graphics interface).
193. This investment has enabled much greater quality for mobile versions of graphics intensive
games such as Fortnite. Investments in alternative reality (AR) hardware, software, APIs, and
development tools (e.g., ARKit) have made it possible for new types of games to appear on iOS
devices, such as Pokémon GO.
194. The sheer variety of games on the App Store shows that Apple’s innovations across the iOS
ecosystem have improved the types and quality of games feasible on iOS devices. Many new
games (like Fortnite, PUBG, and Call of Duty Mobile) that previously only would have been
available on consoles and PCs are now available on iOS devices due to Apple’s innovations. At
the same time, top iOS games include a wide variety of different types of games. These games
range from content creation games like Minecraft to tower-defense games, board-game
adaptations, resource-management simulators, console game adaptations, and more. Mr.
Sweeney’s 2010 prediction that “over the next decade, iPhone and iPad games will grow ….”
has been proven correct. 54
195. Even looking just at store services, as Dr. Evans does, Apple has been a driver of
innovation. For instance, Apple was the first major platform to introduce in-app payments, the
now dominant form of game monetization in the App Store.
196. Developers understand that iOS is an attractive platform, especially for games, because of
how much revenue they can generate on iOS compared to other transaction platforms. For
instance, Fortnite data show that iOS generated the most revenue for Epic per hour of user play
time among all platforms (see Figure 49). This suggests that the App Store provides higher
53
Jeff Grubb, “Nintendo Switch is on a blazing sales pace for 2020 in the U.S.,” VentureBeat,
November 13, 2020, https://venturebeat.com/2020/11/13/nintendo-switch-is-on-a-blazing-sales-
pace-for-2020-in-the-u-s/.
54
DX-4620.002
quality from the developer perspective as indicated by the ability to more effectively monetize
play time than other alternatives. The fact that consumers are willing to spend more per hour
played on iOS than on other platforms also indicates greater relative value.
Worldwide Fortnite revenue per hour played by platform (May 2020 – July 2020)
Platform Revenue per hour

1. iOS $0.22
2. Xbox One $0.19
3. Google $0.18
4. Android $0.16
5. Switch $0.15
6. PS4 $0.13
7. PC $0.10
8. Other $0.10
197. All of the evidence presented on market outcomes presents a clear story. Apple has not
raised prices, Apple has not restricted output, and Apple has not reduced quality in the game
transaction market. This is entirely inconsistent with Apple possessing market power in the game
transaction market.
iv. Purported estimates of the App Store’s profits do not show Apple has market power
198. Dr. Evans asserts that his measures of the App Store’s profit margin demonstrate that
Apple has market power. I disagree. As Professor Schmalensee and Professor Lafontaine
discuss, profit margins do not provide insight into whether a firm has market power.
199. Moreover, the analysis Dr. Evans’ conclusion is based on is flawed. Among other flaws, Dr.
Evans compares what he claims to be the App Store’s profit margin to five allegedly comparable
companies selected by another Epic expert, Mr. Barnes. However, none of these five companies
are truly comparable. Unlike the App Store, these five companies are platforms for transacting
physical goods or services, not for digital transactions. These five companies also differ from the
App Store along other dimensions, such as whether the platforms are integrated into hardware
provided by the company, the demographics of platform users, the geographic markets in which
they compete, and brand value. These comparable companies cannot be used as benchmarks to
determine what Apple’s profits would be in any but-for world, and without a valid benchmark,
Dr. Evans cannot conclude that Apple earns outsized profits from the App Store.

C. Apple’s ability to exercise market power would be limited by a number of factors

even if the relevant two-sided market were limited to iOS apps
200. Even if one considered, contrary to the evidence, a two-sided market limited to the
distribution of iOS apps (but not limited to games), Apple would still be constrained from
exercising market power in that market by many factors.
i. Developer choice constrains Apple from exercising market power
201. Game developers (and developers of other apps) face few restrictions on monetization. Like
many game developers, developers of popular iOS apps have chosen to let consumers pay for
content outside of the iOS app but still consume the content within the app, which constrains
Apple’s ability to exercise market power and charge supracompetitive prices. Amazon Prime
Video, YouTube, Microsoft Office, and Netflix are just a few examples of popular apps that
employ this model.
202. Developers can also incentivize consumers to purchase content outside of the iOS app. For
instance, Apple allows developers to choose how they would like to price their apps and app
content, including how they price across different transaction platforms and how they price when
selling directly to consumers. Some developers have chosen to charge consumers more within
the iOS app than in other avenues for payment, encouraging consumers to shift their purchases to
these other locations.
203. Many developers choose to allow consumers to purchase content outside an iOS app but
allow those users to use the content within the app. My analysis of top apps on the App Store
shows that many allow consumers to purchase content outside of the App Store that can be used
within the iOS app, which does not require the developer to pay a commission to Apple. I
illustrate this analysis in Figure 50. In addition to game apps such as Fortnite, the list of the top
25 App Store apps by revenue includes streaming apps such as YouTube and Netflix and dating
apps such as Tinder and Bumble, while the list of top 25 App Store apps by downloads includes
Uber, Amazon Shopping, Snapchat, Instagram, and Facebook.

Smartphone OEMs offer services that facilitate transferring data from an iOS device
to a new Android device.
As I have already shown, most top apps are available on both Android and iOS, and
many developers allow users to transfer content between devices through single-sign
on.
Multiplatform services such as WhatsApp and Zoom provide alternatives to Apple’s
iMessage and FaceTime, and studies confirm that having friends or family that use
iOS devices is not a significant reason why iOS device owners choose not to switch
to Android devices. 63
D. Market outcomes, such as price and quantity, are inconsistent with Apple exercising
market power in an alleged iOS app distribution market
215. While the relevant market in this matter is the market for digital game transactions, I also
assumed for the sake of argument that Dr. Evans is correct and that the relevant market is his
alleged market for iOS app distribution. I then analyzed market outcomes such as price and
output to determine whether Apple possesses market power in this alleged market. I find that the
evidence is inconsistent with Apple possessing market power even if Dr. Evans’ flawed market
were correct.
216. Apple does not charge a supracompetitive commission rate even when considering an
incorrect market for iOS app distribution. These transactions are subject to the same commission
rates as game transactions, which Apple has only lowered over time.
217. Additional evidence based on the Apple transaction data backs up the fact that prices have
declined over time, not increased.
Figure 51 shows that the percent of initial downloads that are paid on the App Store
has declined over time and that the average commission rate on initial downloads for
all apps on the App Store has declined over time.
Figure 52 shows that the average commission rate on in-app purchases for all apps on
the App Store has declined over time.
63
DX-3598.027
Average commission rate for App Store initial downloads (July 10, 2008 – September 30, 2019)
Average commission rate for App Store in-app purchases (July 1, 2009 – September 30, 2019)

218. Just as Apple has not restricted output for game transactions, Apple has not restricted output
for transactions on the App Store for apps in general. In Figure 53 and Figure 54, I show that the
total number of transactions and total developer revenues for all apps on the App Store over
time. The increase over time is dramatic. Since the launch of in-app purchases in mid-2009, the
total number of annual transactions for apps has increased nearly 500% and total revenue earned
by developers has increased by over 3,700%.
App Store initial downloads and in-app purchases (July 10, 2008 – September 30, 2019)

Developer revenue from App Store initial downloads and in-app purchases (July 10, 2008 –
September 30, 2019)
219. Dr. Evans claims that Apple obtained monopoly power in iOS app distribution around 2010
(Evans Deposition, p. 334, Evans WDT, ¶91). If that were true, one would expect to see Apple
restricting output or raising prices around that time. But the data show the exact opposite. The
average commission rate on both initial downloads and in-app purchases continued to decline
after 2010, and the growth in free-to-download apps only accelerated. The graphs on game
transactions and developer revenues make obviously clear that transactions and revenues
continued to increase after 2010 without any decline in the growth rate. In fact, as mentioned,
total annual game revenue for developers from mid-2009 to FY2019 increased by more than
3,700%, from $282M to $10.9B. In the period from the first quarter of FY2010 to the first
quarter of FY2011, total game revenues for developers increased from $30.1M to $69.4M, or
130%, which is again inconsistent with Apple exercising market power.
220. Overall, Dr. Evans’ assertion that Apple obtained monopoly power in iOS app distribution
around 2010—or at any point, for that matter—is completely contradicted by the data.

V. Dr. Evans’ market definition and market power conclusions are wrong factually and for
this case
A. Dr. Evans incorrectly concludes that there are no good substitutes to the App Store
for consumers and developers to conduct digital game transactions
221. Dr. Evans’ approach to market definition, and hence market power, is unreliable and
inconsistent with the market facts described above. Dr. Evans ultimately concludes that there are
no substitutes for consumers or developers to transact through the App Store with each other for
games and apps. This flows from his opinion that the vast majority of consumers who own an
iPhone are locked into owning one for as long as they own a smartphone, and that these same
consumers have no substitute options for making game transactions or app transactions in
general other than through the App Store on their iPhone. My analysis above makes clear that
this is not supported by market facts or data.
222. Dr. Evans reaches unreliable conclusions because he does not start with the product at issue,
game transactions—or even app transactions in general—and consider what substitutes
developers and consumers have for making them other than through the App Store. Instead, his
analyses are premised on the faulty assumption that the only way consumers and developers can
substitute for game transactions made in the App Store is if other devices are themselves
substitutes for an iOS device at all times of day and with the same device features. 64 Dr. Evans’
analysis is divorced, both theoretically and factually, from the economic realities of this market,
as I discuss in turn below.
B. Dr. Evans is mistaken to claim that devices need to be direct substitutes to allow for
game, or app, transaction substitution
223. Dr. Evans appears to accept that the App Store provides transactions between consumers
and developers, yet his market definition analysis does not focus on this product or the
substitutes available for it. The product at issue in this matter is digital game transactions
between consumers and developers as that is what Apple provides simultaneously to Epic and
Epic’s customers. The economic question for market definition and market power, then, is to
what degree Apple faces competition in providing digital game transactions to developers and
consumers. Dr. Evans does not assess the relevant economic question and instead defines two
markets—a foremarket for smartphone operating systems and an aftermarket for iOS app
distribution—which are inappropriate and uninformative.
224. For his alleged foremarket for smartphone operating systems, Dr. Evans claims that
consumers and developers do not have practical substitutes for smartphone operating systems
because, in his opinion, other devices (such as computers, game consoles and even tablets using
a similar operating system) are not substitutes to smartphones for consumers. From this he
concludes that consumers do not have substitutes for smartphone operating systems for using
apps. This is faulty logic. Neither operating system substitution nor device substitution is
necessary for game, or app, transactions across platforms to be substitutes. As I have shown
above, consumers and developers can substitute game transactions across a variety of transaction
64
Evans Direct Testimony, ¶¶ 50-53
platforms as well as through websites on an iOS device. Dr. Evans’ focus on device substitution
leads him to misrepresent these options available to consumers and developers for making game
transactions. For example, Dr. Evans acknowledges that consumers generally own multiple
devices on which to make game transactions and app transactions. 65 Yet he argues that this
undisputed fact, which obviously increases the ease of multi-homing across different transaction
platforms, is instead evidence of consumer inability to substitute the devices for each other, and
thus, as his faulty logic follows, to use them to substitute game or app transactions. This is
unreliable economic analysis.
225. For his alleged aftermarket for iOS app distribution on smartphones, Dr. Evans finds that
Apple is a monopolist in his alleged iOS app distribution market. He does so by assuming that
consumers cannot easily switch from owning iOS smartphones to owning other smartphones,
ignoring that consumers often own multiple devices and can make app transactions on those
devices, and assuming that consumers cannot make game transactions on websites and other
game transaction platforms in substitution for making game transactions through the App Store.
As I have shown previously, none of these facts is true.
C. Dr. Evans is wrong to assume that game transactions cannot be substituted across
differentiated devices
226. Even on their own terms, Dr. Evans’ market definition is flawed. He offers limited bases for
opining that consumers lack good substitutes for smartphone apps on other devices, and that, as a
result, developers are unable to replace a substantial part of consumer demand for their services
by turning to substitutes for smartphones apps. He claims that developers “must follow the
customer to their smartphone OSs” because:
Smartphones have “unique” features—mobility, constant Internet connection, camera,

GPS, and accelerometers.
There are portions of the day during which consumers may only have access to their
smartphones in order to access the internet (or might find it “inconvenient” to use
another device in their home).
Developers cannot create web apps through mobile browsers that are good substitutes
to native apps.
227. Dr. Evans provides no empirical analysis to support these assertions. Most importantly, he
does not assess or establish the relevance of these features for the ability of developers and
consumers to substitute across them to make game transactions. Thus, Dr. Evans sets a bar for
substitution that makes no sense in the context of apps and games. Apple faces competitive
constraints in its pricing for game transactions from many sources, not all of which offer every
service or feature that an iOS device offers. Dr. Evans does not show for example, that many (if
any) game apps, or apps in general, require mobility, constant Internet connection, camera, GPS
or accelerometers in order to make game or app transactions. Nor does he acknowledge that
developers have choices about what kinds of games they design and what features those games
65
will have. None of the apps that he states work better when you have them “all the time” (e.g.,
Uber, Tinder, Snapchat, Google Maps, and Chase Bank) are game apps.
228. The empirical evidence I have shown above fills in this gap for Dr. Evans. Developers can
reach consumers to transact outside of the App Store because consumers multi-home in their
game play across devices and use different game transaction platforms. Game developers have,
and actively pursue, these many different options for reaching consumers. Obviously, Fortnite
does not require Dr. Evans’ “unique” smartphone features for its game, as it is available on PCs,
tablets, consoles, and smartphones. And consumers clearly do not find it “inconvenient” to make
Fortnite game transactions on those other devices, as even for those Fortnite players who have
chosen to open an iOS account, over 90% of their game play and game spending is on PCs,
tablets, and consoles. I also showed that the top game apps, and the top apps in general, both by
revenue and by download, are offered on many different types of devices and thus clearly do not
require this “unique” set of features to make transactions with consumers.
229. Operating systems and devices do not need to be perfect substitutes in all functions to
compete with other operating systems and devices for a given function. As shown above, Apple
faces competitive constraints in its pricing for game transactions from a variety of sources, not
all of which offer every service or feature that an iOS smartphone does. Competition occurs in
differentiated products, and the fact that products are not identical does not imply a lack of
competition. Taking a brick-and-mortar example, multi-line retailers such as Target or Walmart
that sell products across a variety of departments compete with specialty retailers that customers
can instead visit to purchase the same or similar products. These multi-line stores compete with
grocery stores in the grocery department, sporting goods stores in the sporting goods department,
and clothing stores in the clothing department. Whether or not a local multi-line retailer faces a
competitor with a comparable selection of departments does not change the fact that this multi-
line retailer will compete with local specialty retailers.
D. Dr. Evans’ smartphone operating system market incorrectly ignores tablets
230. An especially telling example of Dr. Evans’ general approach of assuming away the many
options for developers and consumers to make game transactions is his exclusion of tablets from
both his smartphone operating system and iOS app distribution markets. His basis for this is a
statement that tablets “provide users with different functionality than smartphones.” 66
231. There is no question that tablets can be and are used to make game transactions. One of the
primary use cases for tablets is mobile gaming, consistent with the fact that Epic offered Fortnite
on the iPad. Additionally, some tablets share many of the unique features Dr. Evans assigns to
smartphones. Also, many tablets use the same operating system as smartphones, and have
functions that are integrated with phones that use the same operating system.
232. For example, consumers can transact on both iPhones and iPads through the App Store;
developers use the same tools to create apps for iPhone and iPads, as Dr. Evans notes, and
developers can provide apps that are compatible with both iPhones and iPads.. Also, many apps
on the App Store can be downloaded on either an iPhone or an iPad, and when consumers
66
Evans Direct Testimony ¶ 43, footnote 3
purchase a paid-to-download app on an iPhone, they can install it on an iPad (and vice versa)
without making an additional purchase. Hence pricing is directly connected for many apps.
233. Tablets are an important example of the manner in which Dr. Evans’ approach simply
assumes away key alternatives for game transactions. Tablets are also important because Dr.
Evans’ opinion that the purported smartphone operating system market is a “duopoly” of iOS
and Android is based on an assumption that tablets do not serve as substitutes for transaction
platforms on smartphones. Dr. Evans has offered no evidence to support this assumption. Had he
included tablet operating systems in his operating system market definition his “duopoly” would
disappear. iOS and Android are not the only operating systems for tablets. Other tablet operating
systems include Amazon Fire OS and Windows. And, as with smartphones, there are many tablet
OEMs including, among others, Amazon, Lenovo, Samsung, and Microsoft. Dr. Evans’
exclusion of tablets exemplifies the unreliability of his conclusions about market definition and
thus Apple’s market power.
E. Epic’s experts assume that differentiated game transaction platforms cannot

compete and in doing so ignore the vast majority of developer and consumer choices
234. I understand Dr. Cragg (another of Epic’s experts) may offer an analysis of the degree of
overlap of identical games between mobile and consoles/PCs, concluding that this overlap is too
small to be consistent with substitution for game transactions between these various devices and
game transaction platforms. Dr. Evans also makes this point (albeit cursorily) in his testimony. 67
Such a conclusion would be unreliable.
235. This analysis would set a bar for substitution that makes no sense in the context of apps and
games. Apple faces competitive constraints in its pricing for game transactions from many
sources, not all of which offer every game that developers currently choose to offer on iOS.
Again, competition occurs in differentiated products, and the fact that products are not identical
does not imply a lack of competition.
236. The position that only identical games with identical functionality across game transaction
platforms can be “substitute” game transactions between developers and consumers would be an
extreme one. It would assume, for instance, that even very similar games like Call of Duty
Mobile and Call of Duty Warzone (its PC/console counterpart) cannot serve as substitute game
transactions because they are not exactly the same on different platforms—despite being created
by the same developer and offering as similar a gameplay experience as their names would
imply.
237. Such an analysis would also assume that competition for game transactions between
developers and consumers across game transaction platforms cannot happen across different
games. This would be economically wrong, as is evident from the fact that many consoles offer
exclusive games that are unavailable on other consoles. 68 If all transaction platforms had to offer
the same set of games to be substitutes, then the Nintendo Switch and PlayStation 4 would not
67
Evans Direct Testimony, ¶ 53(iii)
68
Matt Purslow, “Gears of War 3 Was Ported to PS3, but Was Never Going to See Release,”
IGN, May 20, 2020, https://www.ign.com/articles/gears-of-war-3-ps3-port
compete with each other for game transactions because they offer many games that are only
available on one device or the other.
F. Dr. Evans ignores the ability to purchase content on one platform and use it on
another, as well as the relevance of websites as a form of competitive constraint
238. The generally mistaken approach by Dr. Evans is exacerbated by his consistent conflation
of the use of smartphone apps with the ability to perform game transactions (or app transactions
more generally). This is because acquisition of in-game content can often occur at a different
time, and on another platform, from its use.
239. Thus even if there were some games or apps that required what Dr. Evans claims are the
unique features of a smartphone, to be used at special “portions of the day,” the developer could
provide them and the consumer could use them in exactly that manner, but acquire that content
separately. Developers are not constrained from charging their customers outside of the App
Store. If a developer and consumer would prefer to transact outside of the App Store in order to
avoid the relevant commission, the developer can offer this option to consumers on another
transaction platform as well as on a website. Because iOS smartphones have web browsers, this
means that developers can transact with a consumer any time the consumers has their iPhone,
even if the app itself cannot be used or played in a web browser.
240. The fact that game and app developers choose to design subscriptions, in-app currency,
single sign-on and portability that work across transaction platforms is evidence that they are
both used and valued for cross-platform purchases. In general, developers can monetize a game
or app such that transactions can be highly substitutable across transaction platforms,
independent of whether smartphone apps are good or poor substitutes for apps on other devices.
241. Dr. Evans in fact disregards all competition from websites, even though websites are
accessible on iOS devices just like the App Store. He does so by asserting, without any empirical
evidence, that developers cannot create web apps accessible through the mobile browser that are
good substitutes to native apps. He then, inexplicably, lists developers who have chosen to offer
apps on smartphones but continue to operate successful websites on personal computers (e.g.,
Facebook, Match, Twitter, LinkedIn, YouTube, the Wall Street Journal, Spotify, and Hulu) as
evidence of developers’ inability to offer a website version on iOS. This is getting the economics
utterly backwards—in these examples developers have shown that they can both have a
successful website and an app on an iOS device. Moreover, even if Dr. Evans were right,
developers are free to offer the ability to acquire content through a website, separate from game
play or app use.
G. Dr. Evans’ economics are confused when he states that growth is evidence of lack of
substitution
242. Dr. Evans suggests that the fact that smartphones are now ubiquitous, that consumers have
increased the amount of time they spend online on smartphones, and that there has been a large
increase in apps written for smartphones is evidence of the lack of substitutability of other
devices for app transactions between developers and consumers. 69 He claims that the growth of
69
Evans Direct Testimony, ¶¶50–56
developers choosing to offer smartphone apps is because they cannot serve consumers on other
devices.
243. The growth of developers choosing to offer smartphone apps is the hallmark of the
competitive value that smartphones offer developers. To suggest, as Dr. Evans does, that this
reality is instead a sign of an exertion of market power directly clashes with basic economic
theory, which would expect a decline in quantity, not an increase.
H. Dr. Evans’ Fortnite case study cannot support his market definition
244. Dr. Evans assesses a case study of Fortnite users’ behavior after the Hotfix event that led to
new versions of Fortnite no longer being available for download on the App Store and the
Google Play store. He claims this study “confirm[s] key conclusions concerning foremarket and
aftermarket competition in this matter,” and therefore supports his market definitions. 70 Dr.
Evans’ Fortnite study is the only analysis of actual consumer level market data he uses to support
any of his market definition calculations. (He otherwise relies on a survey by Professor Rossi,
which I understand will be discussed by Professor Lafontaine.)
245. Dr. Evans’ Fortnite study frames and asks a narrow question: where did iOS users who
played Fortnite spend their time and money in regard to Fortnite after Fortnite updates were no
longer available on the App Store? An answer to this question by definition cannot depict the
degree to which users can turn to alternative game transaction platforms in the face of a non-
transitory market wide price increase by the App Store or by smartphone game transaction
platforms. This is because the change in circumstance of a single game—Fortnite—does not
measure consumers’ responses to a non-transitory market-wide change in price in the market
setting relevant for this case.
246. Dr. Evans’ case study cannot proxy for the degree of switching that consumers of game
transactions would generally take up if Apple had increased the price for all game transactions
(let alone all apps in general), in part because it does not reflect the full set of options available to
developers and consumers following such a price increase. Consumers and developers would
generally be able to both continue to use game apps on an iOS device but substitute payment for
those game apps to another game transaction platforms or shift both app use and payment to
another game transaction platform and device. Consumers could also substitute game
transactions to a website that users can access through their iOS browser but that would not lead
to commissions for Apple. Additionally, Dr. Evans’ conclusion ignores that many consumers in
his case study may have simply substituted away from Fortnite to other games that remained
readily available on iOS and that did not have a change in circumstance.
247. Dr. Evans has not empirically tested what would happen if the App Store, or all smartphone
game transactions platforms, jointly raised their prices for all game transactions. Dr. Evans’
calculation also does not allow one to understand the effect of a non-transitory price increase in
game transactions because such a change does not necessitate depriving developers of a
smartphone app: It only affects the prices paid for in-app purchases or downloads, while free
downloads are not affected at all.
70
Evans Direct Testimony, ¶¶ 124-134
I. Dr. Evans overestimates the degree of iOS single homers and underestimates their
degree of switching game transactions
248. Based on an analysis of a sample of Fortnite user data, Dr. Evans has concluded that iOS-
only users play the majority of iOS Fortnite minutes played and that following the Hotfix, iOS-
only users in the U.S. switched their purchases to non-iOS platforms. While Dr. Evans’
calculation overstates the degree that iOS Fortnite users single-home their purchases on iOS, he
nevertheless finds that these single-homing users substituted to other transaction platforms after
the Hotfix.
249. The proportion of single homers. According to Dr. Evans, in the U.S., iOS single-homers
accounted for 65.0% of iOS Fortnite revenue during the 32 week period prior to the Hotfix
event. 71 In this calculation, Dr. Evans unusually categorizes as iOS single-homers those iOS
users who played Fortnite on multiple platforms but chose to make all their purchases through
the App Store. When I calculated the share of iOS single-homers using the appropriate definition
of single-homing (one based on whether users accessed Fortnite on multiple platforms) and a
dataset that includes all Fortnite user accounts (rather than a sample like Dr. Evans’ data), the
share of revenue from iOS single-homers is substantially lower at 38.6%. My estimate more
accurately reflects the degree of revenue that is represented by single-homing Fortnite users.
250. Dr. Evans provided an additional statistic on single-homing among Fortnite iOS users that
is likewise uninformative. He calculated that for players who use iOS as their “primary platform”
(i.e., by his definition, those who spent most of their Fortnite playing time on iOS devices),
90.9% only ever used Fortnite on iOS. 72 This statistic is uninformative—it is equivalent to the
tautology, “when I focus on users who have a preference for playing Fortnite on their iOS
devices, a significant share prefer playing the game only on their iOS device.” But worse than
that, the statistic is misleading: It is undisputed in the Fortnite data that iOS Fortnite users spent
the vast majority of their time and money with regard to Fortnite on non-iOS devices and
through game transaction platforms other than the App Store (as I show in Figure 14).
Additionally, Dr. Evans has not defined a market centered around only “primary users” of iOS
apps—to the contrary, he claims that all consumers of all iOS apps are locked-in to the App
Store.
251. Switching of iOS users after Fortnite Hotfix event. Dr. Evans estimates that after the Hotfix,
U.S. iOS-only app users (by his definition described above) shifted almost one-third of the total
spending they would have made on iOS in the post-period to consoles and personal computers in
little more than two months. 73 While he and I do not take the same approach when using the
Fortnite data or estimating the amount of substitution to other transaction platforms after the
Hotfix, it is important to understand that what he has found is that iOS-only app users who had
never made a purchase outside the App Store began making purchases through other game
transaction platforms on other devices. This is substitution.
71
72
73
Evans Direct Testimony, ¶¶ 127, 132

252. Dr. Evans continues on to adjust the amount of substitution he finds for this group
downward. Why? Because it is his opinion that over time some iOS-only players, like consumers
in general, will naturally decide to buy game consoles and PCs on which they can, and do, make
game transactions on Fortnite. 74 This shows substitution as well, as these consumers have now
chosen to make payments on an alternative game transaction platform for content that can be
used on the iOS Fortnite app. His attempt to account for the fact that some iOS-only Fortnite
users would have moved to consoles and PCs even absent the Hotfix reduces his estimates of the
share of iOS-only revenue prior to the Hotfix that switched to other platforms by almost half, to
16.3%. 75 Regardless of the reliability of his calculation, his finding implies both that substitution
exists even for users who previously only played Fortnite on iOS and that there are substantial
increases in iOS-only users adding payments through PCs and consoles over time even without a
Hotfix event.
74
75
Game developers and game transaction platforms for PCs and Macs—notably
including the Epic Games Store—often enter into exclusive relationships, which
limits consumer choice and reduces consumer value.
As Dr. Evans notes, many top game developers have opened their own transaction
platforms to distribute directly to consumers and have removed their games from
other transaction platforms, increasing search costs for consumers who want to play
these developers’ games.
Competition is skewed. While some of the largest game developers have extensively
used direct distribution of games to consumers or negotiated a lower commission rate
with existing platforms, small game developers usually cannot, putting them at a
competitive disadvantage.
Prices of digital game transactions are not lower. While the Epic Games Store charges
12%, it operates at a loss, is not projected to make a profit for at least the next two
years, and has grown slower than the market as a whole. Steam’s reduced
commission rate only applies to the largest developers, and its average effective
commission rate in 2019 was estimated to be . 80
B. Professor Athey proposes the theory that forced interoperability will lead to pro-
competitive benefits without conducting any cost-benefit analysis or assessing case
specific information
257. Professor Athey provides a conceptual analysis of “middleware,” namely “multi-platform

app stores,” and their potential role in app delivery. She imagines that if there were greater cross-
platform interoperability, competition across platforms would increase, benefiting consumers
and developers. Professor Athey’s theoretical argument assumes developers would no longer
need to develop apps for multiple operating systems. Her argument also implies a form of forced
interoperability for operating systems: under her framework all devices and operating systems
(including those for Nintendo, Sony, and Microsoft consoles) would be required to be
interoperable with each other.
258. She assumes that a hypothetical “multi-platform app store” would offer meaningful
middleware solutions that cannot currently be offered separately and are not currently available
or offered by Apple itself.
259. It is unlikely that “multi-platform app stores” would have a significant impact on developer
multi-homing costs because multi-platform app stores’ provision of technical middleware in
open platforms is limited. “Multi-platform app stores” may provide store-specific APIs that
extend across devices. But even with these APIs, a developer must still develop a game or app
and optimize it across different operating systems. Store-specific APIs do not solve that
challenge.
260. Many features highlighted by Professor Athey that are provided by “multi-platform app
stores,” such as chat, are not relevant to developing cross-platform games and would not reduce
developer “multi-homing costs.” Others do not require a “multi-platform app store” at all. For
80
DX-5322

also offers forms of technical middleware that help developers to develop games across its own
platforms. Professor Athey’s theoretical “economic middleware” concern is, in short, a solved
issue.
266. Finally, like Dr. Evans, Professor Athey fails to consider the many benefits that the “walled
garden” nature of the iOS ecosystem and the App Store provides when arguing that “multi-
platform app stores” or “economic middleware” would provide a competitor to the App Store
that would reduce Apple’s market power. As Apple’s fact and expert witnesses will testify,
Apple’s business model serves to assure the safety, security, privacy and quality of the iOS
ecosystem. Apple’s “walled garden” ensures that all apps on the App Store have been reviewed,
both by an algorithm and by a human, to guarantee apps are safe, secure, and high quality. This
promotes trust in the overall iOS ecosystem, benefitting both consumers and developers. A but-
for world in which other app transaction platforms existed on iOS devices, including app
transaction platforms that do not insist on the same level of quality, security, safety, and privacy,
would erase these benefits, harming consumers and developers.
C. Apple has not caused substantial anticompetitive effects: prices have not increased,
quantity has risen, and quality has improved over time
267. No Epic expert has offered empirical evidence that the Apple policies and conduct Epic
challenges has had any anticompetitive effect in the digital game transaction market—or on
competition for app transactions more broadly. On the contrary, the empirical evidence shows
that this market is thriving by any measure. I summarize several data points showing that the
market is thriving in Figure 55.

Summary of App Store statistics
Entry
New Game Transaction Platforms
7+
(since launch of App Store)
New Game Streaming Services
4+
(through iOS Web Browser)
Price
Game App Effective Games Commission Rate
8.1%
(FY2019)
All App Effective Commission Rate
4.7%
(FY2019)
Game App Initial Download Commission Rate
0.2%
(FY2019)
Quantity
Increase in Number of Game App Transactions
1,200%
(mid-2008 – FY2019)
Increase in Game Developer Revenue
2,600%
(2010 – 2018)
Increase in Game Developer Revenue
130%
(FY2010 Q1 – FY2011 Q1)
Increase in Number of All App Transactions
500%
(mid-2009 – FY2019)
Increase in All App Developer Revenue
3,700%
(mid-2009 – FY2019)
Source: ¶¶129, 131, 167, 170, 205, 206, Figure 33
268. First, I have seen no evidence of supracompetitive pricing. Apple’s commission rate for
game transactions on the App Store is competitive with other platforms offering game
transactions, and Apple has lowered its commission rate to 15% for many transactions. 83
269. On average, the shift from paid to free-to-download apps has caused Apple’s average
commission rate on initial game app downloads to decrease to less than 1%. And Apple’s
average commission on game app in-app purchases has stayed constant. In 2019, Apple’s
effective commission on game transactions was 8.1%, and on all app transactions, it was 4.7%.
83
Lauren Goode, “Apple’s new subscription offerings are now available to App Store
developers: The changes were first announced in June,” The Verge, September 2, 2016,
https://www.theverge.com/2016/9/2/12774758/apple-developers-app-store-new-subscription-
rules
These numbers will only decrease as a result of the commission reductions for small developers
introduced in the new App Store Small Business Program.
270. Second, there is no evidence of reduction in output. The total revenue of digital game
transactions has risen for the market overall, and the number of transactions and developer
revenues on the App Store have also increased, growing at a faster rate than the market overall.
On the App Store, the amount of revenue earned by game developers has increased by 2,600%
between 2010 and 2018, while developer revenues from all app transactions have increased by
3,700% between mid-2009 and FY2019. And consumers can choose from a variety of
high-quality apps; even among game apps, they have a variety of options in each game genre.
271. Third, there is no evidence Apple has harmed competition by reducing quality. Apple has
not reduced the quality of transactions on the App Store and has made significant investments in
innovation that improve the quality of the iOS ecosystem and App Store transactions. In sum,
there is no evidence that developers or consumers have suffered substantial anticompetitive
effects.
VII. Oath
I declare under penalty of perjury under the laws of the United States that the foregoing
is true and correct.
Respectfully submitted,
April 23, 2021
Word count: 25626

OAKLAND DIVISION

FRANCINE LAFONTAINE, PH.D.
vs.

05640-YGR-TSH Francine Lafontaine, Ph.D.
I. Summary of opinions
1. As an economist and former antitrust enforcer, I offer opinions in three areas: product market
definition, geographic market definition, and market power. As explained below, my overall
conclusion is that Apple does not possess monopoly power or charge supracompetitive prices in
the relevant market for digital game transactions in the United States.
Product market definition
2. Opinion 1. The App Store is a two-sided transactions platform (as Dr. Evans acknowledges).
As a result, market definition must be evaluated from the perspective of both consumers and
developers, and the product is transactions, i.e., the analysis must evaluate where content is paid
for, not where it is used (which can be different) (§ III.A, p. 5).
3. Opinion 2. Epic’s experts have failed to perform the analysis that would permit a finding that
all iOS apps—many of which are not substitutes for one another, such as game apps and other
types of apps—may be clustered together in the same antitrust market. Distinct product markets
can only be clustered for purposes of market definition if competitive conditions are similar for
all product markets that are grouped into the cluster market (§ III.B.ii, pp. 7–8). Since games
face different competitive alternatives from other types of apps, it follows that grouping all types
of apps together, as Epic proposes to do, would be inappropriate as a cluster market
(§ III.B.iii, p. 8).
4. Opinion 3. Game transactions on digital transaction platforms constitute a relevant product

market that is distinct from other types of app transactions, and inconsistent with Epic’s attempts
to define a single market that includes all types of apps (§ III.B.i, pp. 5–7). Game developers and
consumers substitute between transaction platforms when seeking out game transactions,
supporting a conclusion that the relevant market is the market for game transactions across
digital transaction platforms, rather than an iOS-only market for all app transactions
(§ III.C.i, pp. 9–10). The evidence also refutes Epic’s experts’ assertion that platforms are
complements rather than substitutes (§ III.C.ii, p. 10).
5. Opinion 4. Dr. Evans proposes a foremarket/aftermarket framework to support a single-brand

market definition. But substitution between transaction platforms means that developers and
consumers are able to switch in the alleged aftermarket without having to also switch in the
alleged foremarket, critically distinguishing the present case from historical applications of the
foremarket/aftermarket framework (§ III.C.iii, pp. 10–11).
6. Opinion 5. Even if the foremarket/aftermarket framework were to be applied, anticompetitive

concerns are absent here, because neither consumers nor developers face a lack of competition in
a more appropriately defined foremarket; they do not lack information about aftermarket terms;
and they have not been subjected to unexpected adverse changes in aftermarket terms
(§ III.C.iv, pp. 11–12).
7. Opinion 6. Because game developers have many ways to transact with consumers, the dispute
here is not about access, but rather about the terms of access (§ III.C.v, p. 12).

8. Opinion 7. Dr. Evans’ hypothetical monopolist “tests” are fundamentally flawed and do not
establish relevant antitrust product markets. While they may appear to bring a sense of scientific
rigor to the market definition exercise, none of these “tests” contain valid, formal analyses of
consumers’ or developers’ substitution patterns in the face of a real-world change in price or
quality. The “tests” are without probative value (§ III.D, pp. 13–19).
Geographic market definition
9. Opinion 8. Geographic restrictions and differences in market conditions mean that even
though the product is digital, geographic markets should be defined on a country-by-country
basis. The appropriate geographic market here is the United States (§ IV, p. 20).
Market power
10. Opinion 9. Output, properly measured and benchmarked, is the metric that generally best
captures the relevant net effects when evaluating the existence or extent of market power
(§ V.A, pp. 21–22).
11. Opinion 10. Output on the App Store has grown at a rate substantially faster than the overall
market between 2010 and 2018. This indicates that Apple does not restrict output on the App
Store as a monopolist would, and that it offers a lower quality-adjusted price than many of its
competitors. Moreover, Apple has not increased commission rates since introducing the App
Store in 2008, when Dr. Evans acknowledges it had no market power; Apple does not have a
large share of the properly defined relevant market; and the market has experienced entry and
expansion by multiple competitors (§ V.B, p. 22).
12. Opinion 11. Dr. Evans relies on five “indicators” to argue that Apple has monopoly power,
but they are all flawed and cannot support such a conclusion (§ V.C, pp. 23–24).
13. Opinion 12. In addition to Dr. Evans’ faulty comparison to China, his “analysis” of a but-for
world involving app store entry suffers from fatal methodological and conceptual flaws, as
evidenced by several results that do not comport with intuition, standard economic outcomes, or
reality. That analysis should be disregarded (§ V.D, pp. 24–25).

14. I am the Associate Dean for Business + Impact, and the William Davidson Professor of
Business Economics and Public Policy, at the Ross School of Business at the University of
Michigan. I am also Professor of Economics (courtesy) at the University of Michigan’s
Department of Economics.
15. My academic research is in the field of economics called Industrial Organization. Within
Industrial Organization, my research focuses on several related topics: vertical relationships,
franchising, and entrepreneurship, and associated policy issues.
16. From Fall 2014 to the end of 2015, I served as Director of the FTC’s Bureau of Economics.
In that position, I oversaw the economic analyses for all regulatory and enforcement matters that
bureau staff were involved with, including those involving mergers, monopolization, franchising,
and vertical relationships. These matters included five merger cases that focused heavily on
market definition: three hospital mergers, Sysco/US Foods, and Staples/Office Depot.
17. I have published many papers in peer-reviewed academic journals that address economic
questions related to franchising and vertical contracts; coauthored a book titled The Economics of
Franchising; am currently a co-editor of the Journal of Economic and Management Strategy;
have previously co-edited the Journal of Law, Economics, and Organization; and served in other
editorial roles and as a reviewer at several other peer-reviewed academic journals.

III. The relevant antitrust product market is the market for game transactions on digital
transaction platforms
18. The relevant antitrust product market in this case is the market for game transactions on
digital transaction platforms, including downloads, updates, and in-app purchases. Dr. Evans
disagrees, proposing instead three markets that collectively focus on all apps without regard to
subject matter, but limited to iOS rather than all digital transaction platforms where game
transactions are offered.
A. The App Store is a two-sided transaction platform and so the relevant product is
transactions
19. Professor Schmalensee and I have explained, and Dr. Evans acknowledges, that the App
Store is a two-sided transaction platform. Historically, video games were sold on physical media,
such as discs and cartridges. In the past two decades, video games have been increasingly sold
digitally, meaning as downloads over the internet. The App Store facilitates download, upload,
and in-app purchase transactions between developers and consumers. The product, in other
words, is transactions. Dr. Evans is wrong to focus market definition on “distribution,”
disregarding the transactional nature of the relevant market.
20. Market definition is fundamentally about demand substitution—whether customers are

willing and able to substitute if terms of trade worsen, including if quality decreases or price
increases. Therefore, the “perspective” by which to evaluate any proposed market definition is
the perspective of the customers of the products. Defining markets in the context of two-sided
transaction platforms requires that customers on both sides of the platform be taken into account.
The inquiry must therefore consider the competitive alternatives available to developers and
consumers for transactions—where app content is paid for, not where it is used. Consumers and
developers can substitute to platforms other than the App Store for transactions even if they
continue to play a particular game on the iPhone.
21. Because the relevant product is transactions, much of the evidence that Dr. Evans presents is
irrelevant to market definition, as it focuses on the availability of specific games on different
devices, not on whether developers and consumers can transact across platforms. Even the rare
consumer who has access to only an iOS device has a readily available game transaction
alternative to the App Store—the Safari browser. For example, any Fortnite player can use Safari
(or Chrome) to purchase Fortnite’s in-game currency, “V-Bucks,” a transaction that generates no
commission for Apple.
B. Epic’s market definition improperly clusters game transactions with other types of
app transactions that face different competitive conditions
22. The role of market definition is to provide a conceptual framework to guide and discipline
antitrust analysis. It is not an end in and of itself, but rather one of several tools to evaluate the
theories of harm at issue. Markets that are too broadly defined may fail in that role by overstating
the competitive alternatives available to customers facing the exercise of market power, while
markets that are too narrowly defined may understate such alternatives.

23. Market definition focuses on customer substitution.1 In this case it must therefore focus on
the set of options available to consumers and developers for their transactions. Dr. Evans thus
fundamentally misses the relevant economic issue when he states that Apple’s policies “apply to
all apps.”2 Even setting aside their erroneous focus on game distribution as opposed to game
transactions, Epic’s experts have failed to establish that all iOS apps are reasonable substitutes
for one another.
24. As one example that I explain further below, game transactions—the type of transactions that
Epic primarily engages in—are distinct from other types of digital transactions. Consumers who
play games and are facing higher prices or lower quality are unlikely to substitute to a non-game
app or non-game transactions. Similarly, game developers typically focus on game apps, and are
therefore also unlikely to respond to higher prices or lower quality by shifting efforts to develop
types of apps other than games, or to substitute to non-game transactions. Moreover, game
developers and consumers who play games have a distinct set of competitive alternatives
compared to developers and consumers of other types of apps.
25. To address the specific theory of harm in this case, market definition cannot ignore the
competitive alternatives available to Epic and similarly situated developers (i.e., game
developers), as well as the consumers who transact with them. I show below that this is a distinct
and identifiable product. Indeed, Dr. Evans acknowledges that Fortnite has more “substitution
possibilities” than other types of apps do.3
i. Qualitative and quantitative evidence supports game transactions on digital transaction platforms
as a distinct relevant product market
26. Professor Hitt provides qualitative evidence that game transactions are economically distinct
from transactions for other types of apps. The economic analysis is informed by commercial
realities, including that industry players treat games differently from other types of apps. Game
app developers often use specialized technology to create game apps. Game transactions are
offered via both general storefronts (which present games as a specific category) and specialist
storefronts (that only or predominantly offer games). For example, many of the transaction
platforms’ user interfaces, including those of the App Store, Google Play, and the Amazon
Appstore, reflect the distinction between game transactions and non-game transactions, often
categorizing “games” into a separate tab of apps. This implies that platforms recognize that
consumers sometimes visit the platform looking for games and benefit from having games
collected in one location. This is similar to how Walmart separates wine and tools into separate
sections of its store. Apple’s internal business strategies provide further qualitative evidence that
games are distinct from other app transactions: on the App Store, editors consider special factors
when curating games, and Apple has two separate heads of business development, one for games
and one for all other apps.
1
U.S. Department of Justice and the Federal Trade Commission, “Horizontal Merger
Guidelines,” August 19, 2010, p. 7.
2
Written Direct Testimony of Dr. David S. Evans, (“Evans written testimony”), ¶ 39.
3
Evans written testimony ¶ 125.
27. Professor Hitt also provides quantitative evidence that game developers tend not to diversify
their offerings beyond game apps (consistent with a lack of substitution on the part of game
developers to transactions for other types of apps), and that they interact with various platforms,
including those that offer mostly or only game transactions, to transact with consumers who play
games. This further supports a finding that game transactions are a distinct product market.
ii. Where competitive conditions for products are not similar to competitive conditions for other
products, it is inappropriate from an economics perspective to cluster them into a single antitrust
market
28. One aspect of my assignment was to address questions posed by the Court in its October 9,
2020 preliminary injunction order, including whether “clustering” warrants expanding the
relevant product market beyond games. Dr. Evans stated in his deposition that he is not taking
the position that “iOS app distribution” is a cluster market, yet he also does not offer an opinion
on whether transactions for other types of apps could be substitutes for game transactions. If two
products are not substitutes for one another, they belong to distinct product markets and there is
no economic basis to group them into a single antitrust market, unless that market is a cluster
market. If all of the transactions in Dr. Evans’ market definition can be assembled into a single
market, without invoking a cluster market, then it must be that when game developers or
consumers who play games are faced with higher prices or lower quality they would substitute to
transactions for other types of apps on iOS, instead of substituting to game transactions on other
platforms.
29. Cluster markets refer to combinations of individual product markets where products are
bought and sold independently. Though the products in a cluster market are not reasonably
interchangeable and so are not substitutes, economists and courts sometimes group them together
for analytical convenience. Clustering products that are not substitutes only makes sense as an
economic matter if the products are subject to similar competitive conditions and, effectively, the
answers one would get from analyzing any one underlying product market would be the same as
those obtained in analyzing another underlying product market, or the cluster market itself.
30. For example, consider hospital services. Cardiology and trauma services are not substitutes,
nor are they bought as a bundle. Yet they could be grouped together in a single cluster market for
analytical convenience if the competitive conditions that are faced by patients are similar for
both sets of services. It would not be appropriate to group them together if all hospitals offer
cardiology services, yet only some hospitals offer trauma services. Clustering these together
under such conditions would likely lead to understating competition for cardiology services or
overstating competition for trauma services.
31. The 2016 Staples/Office Depot case, the economics of which I oversaw in my role at the
FTC, helps clarify when clustering is and is not appropriate. In that case, the FTC successfully
argued that it was appropriate to cluster consumable office supplies (e.g., pens, file folders, Post-
it notes, binder clips, paper for copied and printers) in a single antitrust market because these
products faced similar competitive conditions, even though they are not functional substitutes for
each other. Put another way, pens and paper clips are not reasonable substitutes and hence do not
belong in the same product market, but distinguishing between those separate product markets
would have made no difference in understanding competition in the office supplies industry.

However, ink and toner were excluded from the cluster market, because there was evidence of
additional competition in the product market for ink and toner. A cluster market that included ink
and toner thus would lead to either understating competition for ink and toner or overstating
competition for other consumable office supplies.
32. Another example from my time at the FTC relates to pharmaceutical mergers. Consistent
with market definition being about demand substitution, the FTC defines separate markets
around specific molecules in analyzing such mergers, they do not define a single market
encompassing all the molecules that the firms produce. Pharmaceutical companies may well
apply similar policies (sales terms, quality control, etc.) to all of their products, but this is not a
reason to cluster all molecules into a single market. Instead, customers typically face different
competitive alternatives for each molecule, leading to separate relevant antitrust markets. A
merger might then raise concerns in some but not all relevant antitrust markets.
iii. Competitive conditions faced by game developers and consumers who play games differ from
competitive conditions faced by developers and users of other types of apps, and so clustering all
apps together would be inappropriate as an economic matter
33. In the present matter, clustering all app transactions would be appropriate if and only if all
the different types of apps in the App Store—such as gaming apps, dating apps, wellness apps,
etc.—faced similar competitive conditions regarding transactions. Epic’s experts have not
analyzed this question or offered any support for clustering all apps, without regard to subject
matter, in a single market.
34. Professor Hitt has provided extensive evidence that many other types of apps do not face
similar competitive conditions with game apps. (Indeed, other types of apps also face different
competitive conditions from one another.) There exist many transaction platforms that focus on
games (e.g., PlayStation, Xbox, Steam) and that are meaningful substitutes for game developers
and consumers who play games, but not for developers and consumers of other types of apps.
There are also important monetization differences (such as the share of games relative to other
types of apps that use subscriptions) that lead to important differences in market outcomes, such
as average commission rate. Other types of apps could have competitive alternatives that are not
available to game apps. For example, consumers can use an internet-connected smartwatch to
replicate the functionality of calculator or weather apps that they would otherwise use on their
phone. They can also replicate the functionality of music or health & fitness apps through fitness
watches and associated transaction platforms. As a result, it is inappropriate to cluster all apps on
the App Store together, and the appropriate economic analysis in the present case must focus on
game app transactions.
35. In contexts where multiple product markets may be defined, one should choose among them
remembering that relevant antitrust markets must be defined with the specific theory of
anticompetitive harm in mind. Consider again the market definition in Staples/Office Depot.
Instead of a merger challenge by the government, which should consider harm to any consumer,
consider a hypothetical private litigation involving a specific business that purchases only ink
and toner from Staples. In such a case, it is irrelevant that one could appropriately define a
cluster market for consumable office products excluding ink and toner. The case instead should
be analyzed using a product market focused on ink and toner. Similarly, the present matter

47. In other words, the foremarket/aftermarket framework fundamentally depends on lock-in

effects in the foremarket, which are absent here. This is because consumers multi-home across
devices, and so the App Store can lose a consumer’s game transactions even if the consumer
does not replace their iPhone. As I have discussed, even consumers who do not own or have
access to a device other than an iOS device can use the Safari browser to transact through the
developer’s website.
48. Meanwhile, developers are not locked in, either. Nothing in Apple’s license restricts
developers from offering transactions on other platforms, or “steering” consumers to other
platforms by charging less on these platforms or more on the App Store. Dr. Evans describes
only what he terms a “nominal” $99 yearly developer fee as the required foremarket payment to
Apple from those customers.
49. Because neither developers nor consumers are “locked in” to their iOS devices, this case
does not fit with historical applications of the foremarket/aftermarket framework, and so the
framework should not be applied to this case.
iv. Aftermarkets give rise to anticompetitive concerns only under particular circumstances that
substantially differ from the circumstances here
50. The economic literature and antitrust agencies explain that when the framework is relevant,
aftermarket concerns nonetheless only arise when (1) there is no robust competition in the
foremarket, (2) customers cannot gather sufficient information about aftermarket terms and costs
before buying in the foremarket, and (3) locked-in customers are surprised by unexpected
changes in aftermarket terms.5 As I discuss below, none of these conditions are met for
consumers or developers, and so even if one were to define a foremarket and aftermarket, there
would be no anticompetitive harm.
51. Starting with consumers who play games, note that:
• (1) While Dr. Evans defines the foremarket as operating systems, consumers do not
purchase operating systems; they purchase smartphones, which would be the product that
consumers buy in a foremarket if one were to be defined in this case. I discuss below
several reasons why Dr. Evans’ support for a smartphone operating systems market is
insufficient. Apple faces strong competition in the smartphone market, competing with
dozens of smartphones designed and marketed by multiple well-funded smartphone
manufacturers. If the iPhone were not a competitive offering, then its customer base
would quickly evaporate.
• (2) Prior to purchasing a smartphone, consumers have sufficient information available to
them about differences in smartphone options, as well as aftermarket terms including
terms related to apps. This information is available in product reviews and publications,
or would be if considered important and different across platforms.
• (3) Consumers are not locked in. Many consumers upgrade their devices and some switch
to an Android device. If consumers were locked in, one might expect them to indicate
5
United States Federal Trade Commission, “Competition Issues in Aftermarkets,” Organisation
for Economic Co-operation and Development Competition Committee, June 2017.
regret or lack of satisfaction, but data on satisfaction and demand from new smartphone
users show that those who do not switch are happy with their iPhones. Furthermore,
Apple has not made unexpected changes to its terms or prices offered to developers,
maintaining or improving its policies from the outset. Nor has Apple changed its terms or
prices offered to consumers, who are not charged by Apple for their use of the App Store.
52. Turning to developers:
• (1) Competition in Dr. Evans’ foremarket cannot be evaluated because game developers
do not buy operating systems. An operating system without a mechanism that facilitates
app transactions is not a two-sided platform because developers have no way of
interacting with consumers on, and have no use for, a closed operating system. Even if
there were a developer foremarket for smartphone operating systems, Apple’s and
Google’s low prices and frequent updates and innovations suggest no lack of competition
in Dr. Evans’ foremarket.
• (2) Terms for developers have also been clear in their contracts and have not adversely
changed. For example, Apple’s Developer Program License Agreement makes clear that
developers set their own prices but must pay a commission up to 30 percent of the prices
they set.6
• (3) Apple has offered developers the same—and in some cases, better—terms since
introducing the App Store in 2008, a point in time when it held a very small share of
phone or smartphone sales, and, by definition as an entrant, held no share in game
transactions (and when Dr. Evans acknowledges Apple had no monopoly power).
v. Multi-homing and substitution by consumers and developers demonstrates that the dispute here
is about the terms of access, not about access itself
53. Dr. Evans describes the App Store as the only way for a developer to “gain access to 1 billion
iPhone users globally.”7 He extends this argument with an analogy, arguing that a manufacturer
must sell in both California and Oregon if it wishes to reach customers in both states.
54. This argument is incorrect. The multi-homing and substitution evidence discussed above
shows that game developers typically have multiple ways to transact with consumers who play
games and Dr. Evans is wrong to argue that the App Store is the only way to transact with
owners of iOS devices. As noted earlier, even consumers who only have access to an iOS device
can use a browser to directly transact with game developers. But as explained by Professor
Hanssens, most iPhone users have access to other devices. This means that the dispute here is
ultimately about the terms of access, including the commission rate, not about access itself. It
also means that Dr. Evans’ analogy does not work. Few customers (literally) multi-home by
living and shopping in both California and Oregon, whereas most game developers and
consumers have the opportunity to multi-home.
6
DX-3256.
7
Evans written testimony, ¶ 295.
D. Dr. Evans’ hypothetical monopolist tests are fundamentally flawed and do not
establish relevant antitrust product markets
55. The hypothetical monopolist test is the thought experiment by which economists
conceptualize the market definition exercise of whether markets are drawn too narrowly.
56. At the center of Dr. Evans’ market definition analysis are six purported hypothetical
monopolist tests in support of his alleged “smartphone operating systems” and “iOS App
Distribution” markets. (Dr. Evans conducts a seventh hypothetical monopolist test for his alleged
“In-App Purchase Payment Processing” market. Prof. Schmalensee addresses this test.) Dr.
Evans offers three “tests” for each of his two alleged markets—one for the consumer side, one
for the developer side, and finally one that purports to test both sides at once, since he recognizes
them as two-sided transaction platforms (and his own writings indicate the two sides should be
tested together in these types of markets). All six “tests” are deeply problematic and misleading,
and should be disregarded.
57. First, Dr. Evans overemphasizes the importance of these “tests” in his market definition
analysis. Antitrust agencies and economists typically do not define relevant antitrust markets
mechanically under the hypothetical monopolist test. Rather, as Dr. Evans ultimately agreed in
his deposition, antitrust agencies and economists tend to define markets based on both
quantitative and qualitative evidence (e.g., documents and testimony identifying competitors or
describing competitive pressures). This typically leads to markets that constitute intuitive, natural
groupings of products that conform to market realities. Thus, Dr. Evans is wrong to state in his
report that the hypothetical monopolist test “involves finding the narrowest set of products that
would enable a hypothetical monopolist to raise price profitably by a small but significant
amount”8 (emphasis added), an argument he walked back from in his deposition.
58. A mechanical application of the hypothetical monopolist test to find the narrowest market
can lead to outcomes that would make market definition unhelpful to evaluating antitrust claims.
For example, the narrowest product market in which Coca-Cola competes may be a narrow
market that consists solely of Coca-Cola and Pepsi-Cola, the two most popular sodas. Such a
market would be unhelpful in evaluating a proposed merger between Coca Cola and Dr. Pepper,
or hypothetical collusion between PepsiCo and Dr. Pepper. Note that, when the FTC challenged
the proposed merger of Coca Cola and Dr. Pepper in 1986, it defined a market of all carbonated
soft drinks rather than a narrower market of some collection of soft drinks that satisfied the
hypothetical monopolist test.
59. In addition, while Dr. Evans’ “tests” may appear to bring a sense of scientific rigor to the
market definition exercise, they are fundamentally flawed and uninformative in establishing
whether either of his markets are relevant antitrust product markets. At the most basic level, and
addressed more fully below, none of these “tests” contain valid, formal analyses of consumers’
or developers’ substitution patterns in the face of a real-world change in price or quality. Dr.
Evans admitted in his deposition that his “tests” do not rely on evidence of actual switching. Nor
did he consider the Apple transactions data in evaluating substitution. Customers’ substitution
patterns in response to price or quality changes are central to the market definition exercise, and
8
Evans initial report, ¶ 236.
any hypothetical monopolist test that fails to identify these patterns cannot have probative value.
In most cases, as discussed below, Dr. Evans merely postulates a price change and asserts,
without support, that few customers would substitute in response.
60. Note that Dr. Evans does not profess to apply a hypothetical monopolist test for game
transactions, and so he has not shown that a hypothetical monopolist test would invalidate game
transactions across digital transaction platforms as too narrow.
i. Dr. Evans’ consumer-side “test” for his alleged “smartphone operating systems” market
61. Most of the problems with Dr. Evans’ first “test” follow from the fact that consumers do not
buy smartphone operating systems, a fact that Dr. Evans acknowledges. He nonetheless proceeds
to “test” the consumer side of this alleged market, which he claims is a two-sided transaction
platform. In his direct testimony, he argues that defining his foremarket around OSs rather than
devices is appropriate because the OSs “have differentiated accompanying ecosystems,
and…consumers first and foremost choose the ecosystem experience they want.”9 This is not
sufficient grounds for defining his market. First, he offers no analysis of device choice to show
whether consumers do, in fact, “first and foremost choose the ecosystem.” Second, even if the
ecosystem were an important factor in device choice, this does not mean that it is the only factor
and that the market should be defined around ecosystems. Even ignoring both of these issues,
this approach would suggest a market for “ecosystems,” which are related to but not the same as
smartphone operating systems. For example, iOS users can participate in the Google ecosystem
through the (free) Gmail, Google Photos, and other apps and services, without ever purchasing
an Android device.
62. To underscore how artificial his alleged market is, in devising the “test,” Dr. Evans does not
consider consumer substitution to other operating systems (his alleged product), but rather to
“other devices”; indeed, Dr. Evans admits this in his direct testimony, when he says that he looks
at “whether smartphones, personal computers, and game consoles—and their respective OSs—
sufficiently substitute for one another to comprise a single relevant market.”10 Furthermore,
while Dr. Evans presents his SSNIP as an increase to the price of smartphone operating systems,
he is forced to apply the price change to the smartphone overall.
63. The first problem with “testing” a market for a product that consumers do not actually buy is
finding the product’s price. Dr. Evans admits that “Apple and Google do not charge explicit
prices for their OSs.”11 Dr. Evans is forced to resort to critically flawed comparables—
specifically prices for Windows and the (defunct) Windows Phone operating system. For
Windows, Dr. Evans bases his SSNIP on the prices that Microsoft charged manufacturers for its
operating system, even though his “test” is purportedly focused on consumers, and even though
Microsoft sells Windows to consumers at retail, charging $139 for a basic “Home” version, $200
for a “Pro” version, and $309 for a “Pro for Workstations” version. These prices are all
substantially higher than the prices charged to manufacturers that Dr. Evans relies upon. Thus,
9
10
11
even setting aside the other flaws with his “test,” Dr. Evans substantially understates the
consumer price of operating systems and thus also understates the size of the SSNIP.
64. Even if one were to base the SSNIP on manufacturer prices, Dr. Evans uses the price charged
for a failed smartphone operating system (Windows Phone) and an operating system for PCs
(Windows). Neither is a good comparable, and both are from roughly a decade ago (2012 and
2009, respectively).
65. Android would provide a more current comparable, but it is still an operating system that is
offered to smartphone manufacturers, not consumers. Moreover, as Dr. Evans acknowledges,
Android is offered to manufacturers at a price of $0. While this $0 price does not reflect
Android’s true value—iOS and Android clearly both have substantial value—this $0 price would
imply a $0 SSNIP, which highlights that “smartphone operating systems” is not a well-defined
market.
66. Had Dr. Evans focused his foremarket on a product that consumers actually purchase, the
smartphone, the SSNIP would have been based on substantially higher prices. Dr. Evans presents
a $551 average price for U.S. smartphones but also acknowledges that “Apple specializes in
selling premium smartphones that appeal to consumers who are more likely to use and spend on
smartphone apps.”12 For example, the price of Apple’s current generation iPhone 12 ranges from
$699 for the baseline iPhone 12 Mini to $1,399 for an iPhone 12 Pro Max with 512 GB of
memory. A 10% SSNIP on a $1,000 phone would be $100, not the $3–$5 that Dr. Evans uses.
67. Having postulated the price and level of SSNIP, Dr. Evans then simply asserts, with no
formal analysis and no data on the substitution relevant to his purported market, that it would be
“implausible” that a hypothetical monopolist could not profitably raise price by $3 or $5.
ii. Dr. Evans’ developer-side “test” for his alleged “smartphone operating systems” market
68. Dr. Evans makes a similar error in “testing” a market for a product that developers also do
not actually purchase. Dr. Evans’ alleged foremarket (smartphone operating systems) is not
actually a two-sided transaction platform without his alleged aftermarket (i.e., without the App
Store or other means to transact with consumers on iOS). Smartphone operating systems only
have value to developers insofar as they present an opportunity to transact with consumers.
Therefore a hypothetical monopolist test that focuses on developers, yet is divorced from
transactions, is ill-conceived.
69. As with his previous “test,” this causes problems with finding an appropriate price. Dr. Evans
bases his SSNIP on a $99 price, which he acknowledges is “nominal,” remarking that the
resulting SSNIP would be “negligible” and “could not have any plausible impact” on developers
(again without performing any formal analysis based on actual substitution in his alleged market,
because there is no such market). This is inconsistent with Dr. Evans’ characterization of Apple
and Google holding a “duopoly” of smartphone operating systems with few competitive
constraints, and would imply that the “duopolists” are leaving money on the table. Of course, the
price that developers actually care about is the commission associated with “iOS App
12
Distribution.” Yet Dr. Evans ignores commissions within this “test,” because he attributes them
to a separate market—“iOS App Distribution.”
70. Dr. Evans refers to a sensitivity analysis where he does calculate “prices” to include
commissions. But he does so in a nonsensical way, by spreading commissions across all
developers and all smartphones. No developer pays a commission on a “per smartphone per
year” basis or views this “price” as a salient feature of its interaction with the App Store.
71. The above demonstrates another critical flaw in Dr. Evans’ “test”—it treats all developers (in
his alleged “all apps” market) as if they are the same. However, as Professor Hitt explains,
developers pursue many different monetization strategies, and an increasing number of them pay
no commission, monetizing instead through ads, on other platforms, through traditional retail
channels, or offering free apps altogether. Thus, any analysis appropriately reflecting that
commissions are the prices relevant to developers would analyze a substantially larger SSNIP on
a subset of developers that pay substantial commissions. This likely includes Epic and many
other game developers that are participants in the relevant antitrust market for game transactions,
since as Professor Hitt shows, game apps are more likely to monetize through the App Store,
more likely to use in-app purchases, and face high average commissions on the in-app purchases
they offer.
iii. Dr. Evans’ consumer-side “test” for his alleged “iOS App Distribution” market, which relies on
inputs from Professor Rossi’s flawed survey
72. Conceptually, Dr. Evans’ “test” highlights the reasons economists generally do not recognize
single-brand markets. To demonstrate that a SSNIP would be profitable in a single-brand
candidate market is necessarily to argue that a firm is not currently maximizing its profits, an
argument that deviates from a key assumption underlying microeconomics in general and
industrial organization economics in particular. Dr. Evans reaches precisely this erroneous
conclusion in the present case, finding that Apple has “left some money on the table” to the tune
of nearly a billion dollars in 2019 alone. Dr. Evans provides no explanation as to why Apple has
not raised its commission given his prediction that doing so would be profitable. What is more
likely is that Dr. Evans’ conclusion is incorrect—that Apple in fact is constrained by existing
competition and potential entry in transactions so it cannot profitably raise its commission rate.
73. Additionally, this “test” again treats all developers as if they are the same. In this case, he
assumes a specific behavior by all developers, namely that they will “pass through” 50% of any
increase in commissions to final consumers in the form of higher app prices. One should not
expect that developers across all types of apps and in-app purchases available on the App Store
would pass through increased commissions at the same rate.
74. Furthermore, this “test” relies extensively upon fundamentally flawed survey data inputs
from Professor Rossi. In particular, Dr. Evans relies on Professor Rossi for purported substitution
data, namely on the extent to which consumers will supposedly reduce their purchases on the
App Store or switch away from their iPhones entirely. This survey data is the only data on
substitution in response to a price change that Dr. Evans uses across any of his “tests,” but it is
entirely unreliable for the purpose to which Dr. Evans puts it. I leave the critique of the design
and implementation of the survey to Professor Hanssens. Here I discuss, from an economic

perspective, three main shortcomings of the survey data as the basis for Dr. Evans’ hypothetical
monopolist “test.”
75. First, Professor Rossi’s survey (and therefore Dr. Evans’ entire “test”) focused on the wrong
product. As Dr. Evans confirmed in his direct testimony, this “test” is purportedly about “iOS
App Distribution,” which Dr. Evans defined in his deposition as only downloads (installs and
updates). Yet Professor Rossi’s survey is entirely about in-app purchases, as Dr. Evans
acknowledged in his direct testimony. In other words, the survey cannot speak to substitution
relevant to Dr. Evans’ alleged “iOS App Distribution” market because it ignores app downloads
and updates.
76. Second, Professor Rossi’s survey focused on the wrong price change. A hypothetical
monopolist test should consider a non-transitory price change, but in Professor Rossi’s survey,
respondents were explicitly asked about prices in the last 30 days. This is a critical error because
consumers’ responses to longer-run price changes can be substantially different from their
responses to shorter-run price changes. This makes the survey responses unusable for any
hypothetical monopolist test.
77. Third, Professor Rossi’s survey failed to present to participants important choices that would
be available to consumers. For example, Professor Rossi did not account for the fact that,
according to Dr. Evans, consumers consider buying a new smartphone only every two to three
years.13 Thus, even if most participants in Professor Rossi’s survey would not have considered
switching platforms during the particular month of the survey, they might have considered this
option later, when they were due to upgrade their phones. This omission leads the survey to
underestimate the degree of switching in the long run, making it more likely that the alleged
market “passes” the test when it should not. (In early participant interviews, before he ran the
actual survey, Professor Rossi asked participants to assume that it was time to purchase a new
phone.) As Professor Hanssens explains further, Professor Rossi also did not clearly and
transparently offer consumers what might be the most attractive response to the price increase—
consuming the same apps and premium in-app content on iOS, but paying for them on another
platform.
78. Fourth, Professor Rossi’s survey focuses on respondents in the United States only, and he
testified in his deposition that he would hesitate to extrapolate to other countries, but Dr. Evans
purports to conduct a hypothetical monopolist test for a global market.
79. Finally, setting aside the bias Professor Rossi has likely introduced, it is useful to put Dr.
Evans’ estimate of device switchers into context. Dr. Evans claims that “98.6% of spending-
weighted respondents would not respond by switching to an Android device.”14 In context,
however, Professor Rossi’s estimates imply that a five percent increase in in-app purchase prices
on the App Store would cause the number of iPhone owners purchasing new Android phones to
more than triple over the usual rate. In other words, even with all its errors, Professor Rossi’s
survey finds a sizeable effect on consumers’ willingness to switch away from iPhones in the
13
Evans initial report, ¶ 368.
14
event of an in-app purchase price increase, which is inconsistent with the notion that Apple can
profitably raise its commission rates.
iv. Dr. Evans’ developer-side “test” for his alleged “iOS App Distribution” market
80. This “test” also highlights the problem with purported single-brand markets, with Dr. Evans
again concluding that Apple is an actual monopolist but that it has failed to choose the profit-
maximizing price. Again, it is far more likely that Apple is constrained by existing competition
or potential entry in the market for transactions than that it has foregone a profitable price
increase for no particular reason. Dr. Evans does not offer any evidence to the contrary.
81. Also as before, this “test” considers the wrong product, with Dr. Evans again focusing
entirely on in-app purchases rather than initial downloads (presumably because Fortnite, the case
study on which the “test” is based, is free to download and so pays nothing for “iOS App
Distribution”). However, despite the focus on in-app purchases, this “test” purportedly is for his
alleged “iOS App Distribution” market rather than his “In-App Purchase Payment Processing”
market.
82. Much as Professor Rossi ignored important options available to consumers, Dr. Evans
considers only one potential developer response to a commission increase—exiting iOS
altogether. But developers could avail themselves of several other potential and less disruptive
responses should the App Store increase commission rates. These include passing through the
App Store commissions to consumers who buy through iOS (e.g. increasing the price they pay
by the level of the commission), thus also steering consumers to websites and other platforms
(i.e., more strategically utilizing the multiplatform approach); changing the monetization
strategy, such as moving toward advertisements or subscriptions; or moving to a web app. Dr.
Evans has not conducted any analysis of these possibilities, which would likely make the SSNIP
unprofitable (and would explain why Apple is not charging higher commission rates).
83. Finally, the only substitution data Dr. Evans uses in this “test” is, like the participants’
responses to Professor Rossi’s survey, in response to an event decidedly different from a SSNIP.
In this case, the data come from Epic’s “hotfix.” It is therefore not substitution in response to a
quality or price change, and so is not analogous to a SSNIP.
v. Dr. Evans’ two-sided “tests” for his alleged “smartphone operating systems” and “iOS App
Distribution” markets
84. In addition to these one-sided “tests,” Dr. Evans offers two two-sided “tests,” one for his
alleged foremarket for “smartphone operating systems” and another for his alleged aftermarket
for “iOS App Distribution.” These “tests” are purportedly meant to account for indirect network
effects, i.e., for the two-sided nature of each alleged market.
85. In reality, neither properly does so. In both “tests,” Dr. Evans essentially argues (using no
formal analysis) that, because the separate effects of price changes on each side are small based
on his “tests” addressed above (subject to the numerous errors already discussed), there can be
no meaningful indirect network effects.

IV. The relevant antitrust geographic market is the United States
90. The general principles underlying the definition of relevant antitrust product and geographic
markets are the same. The geographic market inquiry thus centers around the willingness and
ability of customers on both sides of a two-sided platform—in the present context, game
developers and consumers who play games—to substitute between products that are offered in
different geographies.
91. Game developers generally face few geographic constraints, because they can transact with
consumers in another country by publishing on platforms within that country. No matter their
country of origin, however, developers must compete for U.S. consumers on platforms’ U.S.
storefronts. Geographic constraints are more important for consumers. Due to regulatory and
other restrictions (censorship, ratings, taxes, etc.), game transaction platforms typically operate in
country-specific storefronts that offer different apps and features, and consumers have limited
capabilities to switch to a storefront other than the storefront of their home country. On the App
Store, for example, consumers are allowed to change country or region through the software on
their phones, but the process involves a number of steps and conditions that most consumers
would find too inconvenient unless strictly necessary (such as spending all App Store credit,
cancelling subscriptions, and get a new payment method and address for the new country or
region). While certain consumers do get around such restrictions, this requires that they
technically violate terms of service, provide incorrect registration information, or register
without a credit card and forgo the ability to download any paid apps or engage in in-app
purchases. The typical consumer, therefore, is generally restricted to purchases from platforms
that operate in their own country.
92. Furthermore, I understand from my time at the FTC that U.S. antitrust laws are concerned
with U.S. consumer welfare. Ultimately, this leads me to conclude that the relevant geographic
market is the U.S.
93. Finally, Dr. Evans’ “global except for China” geographic market is an inappropriate
application of cluster markets because competitive conditions differ across countries (e.g., Dr.
Evans shows differences in income and market outcomes) and so it would be inappropriate to
cluster all countries into a single market as Dr. Evans does.

V. The evidence is inconsistent with Apple possessing monopoly power or charging

supracompetitive prices
A. Economic principles relevant to market power analysis imply that output best
captures the relevant net effects
94. The economic literature defines “market power” as “the ability of firms to raise price above
the competitive level for a sustained period,” i.e., the ability to profitably charge
supracompetitive prices.16 Market power may affect terms of trade besides price, such as quality,
and so is best thought of as reflecting the ability of firms to charge supracompetitive quality-
adjusted prices. The term “monopoly power” can be used to indicate a high degree of market
power.
95. Economists generally evaluate the presence or absence of market power through evidence on
“market structure” (including market shares) and “market outcomes” (e.g., price, output and
quality). Evidence on market structure can help establish that an unconcentrated market with
many small competitors is likely to be competitive, but the inverse is not always true—a
concentrated market does not always imply the existence of market power. The economics
literature considers market outcomes to generally be better suited than market structure to
evaluating claims regarding market power.17
96. As price increases and output restrictions are two sides of the same coin, market power can
also be thought of as the ability to profitably restrict output. Output, properly measured and
benchmarked, can be a particularly useful market outcome to analyze in evaluating the existence
or extent of market power. Dr. Evans agrees that output is a “headline” metric.
97. Output is an especially important metric for evaluating the presence or absence of market
power because it can be more easily interpreted relative to other measures. For example, if the
net effect of any price and quality changes is that the product is better for consumers, they will
purchase more of it and output will rise, indicating a net benefit to consumers overall.
98. Output measurement is not simple, but other market outcomes often are even more difficult
to measure appropriately. For example, most measures of profits are accounting measures, which
are not the same as economic profits. Price may also be measured in different ways, and
determining which measure of price is most informative or whether multiple measures must be
jointly considered can be critical to drawing appropriate conclusions. Further, consumers care
about quality-adjusted price, not just about absolute price. Measurement issues aside, in any
case, profitability and price are only two considerations among many and often must be
considered alongside other factors, such as output. If a firm increases the quality of its product,
for example, a corresponding increase in price is not necessarily anticompetitive.
99. Finally, as the App Store is a two-sided transaction platform, Professor Schmalensee explains
that market power analysis must account for the fact that, even if market power existed on one
16
Jonathan B. Baker and Timothy F. Bresnahan, “Economic Evidence in Antitrust: Defining
Markets and Measuring Market Power,” in Handbook of Antitrust Economics, ed. Paolo
Buccirossi, (Cambridge, MA:MIT Press, 2008) (“Baker and Bresnahan”), p. 15.
17
Baker and Bresnahan, p. 15.
side of a platform (here, either consumers or developers), the platform nonetheless may be
constrained from setting a supracompetitive price on that side due to the effects that feedback
loops might have on both sides of the platform.
B. Evidence on output and other outcomes is inconsistent with Apple possessing

monopoly power or charging supracompetitive prices
100. Output being the most informative measure in evaluating market power, it is critical to note
that all output measures relevant to Epic’s claims have increased substantially over time.
Professor Hitt explains that App Store output measured by third-party game developer revenue
increased by 2,600% from 2010 to 2018. As benchmarks, consider that game developer revenue
across the relevant market increased by only 448% during the same time period, and game
developer revenue on Google Play increased from 2013 to 2018. That the App Store
growth rate is much higher than these benchmarks is evidence not only that Apple does not
restrict output on the App Store, but rather that it offers a lower quality-adjusted price than many
of its competitors and so has contributed to the expansion of market output to a greater degree
than many of its competitors.
101. Before moving to other evidence, I note that while Dr. Evans argues that Apple’s
smartphones are more appealing to consumers who are more likely to use and spend on
smartphone apps,18 and that the digital economy is “vast, growing, and heavily reliant on
smartphones and apps,”19 he ignores Apple’s role in driving this growth and the statistics I just
discussed. He instead asserts, with no analysis, that Apple has diminished output in “iOS app
distribution,” even though he acknowledges that he has not studied output growth nor whether it
declined when, according to him, Apple obtained monopoly power.
102. Focusing on commission rates, as Dr. Evans has done, the App Store’s commission rate for
game transactions is competitive with pricing on other game transaction platforms. Professor Hitt
explains that almost all digital game transaction platforms charge commissions of 30% or more.
Apple has never increased its commission since it opened the App Store in 2008; instead,
Apple’s effective commission has fallen as a result of a sizeable increase in zero-commission
transactions and other commission rate decreases over time.
103. Evidence on market structure also clearly indicates that Apple lacks monopoly power.
Professor Hitt explains that the market includes multiple sizeable competitors and that Apple’s
share of the relevant market of game transactions is at most 37.5% and as low as 23.3% when
accounting for competition from one-sided alternatives, as Dr. Evans suggests one should.
104. Evidence on entry is also inconsistent with Apple having monopoly power in the relevant
market. Professor Hitt documents the entry and expansion of numerous competitors, including
participants in adjacent markets that impose competitive constraints, such as game streaming
services.
18
19
C. Dr. Evans’ five “indicators” fail to establish that Apple has monopoly power or is
charging supracompetitive prices on the App Store
105. Dr. Evans relies on five indicators related to market outcomes and market structure to
purportedly demonstrate Apple’s monopoly power. He claims that Apple is a monopolist with a
virtually 100% share, high profit margins, high commission rates and low investments in quality,
that its dealings with developers are evidence of market power, and that there are barriers to
entry. I discuss these indicators below, none of which establishes Apple’s monopoly power.
106. First, Dr. Evans’ allegation that Apple has a market share of virtually 100% rests entirely on
his incorrect market definition. As Professor Hitt explains, and as I explained above, Apple has a
much lower share of a properly defined relevant market of digital game transactions. Moreover,
the fact that game developers and consumers multi-home creates opportunities for substitution,
especially substitution of transactions, which constrain Apple’s market power on iOS. In other
words, if consumers were given incentives to substitute (e.g., via developers setting higher prices
for iOS transactions), they could choose to make purchases on any platform they have access to
even if they mostly play on another platform. Even consumers who do not have access to a non-
iOS device have the option of entering into transactions that are not subject to Apple’s
commissions through the Safari browser (e.g., to buy V-Bucks).
107. Second, Dr. Evans is also wrong to argue that Apple earns an unusually high profit margin
on the App Store as further evidence for monopoly power. There are both methodological and
conceptual problems with this claim. Economists recognize that in differentiated products
industries such as the classic example of entertainment products, or the industry at issue here,
price premiums well above marginal costs are to be expected and are not, on their own, evidence
of supracompetitive pricing. This is even more so for firms in highly dynamic industries, where
sustained investments are needed to remain competitive. Economists also generally do not
consider accounting profits to be an accurate measure of economic profits, especially for firms
like Apple that invest substantially in intellectual property. Note that Epic is a good example of
another differentiated product firm that enjoys high accounting margins. Moreover, as Professor
Schmalensee explains, Dr. Evans relies on App Store margins information that is inappropriate.
Apple does not use (or generate) information on fully-burdened margins for the App Store, or
other parts of its ecosystem, in the ordinary course of business. Apple does not view these as
discrete lines of business, but instead as interconnected parts of its ecosystem, all of which work
together to provide a high-quality consumer experience.
108. Third, contrary to Dr. Evans’ claims, Apple does not charge a supracompetitive commission
rate on the App Store. As I have explained, numerous comparable game transaction platforms
charge commission rates that are identical to those charged on the App Store, including several
platforms that, like Apple, operate “walled gardens” that do not permit third party stores or direct
distribution by developers to consumers. His claim is also inconsistent with his market definition
claim that Apple is charging less than the profit-maximizing commission rate—that, as an
alleged monopolist, it could have increased its profits by nearly a billion dollars in FY 2019.
Finally, Apple set its commission rate of 30% at a time when it had a very small user base—and
a time when even Dr. Evans agrees Apple had no monopoly power. As Professor Hitt explains,
the average commission rate charged by the App Store also has decreased over time, and is now

less than two cents per download. Including app updates, all of which are free, the average
commission across initial downloads and updates is substantially lower—virtually zero.
109. Fourth, Dr. Evans is wrong to claim that Apple’s interactions with developers are evidence
of monopoly power. Dr. Evans claims that large developers not being able to negotiate better
terms on the App Store is evidence of Apple’s monopoly power. Like the previous one, this
argument ignores that, as acknowledged by Dr. Evans, Apple set its commission rate of 30% and
its policy against sideloading in 2008—a time when it had a very small user base and hence no
monopoly power. It also similarly ignores that Apple has reduced its commission rate over time.
For example, the introduction of the “reader” rule for certain types of apps (2011), the 15 percent
commission on subscription renewals after the first year (2016), and the 15 percent commission
charged to developers that gross less than one million dollars (2021). Most importantly, treating
large and small customers differently in certain ways and not in others is consistent with Apple
acting to maximize the value of the App Store platform. Providing better terms to large
developers might hurt Apple’s goal of having variety in its app offerings for consumers. This
variety enhances indirect network effects and hence the quality of transactions to consumers and,
indirectly, to developers.
110. Finally, what Dr. Evans claims are barriers to entry to the App Store are, instead, part of
Apple’s business model, which is not anticompetitive. Dr. Evans claims that Apple’s alleged
monopoly power is protected by the terms that prevent certain app transactions on iOS other than
through the App Store. These terms, however, do not constitute anticompetitive conduct—they
are part of Apple’s “walled garden” business model, which is shared by several other platforms
and is not by itself evidence of anticompetitive behavior. Moreover, as already mentioned, Apple
set these terms in place at the time the App Store was created, when Dr. Evans agrees it had no
monopoly power. Dr. Evans’ argument regarding market structure and barriers to entry in his
“iOS App Distribution” market amounts to saying that “walled garden” business models always
entail monopoly power. Specifically, Dr. Evans’ logic implies that all consoles (Sony
PlayStation, Microsoft Xbox, and Nintendo Switch) are monopolists with 100% share of a
relevant antitrust market for distribution of apps on their platforms, with barriers to entry created
via contract. Lastly, his classification ignores the value that walled-garden platforms provide to
developers (including platform investment in tools and other innovations) and consumers
(including privacy, security, compatibility, and quality benefits). Platforms invest in generating
such value in order to compete with other platforms in both hardware sales and game
transactions.
D. Dr. Evans fails to establish that market outcomes would be improved in his but-for
world
111. Dr. Evans offers two approaches to describe the but-for world (the world without the
challenged conduct), both of which fail to establish that market outcomes would be improved.
Professor Hitt addresses the first approach, which relies on China as a but-for world. Professor
Hitt explains both why China is a flawed competitive benchmark and, most importantly, that it
actually suggests a but-for world that would be worse for game developers and consumers who
play games.

112. Dr. Evans’ second approach is an “analysis” suggesting that, absent Apple’s alleged
conduct, other app stores could enter and compete profitably for the distribution of apps on iOS,
which he claims would lead to lower prices and higher quality. This “analysis” suffers from fatal
methodological and conceptual flaws, and should be disregarded.
113. What Dr. Evans has produced in reality is an algebraic exercise, not an economic analysis.
Instead of modeling the economic behavior of firms, Dr. Evans looks for the commission rate
that makes a certain equation true. However, that equation is derived not from standard economic
principles, industry data, or optimal firm behavior, but rather from a long list of arbitrary
assumptions that Dr. Evans has not properly supported and that are not based on economic logic.
114. Given Dr. Evans’ flawed and non-standard approach, it is not surprising that it leads to
results that do not comport with intuition, standard economic results, or reality. For example, Dr.
Evans’ “analysis” predicts that Apple and the entrants would make higher profit margins by
increasing their commission rates above his “prediction.” Holding everything else about Dr.
Evans’ analysis the same, they would make margins of 88.3% instead of 45.8% if they raised
commission rates to 90%. As another example, if he were to assume that there were five entrants
instead of two, his exercise predicts that entry would cause commission rates to increase rather
than decrease, from 26.4% before entry to 28.9% after entry. If he were to assume ten entrants,
he would predict that commission rates would increase to 66.3%. These results go against his
claims—that entry would lower prices and benefit consumers—and underscore how his
assumptions and calculations produce nonsensical results.
115. The reason for these nonsensical results is that Dr. Evans’ algebraic exercise depends on
many assumptions, including one about fixed costs. Intuitively, Dr. Evans’ equation yields
higher commissions when fixed costs for the entrants are higher, or when they are spread across
fewer customers due to a larger Apple share or a greater number of entrants. This relationship
implied by Dr. Evans’ approach contradicts standard economic principles about the relationship
between prices and costs: price should depend only on marginal costs (often proxied by variable
costs); fixed costs should have no direct effect on price (although they do in Dr. Evans’
approach). Instead, fixed costs in economic models affect whether, or how many, firms choose to
enter the market (although they do not in Dr. Evans’ approach, because the number of entrants
and the shares they each garner are assumed, not generated by a model based on optimal firm
behavior as they should be).
E. Conclusion on market power
116. The evidence demonstrates that Apple is constrained from charging supracompetitive
quality-adjusted prices in the App Store. Apple’s alleged conduct in the App Store therefore
cannot have harmed competition in the relevant market for game transactions, nor generated
antitrust injury to Epic. Apple introduced the App Store as a two-sided transaction platform in
2008, with a business model that has met with much success. The “walled garden” was an
intentional, publicized feature of the iPhone and its ecosystem from the start, promising safety to
users. The commission rate that Apple charges was introduced at a time when Apple had a much
smaller market share, a time when Dr. Evans acknowledges Apple had no monopoly power. The
commission rate has remained the same or fallen since then, and is still competitive relative to
other platforms. All the while, Apple has continued to add features, increase the quality of its

products and platform services, and it has grown its output at a pace much higher than the
broader market. Collectively, these facts lead me to conclude that while Apple has found much
success with the App Store, this outcome is the fruit of competition on the merits.
VI. Oath
I declare under penalty of perjury under the laws of the United States that the foregoing
is true and correct.
April 23, 2021
Word count: 11718

OAKLAND DIVISION
EPIC GAMES, INC.,

No. 4:20-CV-05640-YGR-TSH
RICHARD SCHMALENSEE, PH.D.
vs.
Time: 8:00 a.m.
APPLE INC., Courtroom: 1, 4th Floor
Judge: Hon. Yvonne Gonzalez Rogers
Defendant, Counterclaimant.
Ex. Expert

05640-YGR-TSH Richard Schmalensee, Ph.D.
I. Summary of Opinions
1. My testimony focuses on the economics of two-sided platforms and their importance to
questions of antitrust market definition, market power, and competitive effects raised by this
case. This testimony is informed by my decades of scholarship on platform businesses—
much of it with Epic’s expert, Professor David Evans. I also conducted an economic
analysis of Epic’s tying allegations. Based on my expertise and available evidence, I have
reached the following opinions:
The Economics of Two-Sided Transaction Platforms
2. Opinion 1. Two-sided platforms serve two different groups who need each other in some
way, and their core business is to facilitate interactions between members of those two
groups. Platforms generally exhibit indirect network effects: the value of a platform to one
group of users increases as the number of users in the other group increases. To ensure
balanced participation, two-sided platforms must carefully choose the prices that they
charge each side and must take account of feedback loops that can flow from price increases
on either side. (§ III.A, pp. 5-6).
3. Opinion 2. Two-sided transaction platforms facilitate simultaneous, observable transactions
between members of the two groups or “sides.” Transaction platforms exhibit strong
bilateral indirect network effects, which make them more complex than ordinary, one-sided
businesses. Strong network effects create feedback loops that amplify demand reductions
when prices are raised. (§ III.B, pp. 6-8).
4. Opinion 3. The App Store is a two-sided transaction platform. In economic terms,
transactions platforms such as the App Store should be analyzed as supplying only one
product—transactions—and operating in a single market where both sides must be assessed
together. In this case the relevant product market in which the App Store operates is digital
game transactions. (§ IV, pp. 8-13).
5. Opinion 4. Two-sided platforms often set rules that may directly inconvenience one of the
groups they serve but ultimately benefit that group via indirect network effects. For
instance, the App Store’s app review process may inconvenience some developers, but it
makes the Store more attractive to consumers, which ultimately benefits developers. (§
IV.B, pp. 11-12).
Dr. Evans’ Treatment of Platforms

6. Opinion 5. Dr. Evans’ identification of a distinct mobile operating system foremarket and
an iOS app distribution aftermarket—both two-sided markets with developers on one side
and consumers on the other—is artificial and misconceived. It confusingly bifurcates a
single transaction platform and ignores transaction platforms on non-smartphone devices
that are accessible by consumers and developers. As for smartphones, Dr. Evans’
identification rests on unsupported assertions regarding device competition and switching
costs, and it ignores indirect network effects and the transactional nature of the product at
issue. (§ V.A, pp. 13-15).

7. Opinion 6. Dr. Evans’ discussion of differences among devices is unrelated to the issues in
this case, where the relevant product is not “distribution” or “apps” but rather digital game
transactions between groups of developers and consumers. (§ V.B, pp. 15-16).
8. Opinion 7. Dr. Evans’ analysis ignores much of his (and my) own work on platforms,
which implies that often unpredictable effects can flow from alterations to platform business
models. Dr. Evans takes no real heed of the risk that his analysis would not just destabilize
Apple’s iOS business model but would condemn all “walled garden” business models,
particularly those of the game console manufacturers. (§ V.C, pp. 16-17).
Market Definition
9. Opinion 8. Dr. Evans’ prior research makes clear that any market definition exercise
involving platform businesses must take explicit account of the strength of indirect network
effects and associated feedback loops to avoid defining markets that are too narrow. The
foremarket and aftermarket SSNIP tests Dr. Evans performed for this case are fatally flawed
because they do not do this. (§ VI.A.1-2, pp. 17-21).
10. Opinion 9. Dr. Evans’ foremarket SSNIP test is unrelated to any issues in this case, and it is
unreliable because one of the products in his imaginary foremarket (iOS) has never been
sold separately while the other (Android) has always been given away for free. Dr. Evans’
aftermarket test asks whether the App Store is maximizing profit (and answers that question
in an unreliable way). This question is not relevant to any issue in this case. (§ VI.A.3, pp.
21-24).
11. Opinion 10. Game transactions face different competitive conditions compared to
transactions for other apps or app categories (which involve other groups of developers and
consumers). The relevant market in this case for evaluating Apple’s restrictions on app
distribution is digital game transactions. (§ VI.A.4, pp 24-25)
12. Opinion 11. Even though it is not possible on this record to conduct a correct SSNIP test,
available evidence makes it clear that the relevant product market for digital game
transactions is broader than transactions on the App Store alone. (§ VI.A.4, pp. 24-25)
Market Power
13. Opinion 12. Dr. Evans’ assessment of the App Store’s market power rests on an estimate of
its operating margin, but operating margins do not measure profitability. Moreover, the iOS
ecosystem abounds with what economists call “joint costs” which cannot be allocated in an
economically meaningful way among the multiple products that those costs support. Mr.
Barnes’ estimate of the App Store’s operating margin is unreliable because it looks in
isolation at one segment of the iOS ecosystem in a way that artificially boosts the apparent
operating margin of that segment. When one looks at Apple’s device and services
ecosystem as a whole, the operating margin falls to an unremarkable level. (§ VI.B.1-2, pp.
25-28).
14. Opinion 13. The fact that Apple’s iOS business has a relatively complicated monetization
strategy and that its strategy differs from those of other operating system vendors does not
imply that it has market power. (§ VI.B.2, pp. 27-28).

15. Opinion 14. Properly calculated, Apple’s market share is certainly below 40 percent and
arguably below 30 percent. Moreover, online platforms with strong indirect network effects
and low barriers to entry can experience rapid changes in market share, so a high market
share would not imply durable market power. (§ VI.B.3, pp. 28-29).
Competitive Effects
16. Opinion 15. Under Apple’s policies at issue, which were established before Dr. Evans
claims the App Store attained monopoly power, the number of iOS app developers and the
volume of iOS app transactions have grown enormously. This is in conflict with Dr. Evans’
past emphasis on the significance of observed output growth in competitive effects analysis.
Dr. Evans has provided no economic evidence of anticompetitive effects and has ignored
voluminous evidence of a healthy, thriving iOS ecosystem. (§ VI.C, p. 30).
Epic’s Tying Claim

17. Opinion 16. IAP is not a “separate product,” but rather an integral part of the App Store
transaction platform that provides benefits to users and developers. IAP is a critical input
into the production of App Store transactions. An in-app purchase of digital content,
functions, or services using IAP constitutes, in effect, an upgrade of a previously
downloaded app. The fact that Apple chooses to monetize some transactions and not others
does not mean that IAP is a separate product. (§ VII.A, pp. 31-35).
18. Opinion 17. The IAP requirement for digital content or services enables efficient collection
of Apple’s commissions. All major online stores that charge sellers commissions also
require the use of their own payment mechanisms to enable efficient collection of those
commissions. The anti-circumvention rules in the App Store and other online stores prevent
free-riding, like the anti-steering rules in the Amex case. (§ VII.A.1, pp. 32-35).
19. Opinion 18. As in other online stores, the App Store’s commissions are not fees for
payment processing. Thus Dr. Evans’ SSNIP analysis comparing the App Store’s
commissions to competitive payment processing rates is fatally flawed. (§ VII.C, pp. 38-
40).
20. Opinion 19. There is no tie between IAP and iOS app distribution because they are not
separate products. The fact that Epic and other developers have tried to circumvent IAP is
not valid evidence of a separate demand for payment processing; it merely shows they want
to avoid Apple’s commission. Even if IAP and app distribution were separate products,
they have not been tied, as developers that use the App Store can—and do—choose forms
of monetization that do not require use of IAP (§ VII.B, VII.D, pp. 36-38, 40).

21. I am the Howard W. Johnson Professor of Management Emeritus and Professor of

Economics Emeritus at the Massachusetts Institute of Technology (MIT), where I have
taught industrial organization and related subjects since 1977. I served as the Dean of the
MIT Sloan School of Management from 1998 through 2007.
22. I was a Member of the President’s Council of Economic Advisers from 1989 through 1991.
I am a Fellow of the Econometric Society and the American Academy of Arts and Sciences
and was the 2012 Distinguished Fellow of the Industrial Organization Society. I have
served as an elected member of the Executive Committee of the American Economic
Association and am currently a member of the Executive Committee of the Board of
Directors of the National Bureau of Economic Research. I have S.B. and Ph.D. degrees in
economics from MIT.
23. I am the author or co-author of 13 books, more than 110 published articles, more than 35
book chapters, and many shorter papers. Since 1977 I have testified in a number of antitrust
cases and related matters. I have also testified before the U.S. Congress and state and
federal regulatory agencies. In the last two decades, much of my work, including my three
most recent books, has focused on platform-based businesses, which are defined below.
Much of this recent work, including the Invisible Engines and Matchmakers books and an
amicus brief to the Supreme Court in the Ohio v. American Express (Amex) case, has been
joint with Dr. Evans.
III. The Economics of Two-Sided Platforms
A. General Properties
24. Owners of iOS devices benefit from being able to acquire high-quality apps. Developers of
iOS apps benefit from being able to reach a large number of consumers. Apple’s App Store
exists to facilitate transactions between members of these two groups.
25. Two-sided platforms serve distinct groups of customers who need each other in some way,
and their core business is to provide a common (real or virtual) meeting place that facilitates
interactions between members of the two distinct customer groups. For a platform to be
attractive to both groups it seeks to serve, it must reduce frictions that impede their desired
interaction and structure the prices it charges to both sides so as to make the platform
attractive to both.
26. A key feature of two-sided platforms is that they often exhibit what economists call
“indirect network effects.” Indirect network effects refer to the situation in which the value
realized by members of one group of customers of a platform is higher when they have
access to more members of the other group of customers with whom they could
productively interact. Successful two-sided platforms must ensure that there are a large
number of participants on both sides of the platform and that transactions on the platform
are as easy, safe, and reliable as possible.
27. Because of indirect network effects, the multi-sided platform business model is much more
complex than the business model of a brick-and-mortar retailer or other single-sided

businesses. Platforms must select pricing strategies, service provision, and rules of behavior
so as to ensure balanced participation of the customer groups they serve. If a platform is
required to alter one element of its overall strategy—pricing to one side or rules of behavior
affecting another side, for instance—it may find it desirable or even necessary to respond by
altering other elements of its strategy, with diverse (perhaps unpredictable) effects on
groups on both sides of its platform. Hence, even minor changes to a platform’s pricing
strategy always require careful considerations from various viewpoints.
28. To ensure participation on both sides, so that each side finds participation on the platform
attractive because of participation by the other side, platforms must carefully choose their
price structures on both sides. Platforms’ price structures depend in part on the price
elasticities of demand of the groups served. Like ordinary one-sided businesses, platforms
will tend to charge higher prices, all else equal, to customers who are less sensitive to price,
i.e., that have lower price elasticities of demand. In fact, it is common for one side of a two-
sided platform to be charged less than marginal cost to encourage participation on that side,
in order to make it attractive for the other side, which is charged above marginal cost, to
also participate.
29. Indirect network effects that affect both groups served create feedback effects that amplify
reactions to price changes. Suppose a platform serves groups A and B and it increases price
to group A. This will reduce participation by members of group A. That, in turn, will make
the platform less attractive to members of group B and will accordingly reduce their
participation. And the process continues: less participation from B will make the platform
even less attractive to members of group A, a reduction in A’s participation will make the
platform even less attractive to members of B, and so on. It is possible that the result will be
that both groups completely drop off the platform. For modest price increases, the process
will generally converge, and the decline in participation by A will be greater than A’s price
sensitivity would imply: demand is, in effect, more elastic. In addition, participation by
group B will have fallen.
30. To prevent platform participants in one group from reducing the value of the platform to
members of the other group, platforms often need to establish and enforce rules of behavior.
Such rules are often necessary because the interests and incentives of both sides don’t
always align. For instance, OpenTable suspends a user’s account if the user is a no-show
for four reservations within a 12-month period. That may be a mild inconvenience for some
consumers, but it makes the platform more valuable to restaurants, and if more restaurants
participate, the platform is more valuable to all consumers. Both eBay and Amazon
Marketplace similarly have detailed rules with which sellers must comply in order to retain
access to these platforms.
B. Two-Sided Transaction Platforms
31. Economists recognize a particular type of multi-sided platforms (often referred to as

“transaction platforms”) that primarily facilitate observable transactions, often sales
transactions, between the two groups of platform users.
32. Transaction platforms exhibit particularly pronounced indirect network effects: the value of
participation on a transaction platform increases not only with the number of users on the

other side of the platform, but also with the usage of the platform on the other side. As a
result, transaction platforms are typically able to charge one or both sides both an access fee
for joining the platform and a usage fee for using it. Payment cards and auction houses are
familiar examples of transaction platforms discussed in the academic literature.1
33. It is important to emphasize that the product supplied by a transaction platform to the two
groups it serves is transactions. This is distinct from the exchange the platform facilitates as
a consequence between members of those groups. Payment cards like Amex facilitate
purchases and sales of a wide variety of goods and services, but they do not supply or
demand those goods or services. Transaction platforms generally compete with other
transaction platforms to facilitate transactions between the same general groups—merchants
and consumers in the payment card context, for example.
34. In our amicus brief in the Ohio v. American Express matter, Dr. Evans and I concluded that
for platforms that provide services that are “consumed jointly and unseverably”2 by
members of two distinct groups—a slightly different way of describing transaction
platforms—it makes no economic sense to consider inputs into transactions production in
isolation, as an exclusive focus on conduct with respect to only one of the groups involved
(merchants in the Amex case) would do.
35. An art gallery that displays and manages the sale of paintings owned by others is an
instructive example of a transaction platform. It is a platform because it exists to facilitate
interactions between artists and art buyers by choice of venue, curation, and other means. It
is a transaction platform because its business is to sell paintings. It is properly analyzed as
producing a single product: sales of paintings.
36. As a two-sided platform, a gallery creates value through the indirect network effects that it
offers to artists and art buyers in competition with other galleries and, perhaps, other
channels through which artists and art buyers can interact. Both artists and art buyers
benefit from the platform that the gallery offers for members of the two groups to interact.
In addition to facilitating the sale of artwork, galleries—as transaction platforms—typically
also offer a variety of services to artists and collectors on an ongoing basis. These services
ultimately benefit all parties as they can lead to potential transactions in the future. To
capture some of this value, galleries typically monetize by retaining between 30 percent and
60 percent of the artwork’s sales price. This commission is much higher than the payment
processing fees that the gallery pays as it compensates the gallery for all its services to
artists and art buyers.
1
Not all two-sided platforms are transaction platforms. Newspapers link advertisers and
readers, but no transaction between them happens on the newspapers’ platforms. Similarly,
shopping malls facilitate interaction between shoppers and merchants but are not involved
in any transactions between those two groups.
2
Brief for Amici Curiae Prof. David S. Evans and Prof. Richard Schmalensee in Support of
Respondents, State of Ohio et al. v. American Express Company et al., No. 16-1454,
Supreme Court of the United States, January 23, 2018, at p. 5.

37. A gallery does not need to charge a commission, however. It could instead choose to charge
art buyers a fee to enter the gallery (i.e., an access fee). However, this monetization strategy
could be counterproductive, as it might reduce the number of art buyers, which in turn
would likely reduce the value of the gallery to artists. Charging a percentage commission
per transaction, on the other hand, is much more in line with economic fundamentals.
38. Having chosen to charge a commission, galleries generally require that purchases using a
payment card be processed via the gallery’s payment terminal. If instead the buyer were
permitted to deal directly with the artist, in violation of this anti-circumvention rule, the
artist could avoid paying the gallery’s commission. Of course, if the gallery were able to
collect its commission even if the buyer paid the artist directly, it would make no economic
sense for the artist to take direct payment.
39. An artist who takes payment directly from a buyer for a work that a gallery had displayed
would avoid paying for the gallery’s display and other services and would thus be free-
riding. Free-riding artists would undermine the investments of galleries in display services
for artists, which would ultimately harm art buyers and artists alike. In this sense, the
gallery’s anti-circumvention rule is like the anti-steering provisions that Amex has in place
to deter merchants from “steering” customers away from using their Amex cards and
thereby making the network less attractive to customers and other merchants.
40. In addition, the gallery’s payment card terminal may generate data that is valuable to the
gallery and to artists, but, strictly speaking, it does not process payments. That is, it does
not arrange the transfer of wealth from the buyer to the gallery. The gallery owner—and
probably the artist—would be astonished to hear that the gallery’s commission should be
thought of as a markup on the use of its payment processing terminal. The gallery’s
commission simply reflects what the gallery views as the optimal monetization strategy for
the value that it offers to both sides of its transaction platform.
IV. The App Store Is a Two-Sided Transaction Platform
41. As Exhibit 1 shows, game console manufacturers, among others, had launched online stores
for app distribution before the debut of the App Store in 2008. Many other online game
stores have launched since. Many stores have posted commission rates of 30 percent, which
became a standard in the games industry.3
3
Marks, Tom, “Report: Steam’s 30% Cut Is Actually the Industry Standard,” IGN, January
13, 2020, available at https://www.ign.com/articles/2019/10/07/report-steams-30-cut-is-
actually-the-industry-standard.

Source: DX-4855, DX-4856, DX-4857, DX-4858, DX-5859

42. With the launch of the App Store, Apple’s iOS devices became a multi-sided software
platform connecting developers and users, with the App Store facilitating those connections,
like the platforms operated by the major game console producers and others.
43. Two-sided transaction platforms have three fundamental features: (1) two-sided transaction
platforms need to adopt pricing strategies, service provision strategies, and rules of behavior
to attract two distinct groups of users and to facilitate productive interactions between them;
(2) they derive substantial value from strong bilateral indirect network effects; and (3) they
have as their main purpose the facilitation of observable transactions that simultaneously

connect members of the two groups of users. These are often, but not always, sales
transactions.
44. The App Store displays all three of these fundamental features and is thus clearly a two-
sided transaction platform. First, to be viable, the App Store needs to attract both
consumers and developers. Second, there are clear bilateral indirect network effects here:
consumers want access to good apps (and the more good apps they have access to, the
better), and developers want access to as many potential customers as possible. Third, the
App Store generates value for both groups of customers when there is a transaction between
a consumer and a developer—a download or an in-app purchase of digital content from a
developer.
A. The App Store Works to Attract Consumers and Developers
45. With the launch of the App Store, Apple established a two-sided platform that thrives based
on the participation of both developers and device users. To remain competitive, Apple
must continuously think about how to attract users and developers,4 and may apply different
strategies over time or across product types.
46. One initiative that Apple has taken to make the iPhone live up to this description has been to
actively court game developers. It worked to persuade Epic to offer its games on the iOS
platform, for instance, as Epic’s popular titles would make the platform more attractive to
users, and thus encourage them to buy more iOS devices. In that context, Apple competes
with PlayStation, Nintendo, Xbox, and Android handset makers such as Samsung to make
Fortnite and other games as positive an experience as possible on iOS devices, so as to
attract gamers to the iOS platform.5 An internal document also indicates that Sony
competes with Apple’s App Store in digital games.6 In other contexts, e.g., email
applications, Apple does not compete with Nintendo but does compete with other handset
manufacturers as well as manufacturers of tablets, laptops and desktop computers to attract
email developers on the one hand, and consumers who wish to be able to send secure emails
conveniently on the other.
47. As Apple’s fact witnesses will testify, Apple has understood that consumers value security
and privacy of their information, and it has responded and continues to respond accordingly.
The App Store has been reviewing apps since its inception to provide consumers assurance
that iOS apps perform as advertised and don’t contain malware. Finally, over time Apple
has introduced significant changes to the functionality and design of the App Store,
including changes that facilitate the discovery and purchase of new apps and in-app content,
to make it more appealing to end-users.7
4
DX-4094 at 051-052, 080, 082-087, 098.
5
See, e.g., DX-3796 at 003-004.
6
DX-4389 at 001-002.
7
See, e.g., DX-4526, at 095; DX-3202; DX-4178 at 019.

48. Critically, the competition between software platforms does not end when one platform’s
app is selected by a user at some point in time. In fact, many users own devices running on
multiple operating systems, and can choose whether to use Device A or Device B to play a
game, draft an email, and so on. Hence, the platforms compete for usage in addition to app
purchases, because declining usage will reduce the attractiveness of the platform to
developers.
49. Through the App Store, Apple must also provide app developers incentives to develop and
maintain their apps for the iOS platform, instead of competing platforms such as Android or
Windows or, in the case of games, the Xbox, Switch, or PlayStation systems. Competing
platforms do not generally seek to persuade developers to permanently abandon rival
platforms. Sometimes Apple, Google, Samsung, and other platforms seek to persuade a
developer to develop for its platform first, to provide content that differentiates it from other
platforms, or to provide exclusivity for a limited period.
B. The App Store Has Strong Bilateral Indirect Network Effects That Have Shaped Its
Business Model
50. Greater consumer participation makes the App Store more attractive to developers, and
greater developer participation means more high-quality iOS apps that make the App Store
and iOS devices more attractive to consumers.
51. The App Store’s pricing policy is designed to encourage participation. In order to encourage
consumers’ participation, the App Store does not charge consumers access or transaction
fees, and its pricing structure encourages free apps. On the developer side, Apple provides
an array of powerful tools to create iOS apps at no charge, as well as promotional support,
and when developers pay a nominal fee of $99 per year they can offer their apps on the App
Store. In the art gallery example, this would be like a gallery giving paint, brushes, and
canvas for free to promising artists that agree to let it show their work.
52. Like other transaction platforms, Apple also establishes and enforces clear rules of behavior
to prevent platform participants from reducing its value for others. The app review process
is one example, and illustrates the importance of indirect network effects. Although the app
review process can be burdensome for app developers, once reviewed and certified, a well-
curated App Store with high-quality, bug-free iOS apps attracts more users and, ultimately,
more potential business for app developers. Developers have recognized that, much like
Open Table’s rule regarding diner no-shows, the review process, while onerous, creates
value. For example, in surveys I have reviewed, developers expressed dissatisfaction with
the limited review process in Android Market (the precursor to Google Play), with one
remarking: “Yes, it’s easier on the developer, but it’s detrimental to the whole ecosystem.”8
53. “Stores within a store” could direct consumers to less safe or lower quality apps that had not
gone through the review and curation process. While some developers may chafe at the
App Store’s banning of such stores, according to Mr. Sweeney, the fairness of a policy that
8
DX-4626, at 092.

bans stores within a platform’s store depends on “whether such a decision would have
overall positive consumer benefits or negative consumer benefits and follows the
expectations of consumers.”9 Other platforms through which Epic does business impose
restrictions similar to those imposed by the App Store.10
54. Dr. Evans notes that some of the conveniences that Apple offers for its App Store users
impose costs on developers.11 However, customer-friendly features like “Sign in with
Apple” along with other benefits such as parental controls and easy cancellation of
subscriptions will ultimately also benefit developers through increased customer satisfaction
via indirect network effects. This is similar to the restrictions on consumers for OpenTable
that I discussed above. Even though restrictions on one side of Apple’s two-sided platform
might be a mild inconvenience for some developers, they make the platform more valuable
to consumers, and if more consumers participate, the platform is more valuable to all
developers.
C. The App Store Has as Its Main Purpose Facilitating Observable Transactions
Between Developers and Consumers
55. Lastly, the App Store facilitates transactions between end users and developers—like Amex,
the App Store neither buys nor sells content; it facilitates transactions involving a range of
products. Transactions between iOS users and developers on the App Store include user
downloads of apps that are supplied by developers on the App Store, updates of apps that
are supplied by developers on the App Store, and in-app purchases of additional content by
iOS users. Developers, consumers, and Apple benefit from the App Store only when
transactions occur.
56. Similar to other transaction platforms, the App Store simultaneously connects iOS users and
developers and facilitates observable transactions between these two sides. The App Store
also exhibits particularly pronounced indirect network effects: the value to developers of
participation on the App Store increases with the usage of the App Store by iOS users.
57. As such, the App Store price structure must reflect Apple’s view of relative price
sensitivities as well as the nature and strength of indirect network effects on the two sides.
In order to offer apps in the App Store, developers must pay a nominal annual access fee
($99) to join the Apple Developer Program. When transactions involving apps they have
offered on the App Store occur, developers are charged commissions that vary with the
nature of the transactions. Like most other online stores, the App Store requires the use of
its own payment system (In-App Purchase, or “IAP,” for in-app transactions) in order to
efficiently collect the commissions it is owed.
9
Deposition of Tim Sweeney, February 8, 2021 (“Sweeney Deposition”), at 72:19-73:4.
10
DX-3298; “DX-4922; DX-3583, at 002; DX-3437, at 001, 003.
11
Written Direct Testimony of Dr. David S. Evans, Epic Games, Inc., v. Apple Inc., No. 4:20-
CV-05640-YGR, United States District Court for the Northern District of California,
Oakland Division, May 3, 2021 (“Evans Testimony”), at ¶ 199.

58. Economic analysis of the App Store and its policies must reflect the reality that it is a two-
sided transaction platform with strong bilateral indirect network effects. Because of the
complexities of two-sided platform business models, economic analysis of such businesses
is inevitably more complex than the analysis of familiar single-sided businesses like
manufacturing firms. Concluding that the App Store is a two-sided transaction platform is
just the starting point for analyzing antitrust questions of product market definition, market
power, and competitive effects. The key is the careful application of these principles while
bearing complexities of platform businesses in mind.
V. Dr. Evans’ Treatment of Platforms and Markets is Internally Inconsistent and
Disconnected From the Issues in This Case
59. During the preliminary injunction phase, Dr. Evans declined to express an opinion about
whether the App Store is a two-sided transaction platform. I am pleased that Dr. Evans now
admits that the App Store is a two-sided transactions platform,12 and I agree that the iOS
system, after the introduction of the App Store, is also a platform.13
60. Dr. Evans asserts that Apple participates in two two-sided markets that link consumers and
app developers: a foremarket for “smartphone OSs” and an aftermarket for distribution of
iOS apps.14 He seems to be seeking to direct attention away from Apple’s impacts on
transactions between Epic and similarly situated game developers and consumers—impacts
with which this case is exclusively concerned. As I show in what follows, this artificial
structure makes no economic sense.
A. Dr. Evans’ Assertion of Separate Foremarkets and Aftermarkets Is Artificial and a
Distraction from the Issues in this Case
61. Dr. Evans devotes over 12 pages of his written testimony to discussions of competition and
market power in what he defines as a foremarket for smartphone operating systems.15 Dr.
Evans’ foremarket has one particularly unusual, if not unique, feature. Apple’s iOS is not
licensed to OEMs or sold separately to consumers, and Google’s Android OS is not licensed
to OEMs at a positive price and is not sold separately to consumers. Thus Dr. Evans has
asserted a very unusual market: neither of the products that he asserts dominate the market
are in fact sold or licensed at a positive price.
12
Evans Testimony, at ¶ 7.i.C.
13
Note, though, that Dr. Evans’ revised framing is inconsistent with Epic’s Complaint, which
deals only with a single-sided “iOS App Distribution Market,” in which Apple provides
“services that allow users to find new apps they desire to download and that make new apps
and app updates seamlessly available for download and update.”
14
Evans Testimony, at ¶¶ 40, 43-44, 92.
15
Evans Testimony, at ¶¶ 40-91.

62. Dr. Evans also defines an aftermarket for “iOS app distribution.”16 This terminology misses
the main purpose of the App Store as a two-sided transaction platform. Focusing only on
the services provided to developers is precisely the economic error committed by plaintiffs
in Amex when they focused only on the merchant side of payments card platforms.
63. Dr. Evans’ division of the Apple ecosystem into fore- and aftermarkets is inconsistent with
the existence of strong bilateral indirect network effects, which he seems to accept.17 It is
uncontested that improving the quantity and quality of apps available for a particular
operating system stimulates sales of phones or other devices that use that operating system.
Similarly, improving the devices that use a particular operating system (via better hardware
or improvements to the bundled operating system) will make that operating system more
attractive to developers. With strong indirect network effects, developer behavior affects
consumer demand, even if some consumers face substantial switching costs, and consumer
choices affect developer behavior. If fewer good apps are available for iOS devices, there
will be less demand for those devices when consumers upgrade, as they do on average every
four years. If fewer consumers use iOS devices, developers can easily switch efforts to
other platforms. This does not require abandoning that platform, merely altering priorities
regarding where an app first appears, where marketing efforts are focused, and similar
adjustments at the margin.
64. Dr. Evans asserts that his fore- and aftermarkets are separated in part because the foremarket
for smartphone operating systems is not competitive, so foremarket competition does not
discipline aftermarket prices.18 There is, of course, no real foremarket in which consumers
pay for mobile operating systems. In the real world, consumers choose among bundles that
consist of smartphones (or other mobile devices) plus mobile operating systems. This real
market is far from a duopoly: Apple faces numerous large competitors who get their
Android operating system for free, and Apple accounted for only 15 percent of global
smartphone sales in 2020.19 Apple’s ability to compete in that marketplace depends on its
ability to offer consumers a large selection of high-quality apps—and thus on making the
App Store attractive to developers as well as consumers.
65. Even if competition for bundles consisting of devices and operating systems were not
intense, it would have no implications for the issues in this case. This case is about
transactions between app developers and consumers, not about competition among
operating systems or devices that use them.
16
Evans Testimony, at ¶ 92.
17
Evans Testimony, at ¶ 7.i.C-D.
18
19
“Gartner Says Worldwide Smartphone Sales Declined 5% in Fourth Quarter of 2020,”
Gartner, February 22, 2021, available at https://www.gartner.com/en/newsroom/press-
releases/2021-02-22-4q20-smartphone-market-share-release.

66. Dr. Evans seems to suggest that the mere existence of switching costs on one side of a
transaction platform automatically means that there are economically distinct fore- and
aftermarkets.20 In the first place, it is important to keep in mind that switching costs may
reflect product differentiation, as some consumers are particularly attached to the unique
features or functions provided by their favorite platform. Being reluctant to switch because
I prefer the product I use to available substitutes hardly signifies a competitive problem.
The development of unique features that may make consumers reluctant to switch is a
central focus of competition in technology-intensive products; switching costs that derive
from them are a symptom of competition, not a sign of its absence.
67. Moreover, the magnitude of switching behavior, as measured by the willingness of
participants on either side to switch to platform B due to an increase in price of platform A,
will determine whether or not platforms A and B are in the same market. Dr. Evans’ failure
to estimate switching costs (which surely vary among consumers and developers) or indirect
network effects on either side of the platform (despite accepting that both his markets center
on two-sided transaction platforms, for which indirect network effects are strong by
definition) demonstrates why his division of fore- and aftermarket is artificial.
68. The correct way to analyze iOS and the App Store is as a single, integrated, two-sided
transaction platform, in which the App Store is Apple’s mechanism for facilitating
transactions between developers and end-users. There is only a single two-sided transaction
platform that facilitates interactions between developers and end users in the iOS ecosystem,
rather than two—as Dr. Evans suggests—that overlap almost completely. And, since Epic’s
allegations concern this transaction platform, the product market must center on the
transactions facilitated by this platform. Similarly, any assessment of the boundaries of the
market must include an analysis that seeks to identify transactions facilitated by other
transaction platforms that are sufficiently substitutable from the perspective of developers
and end users.
B. Dr. Evans’ Discussion of Smartphones and Other Devices Is Another Distraction
69. In his discussion of his foremarket for smartphone operating systems, Dr. Evans frequently
conflates smartphone operating systems with smartphone devices.21 In fact, Dr. Evans
devotes a large part of his discussion of the definition of the foremarket whether other
devices are practical substitutes for consumers and whether developers can use other
devices to provide services to consumers who have smartphones.22
70. The relevant question for the purposes of defining antitrust markets is not which devices are
functionally equivalent, but rather what transactions facilitated by other platforms are
reasonably substitutable for those facilitated by the App Store from the perspective of
20
Evans Testimony, at ¶¶ 73, 118.
21
22
Evans Testimony, at ¶¶ 50-53, 55, 58, 60-61.

developers and end users. Some of those other platforms involve desktop computers,
laptops, game consoles, and tablets.
71. Rather than focusing on reasonably substitutable transactions from the perspective of
developers and end users, Dr. Evans focuses on the functional differences between the
hardware associated with various platforms, and argues that developers have no other way
to reach iOS users because users cannot easily replace a smartphone with another device
like a console or a PC. That a console or a PC is not functionally equivalent to a
smartphone is of course true but irrelevant. Substitution does not need to occur literally
with a user replacing their smartphone with a gaming console. Rather, the relevant
dimension of substitution occurs at the transaction level, where, for instance, a user may
choose to purchase V-Bucks on the Microsoft Store on Xbox rather than on the iOS App
Store.
72. Therefore, Dr. Evans’ claim that Apple has foreclosed access to approximately 1 billion
iPhone users globally23 is simply wrong. In fact, as shown by Professor Hitt, the vast
majority of iOS users already have access to multiple types of devices that can be used to
conduct game transactions, including smartphones, tablets, personal computers, laptops, and
gaming consoles. Moreover, not only do consumers have access to multiple devices,
Professor Hitt also shows that consumers also regularly make transactions through multiple
online transaction platforms.
C. Dr. Evans’ Analysis Condemns All “Walled Garden” Business Models, Including
Those of Game Console Manufacturers
73. If one were to follow Dr. Evans’ foremarket/aftermarket logic, one would have to conclude
that all game console manufacturers monopolize the distribution of games on their
platforms.
74. The core economic issue in this case is whether the iOS “walled garden” model, adopted
well before Apple sold its first 4G phone and similar to the models of the game console
makers, now constitutes an antitrust violation. As this Court already noted, all game
console manufacturers impose restrictions designed to ensure that purchases of apps and in-
app upgrades for their operating systems platform must go through their app marketplaces.
Like Apple, they also moderate content and prohibit stores-within-a-store as well as
sideloading.24 All charge commissions on purchases of apps and in-app upgrades. Once a
user has purchased a console and games that can be played on it, she would incur switching
costs to move to another console. In the presence of switching costs, which Dr. Evans
contends limit the impact of “foremarket” competition on “aftermarket” competition,25 their
23
24
DX-4434; DX-3258, at 011.
25

business models contain the same exact core elements as the iOS business model that Epic
contends violates the Sherman Act.
75. Dr. Evans seeks to downplay the drastic implications of his analyses, asserting simply that
“The App Store is not just a typical platform like Nintendo, operating as a walled garden.”26
He does not attempt to explain why the size and importance of the App Store relative to
others that are similarly structured should matter for antitrust analysis, let alone attempt to
indicate what size should be the threshold between legality and illegality for walled garden
business models.
76. In short, Dr. Evans’ analysis implies that the “walled garden” business models operated by
game console manufacturers should be demolished along with the App Store’s model, with
unpredictable effects. The Court noted explicitly in its Preliminary Injunction Order that its
“final decision should be better informed regarding the impact of the walled garden model
given the potential for significant and serious ramifications for Sony, Nintendo and
Microsoft and their video game platforms.”27 Dr. Evans provides no meaningful analysis of
those ramifications.
VI. Implications of the Fact that the App Store Is a Two-Sided Transaction Platform for
Market Definition, Assessment of Market Power, and Analysis of Competitive Effects
77. The conclusion that the App Store is a two-sided transaction platform has important
implications for defining the relevant market, evaluating claims of market power, and
analyzing whether Apple’s conduct has harmed competition. Dr. Evans’ analysis of these
issues does not do justice to those implications.
A. Implications for Defining Product Markets
78. In prior writings, Dr. Evans has shown that market definition analysis of markets involving
platforms must take explicit account of the strength of indirect network effects, but he does
not do so here. Moreover his foremarket SSNIP test is uncoupled from economic reality,
and his aftermarket SSNIP test makes no sense.
1. Dr. Evans and Other Authors Show that Market Definition Involving
Platforms Must Take Explicit Account of the Strength of Indirect
Network Effects
79. Antitrust practitioners often rely on the hypothetical monopolist test to delineate the relevant
product and geographic markets for single-sided businesses, particularly in the merger
context. The test involves assessing whether a hypothetical profit-maximizing firm, not
subject to price regulation, that was the only present and future seller of a set of products
likely would find it profitable to impose at least a small but significant and non-transitory
increase in price (SSNIP) on at least one product in the market. If a SSNIP on a candidate
market does not increase the hypothetical monopolist’s profit, then the candidate market is
26
27
Dkt. 118 at p. 18.

too narrow, and more products should be added into the candidate market until such a price
increase would increase the hypothetical monopolist’s profit.
80. Applying a conventional one-sided hypothetical monopolist test to only one side of a
transaction platform is inappropriate because it ignores the indirect effect that a price
increase would have on participation of the other side, which in turn would lower the
participation of the tested side. By not accounting for indirect network effects, a one-sided
hypothetical monopolist test would underestimate the loss in profits due to an increase in
price. As a consequence, the one-sided hypothetical monopolist test might incorrectly
predict a positive profit for the hypothetical monopolist, and the market would be defined
too narrowly.
81. Even if applied on both sides of the transaction platform, though, correct application of the
hypothetical monopolist test to markets involving platforms turns out to be quite complex.
The presence of indirect network effects may make its quantitative application
impractically complex, since strengths of indirect network effects are typically hard to
quantify.
Second, there is no consensus among economists about the price to which the SSNIP
should be applied. In a two-sided market, a platform sets at least two prices (at least one
for each side), thus determining both a price level and a price structure. It is not clear
how, if at all, the hypothetical monopolist should be thought of as adjusting the price
structure while raising the price level.
Finally, considering the App Store and other online stores, there is no single best
measure of volume. Store revenue is an obvious measure, but it gives no weight to free
downloads even though they benefit both developers and consumers—and thus benefit
Apple.
82. Economists—including Dr. Evans—have proposed different versions of the hypothetical
monopolist test for platform markets, but each is complex and requires multiple inputs that
are generally difficult to estimate. And, as Dr. Evans and I have written, a two-sided SSNIP
test “requires a complete structural model of the firms’ demands, including both cross-price
effects and indirect network effects, a good deal more information than is necessary in the
case of ordinary single-sided firms.”28 These quantities are recognized to be hard to
estimate.
83. In his testimony, Dr. Evans advances multiple SSNIP tests that he claims show that both his
foremarket and aftermarket are relevant antitrust markets. But though his analysis may
appear rigorous, it effectively ignores the indirect network effects that are vital here, is
biased toward overly narrow markets, and thus provides no reliable, relevant evidence.
28
See Evans, David S. and Richard Schmalensee, “The Antitrust Analysis of Multi-Sided
Platform Businesses,” NBER Working Paper 18783, February 2013, available at
https://www.nber.org/system/files/working_papers/w18783/w18783.pdf, at p. 24.

2. Despite His Prior Writings, Dr. Evans’ SSNIP Tests Neglect Indirect
Network Effects and Are Thus Conceptually Flawed
84. Dr. Evans’ foremarket and aftermarket SSNIP tests are conceptually flawed because they
completely neglect indirect network effects that he has admitted exist here and are thus
inconsistent with the literature, including Dr. Evans’ own research.
85. Strong indirect network results imply that the ultimate impact of a price increase to one side
of a platform can be considerably greater than the initial impact. A price increase to one
side, A, of a platform will reduce participation on that side, which will make the platform
less attractive to participants on the other side, B, and they will reduce participation. This,
in turn will reduce participation on side A, and so on. Since the Hypothetical Monopolist
and SSNIP tests turn on responses to price increases, ignoring indirect network effects will
generally lead to markets that are too narrow.
86. A simple example illustrates the problem—and Dr. Evans’ error. Suppose a 10 percent
price increase on side A of a two-sided platform reduces participation on that side from 100
to 95, reflecting an own-price elasticity of -0.5. Such a price increase might well be
profitable. Now suppose that strong indirect network effects give rise to a within-platform,
cross-side externality elasticity of 0.6. That is, a 10 percent reduction in participation on
either side, all else equal, would lead to a reduction of 6 percent in participation on the other
side. If participation on side B was also initially 100, the initial 5 percent reduction in
participation on side A would thus then lead to a 3 percent (=0.6*5) reduction on side B,
from 100 to 97. But that reduction in participation on side B would make the platform less
attractive to the group on side A. With a within-platform, cross-side externality elasticity of
0.6, the 3 percent reduction in participation on side B would lead to an additional 1.8
percent reduction in participation on side A. When the process converges, participation on
side A will have declined by just over 7.8 percent, which may be enough to make the initial
increase in the side A price unprofitable.
87. Dr. Evans, however, does not perform any actual SSNIP calculations that take into account
both sides of the platform simultaneously for either the foremarket or the aftermarket.
Instead, he conducts his foremarket and aftermarket SSNIP tests on the consumer side and
on the developer side in isolation.29 Then, he effectively dismisses indirect network effects
by claiming that a SSNIP on both developers and consumers would be profitable because
neither side would respond to the one-sided price increases he tested. This ignores the fact
that strong indirect network effects imply that any impact on consumers would affect
developers on the margin (and vice versa).
88. This failure invalidates Dr. Evans’ results and conclusions. As Dr. Evans himself showed in
multiple articles, employing a one-sided test and ignoring indirect network effects would
29

generally lead to markets that are too narrow.30 To correct for this error, Dr. Evans
previously prescribed a test calling for a total of 14 inputs, which includes additional inputs
well beyond what a single-sided SSNIP test would require, with many specifically related to
within-platform and cross-platform cross-side externality elasticities to capture the indirect
network effects. Yet, as shown in Table 1, in his analysis for this matter, Dr. Evans
completely abandons this methodology in his attempt to define his candidate markets.
Table 1 - Two-Sided Platform SSNIP Test and the Required Inputs31
Evans and Noel Dr. Evans’

Inputs Side (2005 and 2007) Foremarket SSNIP
Consumer
Percentage Increase in Price
Developer
Consumer
Margin
Developer
Consumer
Revenue
Developer
Own-Price Elasticity of Consumer ?

Demand Developer ?
Cross-Price Elasticity of Consumer

Demand Developer
Within-Platform, Cross-side Consumer

Externality Elasticity Developer
Cross-Platform, Cross-side Consumer

Externality Elasticity Developer

30
See Evans, David S. and Michael D. Noel, “Analyzing Market Definition and Power in
Multi-Sided Platform Markets,” October 21, 2005, available at
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=835504; see Evans, David S. and
Michael D. Noel, “Defining Markets That Involve Multi-Sided Platform Businesses: An
Empirical Framework with an Application to Google’s Purchase of DoubleClick,”
November 2007, available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1027933,
at p. 12 .
31
Some inputs are marked as “?” because Dr. Evans does discuss them, but he offers no
quantitative evidence which his prior writings argue are necessary inputs for a valid test.

89. There is no reason to believe that, had Dr. Evans performed a SSNIP analysis consistent
with his prior writings, inclusion of non-zero indirect network elasticities would not have
reversed his conclusions.
3. Dr. Evans’ SSNIP Tests are Otherwise Unreliable and Uninformative
90. Besides ignoring the importance of indirect network effects, Dr. Evans’ foremarket and
aftermarket SSNIP tests are each deeply flawed in their own right. The first is unrelated to
the issues in this case; the latter simply makes no sense.
a) The Foremarket SSNIP Test Is Uncoupled from Economic Reality
91. Dr. Evans describes his foremarket SSNIP test as an attempt to see whether Apple and
Google could profitably increase the prices of their mobile operating systems by a small but
significant non-transitory increase in price (a SSNIP).32 But whether or not a single seller of
mobile operating systems or smartphones could profitably exercise monopoly power has no
bearing on any issue in this case. Epic’s Complaint makes no claim for relief regarding any
alleged market for operating systems or devices. There is no such relevant market, and Dr.
Evans’ foremarket SSNIP test is merely part of an elaborate attempt to distract.
92. Dr. Evans’ foremarket SSNIP test is defective as well as irrelevant. He ignores the fact that
the relevant product is transactions, and attempts to apply a SSNIP to the price of the OS,
rather than the transaction between developer and user.33 He is forced to engage in
numerical gymnastics to come up with his view of what such a price increase might mean,
since Apple has never licensed iOS separately, and Google has never charged for an
Android license.34 Dr. Evans arbitrarily divides the price of handsets between hardware and
a very low competitive price for iOS, which virtually guarantees that the SSNIP passes.35
Had he used the Android price—zero—as a benchmark, passage (i.e. the SSNIP being
profitable) would have been assured. Alternatively, and no more arbitrarily, one could
decide that the implicit price of iOS should be the difference between, say, the price of an
iPhone and the marginal cost of manufacturing it. With that much higher price, passage
would have been much less likely.
32
33
Evans Testimony, at ¶ 67 (“It is appropriate to base the SSNIP on the price of the OS rather
than the device”).
34
35
Evans Testimony, at ¶ 67 (“It is possible to estimate the implicit competitive price for the
OS based on license fees—about $30—that Microsoft charged for Windows Phone before it
exited. The fees developers pay are small, and they are fixed costs that cover all users, and
can therefore be ignored in calculating the total price to app users and developers. It is
appropriate to base the SSNIP on the price of the OS.”).

b) Dr. Evans’ Aftermarket SSNIP Test Makes No Sense
93. As already noted, the hypothetical monopolist and SSNIP tests were originally designed to
aid merger analysis by testing whether multiple firms in a candidate market, if acting
together, could profitably raise price. The setup in Dr. Evans’ aftermarket SSNIP test is
very different. His proposed aftermarket for iOS app distribution has only one seller, the
Apple App Store, which is thus, by his definition, a monopolist.36 His SSNIP test thus can
only be an attempt to assess whether the App Store, acting alone, could profitably increase
price—as, he says, it could.37
94. Dr. Evans conducts his aftermarket SSNIP test in three steps. First, he raises the price that
consumers pay for all in-app purchases/subscriptions by 5 percent and, holding app supply
constant, argues that the price increase would be profitable for the hypothetical
monopolist.38 Second, Dr. Evans increases the price of iOS app distribution for Epic by ten
percent, and, holding app demand constant, argues that Epic would rather continue to use
the App Store and pay the ten percent increase in the commission rate.39 Finally, Dr. Evans
argues a SSNIP on both developers and consumers would be profitable because neither side
would respond to the one-sided price increases he tested.40 This, of course, neglects the fact
that indirect network effects amplify responses to price changes, so that any fall in consumer
participation would make the App Store less attractive to developers, including Epic.
95. Dr. Evans concludes that “Apple could have increased its profits by by
41
imposing this price increase.” The hypothetical monopolist is of course Apple, and Dr.
Evans holds app demand constant in his analysis because he neglects the fact that a
reduction in development of iOS apps would reduce iOS device sales and thus reduce the
demand for apps. This would lead to a shift in developer attention away from iOS, reducing
App Store revenues—perhaps enough to make the SSNIP unprofitable.
96. But taken at face value, what, exactly does Dr. Evans’ conclusion that the App Store could
profitably raise prices mean? A single-firm SSNIP test is problematic because if any
individual business is maximizing profits then, by definition, it could not increase profits by
raising prices. In 1956, the Supreme Court fell victim to what economists have since
dubbed the Cellophane Fallacy by neglecting this fact.42 Despite other evidence indicating
36
37
38
39
40
41
42
United States v. E.I. du Pont de Nemours & Co., 351 U.S. 377, 1956. See, e.g., Baker,
Jonathan, “Market Definition: An Analytical Overview,” Antitrust Law Journal, Vol. 74,
No. 1, 2007, at p. 164.

that DuPont, the only producer of cellophane, was pricing cellophane well above
competitive levels, the Supreme Court found that it was not a monopoly. It relied on
evidence that consumers considered cellophane to have good substitutes to which many
would have switched if DuPont had increased cellophane’s price. If that had not been the
case, of course, DuPont would have profitably increased price. In modern language, a
proper SSNIP test would have predicted that a price increase for cellophane would have
reduced DuPont’s profits, and this prediction would have led automatically to the
(apparently erroneous) conclusion that a market confined to cellophane was too narrow.
97. Dr. Evans’ interpretation of his aftermarket SSNIP test suffers from an inverse Cellophane
Fallacy. Dr. Evans claims that Apple is a monopolist in the aftermarket, and therefore
Apple’s profit-maximizing price should already be above the competitive level, and yet Dr.
Evans’ SSNIP test finds that Apple as a monopolist could still raise price even beyond the
allegedly monopolistic price it already charges.
98. So how can Dr. Evans possibly conclude that the App Store, acting alone, would make more
money by raising prices? There are only two possibilities. The first is that Dr. Evans is
simply wrong: his neglect of indirect network effects, despite the prescriptions of his own
writings, has, predictably, led him to underestimate the revenue losses due to price changes.
The other shortcomings of his analysis could also have contributed to an incorrect result.
The other possibility is that despite the defects in his analysis, Dr. Evans is right: the App
Store could increase its profits over a short but non-transitory period by charging more. In
this case it is not maximizing short-run profit. The App Store is part of the iOS platform,
after all, not a stand-alone business. It is not implausible that competitive forces may be
compelling Apple to hold the App Store commission rates below the ostensibly short run
profit-maximizing levels in order to encourage app transactions to the benefit of the iOS
business as a whole in the long run. That is certainly plausible behavior, indicating once
again that it makes little sense to analyze App Store pricing in isolation.
99. Dr. Evans argues that, “it is likely that Apple has not increased its commission rate over the
current 30% maximum given the high level of regulatory, media, and developer scrutiny it
has faced over several years.”43 This explanation seems much less plausible than the one
just given: it ignores the fact that the App Store is part of the iOS platform, that the 30
percent rate was established two years before Dr. Evans believes the App Store attained
monopoly power,44 and that the scrutiny to which he refers is a very recent phenomenon.
100. Does it matter which of these possibilities is correct? As a matter of antitrust economics, it
does not. Apple has made the App Store the sole platform connecting iOS app developers
that use Apple software with iOS device users, and it requires the use of IAP for in-app
purchases. The App Store handles a substantial volume of commerce to the benefit of
consumers, developers, and Apple itself. These facts are not in dispute. Whether or not the
App Store is labeled a relevant antitrust market has no bearing on the economic analysis of
43
Evans Testimony at ¶ 142.
44

the effects of the policies of which Epic complains. Once Dr. Evans has affixed that label,
he simply moves on to an analysis of competitive effects.
4. Dr. Evans Ignores the Fact that Digital Game Transactions Face
Different Competitive Conditions Compared to Non-Game
Transactions
101. It is important to recognize that a single transaction platform may face different competitive
constraints for different types of transactions. If those competitive constraints differ
substantially, it will be appropriate to consider the different transactions types to be in
separate relevant markets.
102. This is not a novel point, nor is it specific to platforms. It is my understanding, for instance,
that the Ford Motor Company produces one basic product: motor vehicles. It may face a
rather different set of competitors for its trucks than for its sedans, however. If this is the
case, and if an antitrust case centers on Ford trucks, a proper analysis may focus on the
market for trucks, while taking into account supply-side and demand-side constraints from
outside that market. Firms that do not produce trucks may not belong in the relevant
market, though the ease with which they could enter that market should be considered.
103. It is also important to determine the competitive constraints faced by the App Store and the
extent to which other platforms are sufficiently substitutable from the perspective of
developers or consumers to be included in the market. For a transaction platform, the
relevant question is reasonable interchangeability of transactions involving groups served by
the platform at the center. As a matter of economics, two-sided platforms face competition
from other similarly situated two-sided platforms that facilitate transactions between the
same two sets of users. As Exhibit 1 indicates, a large number of firms have chosen, as
Apple did, to develop their own app stores.
104. While users and developers of certain apps, such as ridesharing, may not have sufficiently
substitutable transactions platforms to smartphones and tablets, for other apps, in particular
games such as Fortnite, developers and end users have many reasonably substitutable
transactions platforms, including consoles, personal computers, streaming services, smart
TVs, and virtual reality gear. The competitive conditions for digital game transactions may
be different from those of non-game transactions because of the presence of significant
transaction platforms that focus exclusively on digital game transactions, e.g., the
PlayStation Store and Nintendo eShop. I therefore disagree with Dr. Evans’ assertion that
“[l]ooking only at one category of demand, such as for games, would not accurately capture
the nature of the competition that these platforms face.”45
105. Apple’s conduct at issue may have different competitive effects depending on the app or
app genre at issue. When it comes to market definition—which is, after all, simply a tool
for helping to measure competitive effects—a market like Dr. Evans’ market that
encompasses every app in the App Store is unlikely to be informative. The App Store
45

contains apps dealing with crocheting and biochemistry, among a wide variety of other
subjects. The notion that transactions involving all these app are reasonable substitutes for
consumers or developers for transactions involving game apps seems implausible. Dr.
Evans has certainly offered no evidence on this point.
106. The relevant product market for game transactions is certainly broader than the App Store. I
do not have sufficient data to conduct an empirical SSNIP test to delineate with precision
the boundaries of the relevant product market; neither does Dr. Evans. Nor is it necessary to
conduct such a test to reach sound conclusions. If the data were available, I believe that a
properly conducted hypothetical monopolist test would lead to the conclusion that the
relevant market for game transactions is broader than the App Store.
107. As Professor Hitt shows, the facts on the ground are clear: the quantitative and qualitative
evidence demonstrates that Epic and similarly situated game developers transact with
customers across a wide range of transactions platforms, including on consoles such as Sony
PlayStation, Microsoft Xbox, Nintendo Switch, as well as on personal computers and
mobile devices such as smartphones and tablets. In addition, cloud-based streaming
services may soon become meaningful competitive constraints to all other game transaction
platforms. I therefore agree with Professors Lafontaine and Hitt that a relevant product
market for game transactions on digital transaction platforms is the appropriate antitrust
market to assess Epic’s claims in this matter.
B. Implications for Assessing Market Power
108. As discussed above, unlike one-sided businesses, two-sided platforms tend to have a skewed
pricing structure. Platforms often charge zero or even a negative price to one side if demand
is highly price-sensitive and/or if its participation is critical to attracting members of the
other side. Furthermore, the profit-maximizing, non-predatory price charged on one side
may be above, equal, or below the marginal cost specific to that side. The App Store’s
asymmetric and complex pricing structure reflects differences in the price elasticities of the
two sides and the strengths of the indirect effect that each exerts on the other, and a desire to
encourage small developers and free apps.
109. If the prices charged by a two-sided transaction platform have not changed, their level
provides no information regarding market power. Whatever market power a platform may
have, if, like the App Store, it has only reduced elements of its price structure over time, it
has not exercised market power.
1. Dr. Evans Wrongly Asserts that High Operating Margins Imply
Monopoly Power
110. As evidence of Apple’s purported monopoly power in app distribution, Dr. Evans points to
the supposedly high profits earned by Apple from the App Store, relying on analysis by Mr.
Barnes, an accountant.46
46

111. First, high accounting profitability, even of a stand-alone business, does not establish the
existence of market power, let alone monopoly power. It is well-known that accounting
profitability can be a very poor measure of true, economic profitability, particularly for
businesses that make substantial investments in intellectual property.47 Moreover, while a
persistently high economic profit rate is suggestive of market power, it is by no means
dispositive.
112. Mr. Barnes’ comparison of the App Store to other online platforms such as eBay, Etsy,
Rakuten, and Mercado Libre ignores the fact that many other successful technology
companies, large and small, earn similar, and at times higher, operating margins than Apple
as returns on their intellectual property.48 Exhibit 2 summarizes operating profit margins of
publicly-traded companies, for which financial statements are publicly available, that
operate online stores in Exhibit 1. Though Epic is not publicly traded, I also include its
operating margins based on financial information produced in this matter.
Source: DX-4861
113. A simple example illustrates the basic problem with using the operating margin as an
estimator of profitability without taking into account firms’ investments in intellectual
47
Fisher, Franklin M. and John J. McGowan, “On the Misuse of Accounting Rates of Return
to Infer Monopoly Profits,” American Economic Review, 73(1), 1983, pp. 82-97, at p. 82;
Bork, Robert H. and J. Gregory Sidak, “The Misuse of Profit Margins to Infer Market
Power,” Journal of Competition Law & Economics, 9(3), 2013, pp. 511–530, at pp. 514-
515.
48
Written Direct Testimony of Ned S. Barnes, CPA, Epic Games, Inc., v. Apple Inc., No.
4:20-CV-05640-YGR, United States District Court for the Northern District of California,
Oakland Division, May 3, 2021 (“Barnes Testimony”), at ¶¶ 25-27.

property and other assets. Consider a set of Kansas wheat farms, none of which has any
market power. Because soil fertility varies, so will wheat harvests per acre, and thus so will
operating margins. But farms with more fertile soil will be more expensive to buy, and
under perfect competition, the higher cost of farms with better soil will just compensate for
their higher operating margins, and buying and operating any of the farms will yield roughly
the same competitive rate of return.
114. In this industry, intellectual property plays the role of soil fertility. Successful investments
in intellectual property create an asset that sustains positive operating margins. If one just
looks at operating margins and ignores the investments in intellectual property on which
they rest, one is biased toward finding market power where none exists, just as ignoring the
costs of Kansas wheat farms and looking only at their operating margins would mislead one
into thinking that some of them have market power.
2. Mr. Barnes’ Estimate of the App Store’s Operating Margin Relies on

an Inevitably Arbitrary Allocation of Joint Costs and is Accordingly
Unreliable
115. It makes no sense to try to measure the App Store’s profitability in isolation; it is a part of
the iOS platform and relies on all of Apple’s intellectual property. The costs associated
with that asset are joint costs that simply cannot be allocated in an economically defensible
way to individual operating segments.49 (The classic example of a joint cost is the cost of
raising a steer that yields both meat and hide. Any allocation of that cost between meat and
hide is purely arbitrary.) Thus any accounting measure of the App Store’s stand-alone
profitability is also arbitrary and thus unreliable as an indicator of anything. 50 As top Apple
executives will testify, Apple does not calculate P&Ls by products and services because
they view it as an unproductive exercise. Exhibit 2 shows that in comparison with other
firms with substantial investments in intellectual property, Apple’s operating margin is
unremarkable.
116. In the iOS business, research and development (R&D) and related expenses are a clear and
important example of a joint cost. R&D investment that adds useful features to the OS or to
iPhone hardware, for instance, is likely to increase iPhone sales. It is also likely to make
49
Epic Games’ financial statements also highlight how margins can vary across different
products and that joint costs cannot be allocated to different products arbitrarily. For
example, in 2018, Games had a gross margin of 66 percent, while the Epic Games Store
had a gross margin of 12 percent. Corporate expenses and operating expenses related to
online services were not divided across Games, Engine, and the Epic Games Store. See
EPIC_00001689.
50
Mr. Barnes’ accounting estimates rely on documents that use economically arbitrary
methods of allocating joint costs, such as revenue-based allocation, which ignores the fact
that the App Store’s share of Apple’s revenue may be a very poor indicator of the extent to
which it benefits from various joint costs.

developing apps for those phones more attractive, thus stimulating transactions on the App
Store. Similarly, investments that makes the App Store more user-friendly, whether or not
they are formally classified as R&D, will both stimulate usage of the App Store and make it
more attractive to own an iPhone, thus stimulating iPhone sales.
117. Dr. Evans argues that the App Store’s operating margin implies that it does not need a 30
percent commission rate to be profitable.51 This suggestion seems unrelated to any
allegations in Epic’s complaint. Moreover, as an economic matter, it is unclear why Dr.
Evans focuses on only one element of Apple’s pricing policy. Like any profitable business,
Apple could, of course, alter its pricing structure, including iPhone, iPad, and App Store
prices, in a variety of ways. If it were forced to change any element of that structure, it
would likely find it optimal to change other elements in response.
118. Platforms’ pricing structures reflect the patterns of price sensitivities and indirect network
effects they face. It is true that many platforms earn most or all of their profit from only one
of the groups they serve. But, despite Dr. Evans’ suggestions to the contrary,52 Apple is
hardly unique in earning profits from both groups it serves. Apple’s pricing structure is not
a sign of market power or of its absence. Nor is any supposed change in the App Store’s
business model after Steve Jobs’ 2008 statement that it would be run on a break-even basis.
3. The Available Reliable Evidence Indicates that the App Store Lacks
Monopoly Power
119. Market share is sometimes used as an indicator of market or monopoly power. In using a
firm’s market share for this purpose, it is important to consider its durability, that is, its
ability to persist for a significant period without being eroded by competitive entry or
expansion. In the absence of significant barriers to new entry or expansion, a firm with
large market share today could find its share and its market power reduced rapidly over
time. As Exhibit 1 shows, many new online platforms—a number of which are focused on
games—have entered since the App Store launched. Not only do entry barriers seem low,
but there are no obvious barriers to the expansion of output by other platforms if the App
Store were to engage in monopolistic price increases. The online platform world is not like
brick-and-mortar retailing: output expansion does not require building and staffing new
physical facilities.
120. As Android and YouTube illustrate, success can come quickly in platform businesses. When
a two-sided platform attains critical mass with sufficient numbers of participants on both
sides, the platform becomes attractive to members of both groups served, and organic
growth ensues. Emerging gaming platforms without a significant output share today such as
Stadia, Luna, or GeForce Now can still constrain the App Store because of the threat that
they can leverage indirect network effects and grow rapidly. On the other hand, when users
on one side of a platform defect, the platform becomes less attractive to the other group of
51
52
Evans Testimony, at ¶ 16, characterizes OSs as having either a “user pays” or “developer
pays” model.

users, and a rapid downward spiral can ensue. Shares of gaming platform sales can be
eroded quickly, and shares of mobile operating systems have also eroded over time.
121. Epic alleges that market imperfections such as imperfect information and switching costs
effectively lock consumers into their initial choice in the market and prevent them from
substituting away from it. Since the App Store is a two-sided transaction platform, both
sides must be considered in order to evaluate that allegation.
122. On the consumer side, one needs to take into account ownership of multiple devices on
which games are played. Even if, hypothetically, a consumer who purchased an iPhone
were to become locked in to using only iOS phones for the foreseeable future, games such
as Fortnite can be played and are played on multiple devices. Many consumers have access
to multiple devices and could acquire games and game-related digital content on other
platforms—and, in many cases, enjoy the content purchased on these other platforms even
on their iOS devices.
123. On the developer side, many game developers that use the App Store also develop for other
platforms and reach consumers through other online platforms. If Apple were to raise prices
for developers to monopolistic levels or to decrease developer support substantially,
developers could switch their efforts away from work on iOS because they can reach
potential customers through other platforms. If developers were to do this, the
attractiveness of the App Store to consumers would be decreased; they would acquire game-
related digital content elsewhere.
124. Apple must also provide app developers incentives to develop and maintain their apps for
the iOS platform, instead of competing platforms. As I noted previously, competing
platforms do not generally seek to persuade developers to permanently abandon rival
platforms. Nonetheless, Dr. Evans assumes that developers have only two options: to
continue operating in the App Store or to leave the platform.53
125. I have seen no evidence to conclude that Apple has monopoly power in the market for
digital game transactions. Moreover, based on the facts and evidence I have reviewed, I
agree with Professor Hitt’s conclusions regarding Dr. Evans’ assessment of market power.
Specifically, Professor Hitt’s conservative definition for the market for digital game
transactions shows that Apple’s share is 37.5 percent, and ranges from 23.3 percent to 27.4
percent when taking into account one-sided competition that Dr. Evans deems relevant.
Moreover, Apple’s share is likely to further erode as new platforms that facilitate digital
game transactions enter the relevant market, illustrating that barriers to entry are low.
126. In the medium to long run, the extraordinarily innovative industry that produces mobile and
gaming apps means that the needs of developers and consumers are constantly evolving.
Apple will need to continue to innovate to remain competitive. Failure to meet the evolving
needs of developers could have drastic consequences.
53

C. Implications for Assessment of Competitive Effects
127. The competitive effects of actions by or policies of a transaction platform such as the App
Store must be analyzed by considering impacts on the market as a whole, on the price and
volume of transactions—the platform’s product. It makes no economic sense to consider a
single side in isolation.
128. The ultimate test of the competitive impact of challenged conduct is its effects on the total
transaction price, taking into account charges to both sides. When it is difficult to
summarize a complex pricing strategy, the volume of transactions can be considered
instead. As Dr. Evans and I stressed in our amicus brief in the Amex case, in the case of a
transaction platform “To determine whether a constraint is anticompetitive, the presumption
at the first stage of the rule of reason should be to consider the impact on both sets of
customers, on how much they jointly pay and, ultimately, on the overall output of the jointly
consumed service.” 54
129. The analyses put forth by Professor Hitt use a reasonable methodology that accounts for the
two-sided transaction pricing, and they show rapid output growth with no evidence of
constraints from the App Store’s pricing policy.
130. Dr. Evans does not address the voluminous evidence indicating that the effect of Apple’s
policies has been to produce a healthy, thriving ecosystem that has brought enormous
benefits to developers and consumers and yielded Apple a return on its investments in
intellectual property. As noted above, despite Dr. Evans’ argument that Apple has had
monopoly power since 2010,55 App Store commission rates established two years earlier
have not increased. However, as Professor Hitt has shown, total output on the App Store in
terms of transactions and revenue from developers have increased dramatically. The
number of game transactions has grown from 0.25 billion in 2008 to 3.52 billion in 2019.
From 2010 to 2018, total revenue earned by developers has increased by more than 2,600
percent. It requires a special lens to look at the evolution of the iOS ecosystem since 2008
and see competitive problems.
VII. Competitive Analysis of a Potential “Tie” Between IAP and App Distribution
131. Dr. Evans contends that Apple conditions app distribution through the App Store on the use
of in-app payment processing for digital content for which a payment is required through
Apple’s In-App Purchase (“IAP”) mechanism, that this ties two separate products, and,
further, that by this tie Apple has restrained trade in or monopolized the market for “iOS in-
app payment processing solutions.”56
54
Brief for Amici Curiae Prof. David S. Evans and Prof. Richard Schmalensee in Support of
Respondents, State of Ohio et al. v. American Express Company et al., No. 16-1454,
Supreme Court of the United States, January 23, 2018, at p. 6.
55
56

132. Based on my review of the available evidence, and understanding of platform businesses, I
conclude that these claims lack merit. IAP is not a separate product, as required for a tie,
but rather an integral part of the App Store transaction platform that enables efficient
collection of Apple’s commission and provides procompetitive benefits to users and
developers. Epic has not shown the existence (let alone competitive harm in) a market for
payment solutions in iOS apps, nor separate demand for such a product. But even if IAP
and app distribution were separate products, they have not been tied, as developers that use
the App Store can—and do—choose forms of monetization that do not require use of IAP.
A. IAP Is an Integral Part of the App Store Transaction Platform and Provides
Benefits to Users and Developers
133. When the App Store launched in 2008, it facilitated only two basic types of transactions—
free and paid downloads. A developer wanting to offer users the ability to upgrade their app
needed to offer two versions of the same app.57 This was inconvenient as it required users
to download multiple apps and could result in loss of progress.
134. In 2009, Apple introduced in-app purchase capability allowing developers to provide
premium features that users would unlock without leaving the app. This innovation made
possible the freemium business model, which has become the second most popular on the
App Store. IAP is the name attached to the feature of the App Store that handles in-app
purchases. The App Store handles in-app purchases exactly as it handled paid downloads
before in-app purchasing was enabled: digital content delivery and (with the assistance of a
third-party payment processor) payment to Apple and the developer are assured.
135. The App Store provides transactions services involving digital content simultaneously to
both developers and consumers. Consumers make payments and receive products, and
developers receive payments and deliver products (or have Apple make delivery for them).
When the developer expects a payment, delivery of that payment is an integral part of the
transaction, and making that payment is an integral part of the transaction for the consumer
involved. Analytically, it makes little more sense to separate payment from delivery than to
treat the two sides of a transaction platform separately. If a developer wishes to earn
revenue from its digital products, app distribution is inseparable from payment.
136. From an economic perspective, IAP is an input into the production of App Store
transactions that facilitates value creation for both consumers and developers. There is
therefore serious tension between Epic’s tying claim and the correct economic teaching in
Amex: in the case of a transaction platform such as the App Store that supplies one
product—transactions—it makes no economic sense to consider inputs into transactions
production that are simultaneously engaged as actually or potentially separate markets. IAP
in the App Store is no different conceptually from, say, the merchant acceptance in the
payment card situation: merchant acceptance was an input that American Express relied
upon to produce transactions. IAP is the App Store’s centralized payment system and the
link to Apple’s secure and centralized system used to record sales, manage payments to
57
DX -3463.

developers, and collect commissions from developers that utilize the App Store. Without it,
Apple would not be able to provide customer support for any issues with in-app
transactions. Hence, IAP is an integrated part of the App Store.
137. Dr. Evans claims that “The in-app transaction between a business and a consumer using an
iOS app does not have to involve Apple . . . .”58 Nonetheless, as described below, Apple
provides a variety of App Store features that are beneficial to both developers and
consumers on an ongoing basis after an app has been initially downloaded. These include
(for developers) the ability to update apps and alert new and existing consumers of those
updates and (for consumers) the ability to redownload apps, share digital content with
family members, and manage subscriptions. Hence, contrary to Dr. Evans’ suggestion, it is
entirely reasonable for Apple to charge a commission for transactions between developers
and consumers that involve in-app purchases.
1. All Major Online Stores That Charge Sellers Commissions Require

the Use of Their Own Payment Systems to Enable Efficient Collection
of Those Commissions
138. There are a number of reasons why many, if not most, online transaction platforms have
integrated transactions handling, but one seems particularly important. When a transaction
involves a contractually agreed-upon payment by the seller to the platform, handling that
transaction internally enables the platform to collect that payment automatically. If the
transaction were instead handled by a third party that did not automatically remit the agreed-
upon payment to the platform, the platform would need to rely on the seller to keep track of
what it owes the platform and make timely and accurate payments. Presumably spot
auditing, with lawsuits when fraud is detected, would be the only available tool to enforce
honesty.
139. For the App Store, which earns commissions on transactions made between approximately 1
billion customers and thousands of developers, many of them small businesses, and earns
commissions on hundreds of millions of transactions annually, this would clearly be a much
more expensive alternative than in-house transactions management. Without automated
processes, a developer using an external payment mechanism could seek to evade a
commission owed to Apple, and Apple would have no technological ability to collect any
commissions on the sale. Apple might well elect to pass some of the increased cost back to
developers.
140. Such restrictions thus serve the same economic function as Amex’s anti-steering rules—if
Apple and other platforms could not use their own payment solutions for efficient collection
of the commissions they are owed by developers, the potential for developers to free-ride
would grow.
58

141. That most online stores require use of their own payment solutions shows that this
requirement has nothing to do with market power, but rather with the simple fact that it
allows for an efficient and seamless way for platforms to collect commissions. For
example, the App Store, Google Play, Sony’s PlayStation Store, and Steam all require in-
app purchases to go through their own payment systems59; yet these platforms operate at a
range of different scales. Indeed, four of the five digital platforms cited by Mr. Barnes60—
eBay, Etsy, MercadoLibre, and Alibaba—charge a commission and require use of their own
payment systems to collect it.61
142. In the case of Epic, every store that distributes Fortnite charges a fee on in-app purchases
made on their platform using their payment method. These stores include Samsung,62
Google Play, Xbox, Sony, and Nintendo. Mr. Kreiner, testifying on behalf of Epic,
confirmed that Sony, Microsoft, and Nintendo require Epic to use their own payment
processing systems to distribute its products on their platforms, and added that with these
platforms, Epic did not take the position that their payment processing functions were
“separate products” that Epic was required to use.
143. Dr. Evans now attempts to distinguish the App Store from all the other stores that also
require use of their own payment systems for in-app purchases of digital content by arguing
that “. . . these stores likely lack substantial market power and developers can therefore
choose whether to take the bundle of services or not.”63 He thus now agrees that this
requirement has nothing to do with market power. It is important to note, though, that this
argument does not work for the game consoles: if a developer wants to sell a Nintendo game
online, for instance, it has no choice but to go through the Nintendo eShop and to use that
store’s payment system for in-app purchases.
59
DX-3505; DX-3585; DX-4335; DX-3437; “Microtransactions (In-Game Purchases)”,
Steamworks, available at https://partner.steamgames.com/doc/features/microtransactions,
accessed on March 10, 2021.
60
Barnes Testimony, at ¶ 25.
61
“Etsy Payment Policy,” Etsy, available at https://www.etsy.com/legal/etsy-payments/,
accessed on February 24, 2021; “Transaction Services Agreement,” Alibaba, January 16,
2021, available at
https://rule.alibaba.com/rule/detail/2054.htm?spm=a271m.8038972.0.0.68df6d82RybX55;
“Form 10-K For the fiscal year ended December 31, 2019,” MercadoLibre, Inc., available
at http://investor.mercadolibre.com/static-files/5cfa8a5d-6208-4cf6-bc76-488daafb7101, at
p. 5; “Managed Payments,” eBay, available at https://pages.ebay.com/seller-center/service-
and-payments/managed-payments-on-ebay.html, accessed on March 10, 2021.
62
DX-3067, at 037; DX-4335, at 007.
63

144. Dr. Evans now argues that IAP is not an integrated part of the App Store “because it is not
provided for most apps that are distributed through the App Store.”64 It is, of course made
available to all apps; many simply chose not to use it. As just discussed, most, if not all
online stores offer similar options to developers, which suggests these options are close to
being an essential feature of online stores in general, even if not all developers make use of
them. Meat counters seem to be an essential, integrated feature of U.S. supermarkets, even
if vegetarians don’t use them.
145. Dr. Evans refers to IAP as “Apple’s one-size-fits all payment solution.”65 But given that a
core function of IAP is to efficiently collect App Store commissions, this function will
always require a one-size-fits all solution. Since collection of commissions via IAP is
automatic, if Apple were forced to cease using IAP and to use another method of collection,
it would incur higher costs. It might well pass some or all of that cost increase back to
developers. Alternatively, Apple might conclude that reliable alternative methods of
collection either do not exist or are prohibitively expensive. In this case, it would decide to
cease charging commissions on in-app sales.
146. It seems likely that this is the alternative that Epic prefers. But if Apple were forced to
make such a fundamental change in its business model, to depart from what has become an
industry standard of charging a commission for at least some in-app purchases, it would
almost certainly find it optimal to change other elements of its business model, to the benefit
of some consumers and developers and to the detriment of others. Apple’s current iOS
business model has worked well for consumers and developers, as growth on both sides of
the platform makes clear. There is certainly no guarantee that an alternative model
compelled by judicial decree would be any better.
147. Because the recording of sales and the collection of Apple’s commission are the core
functions of IAP, Dr. Evans’ claim that the requirement to use it harms competition and
innovation in payment solutions (besides being factually wrong) is largely a red herring.
148. To demonstrate, let me return to the art gallery example above. The gallery is a transaction
platform connecting artists and art lovers. When an art-lover decides to buy a painting with
a credit card, the clerk uses the gallery’s terminal to connect to the gallery’s payment
processor. If the transaction is authorized, the payment processor sends the payment for the
painting to the gallery’s account. The gallery, in turn, deducts its agreed-upon commission,
30 percent for example, and sends the remainder to the artist. But for the level of
technology involved, this is essentially how the App Store operates. Suppose now that an
artist comes with her own payment terminal in hand and asks to be allowed to use it. She
argues that it is much newer than the gallery’s method, has many great innovative features,
is more secure, provides more efficient refunds, and connects wirelessly to the artist’s
payment processor, which charges only 5 percent. The gallery would, of course, refuse this
request, since it would be essentially an attempt to avoid paying the gallery’s commission,
64
65

regardless of how innovative the artist’s terminal is. This refusal would not be suppression
of competition; it would be an insistence on collecting an agreed-upon commission.
2. There Are Other Procompetitive Benefits from Requiring Use of IAP
149. Besides providing an efficient way to implement Apple’s platform business model, the
integration of IAP provides benefits to both consumers and developers. For consumers, it
enables them to conveniently purchase and manage digital content without needing to
provide their payment details multiple times. It also ensures a safe and secure marketplace,
and supports an ecosystem that syncs content devices.
150. As Apple witnesses will testify, by creating a record of iOS purchases, IAP supports the
ability of users to redownload apps and in-app purchase on new devices, share subscriptions
and in-app features with family members, view their entire purchase history, and manage
subscriptions from one place on their phone. I also understand that IAP protects the privacy
and financial information of consumers by withholding such information from developers.
All of these are features (which Dr. Evans ignores) which a third-party processor likely
could not replicate and make the iOS ecosystem more attractive to consumers.
151. One feature supported by IAP worth spotlighting is Family Sharing. Family Sharing also
allows parents to review purchases made by their minor children—an important feature
especially for gaming apps that often go to great lengths to tempt kids into buying a new
skin or the next level of the game.
152. Apple’s IAP integration benefits developers too. Providing payment processing essentially
for free to all developers rather than requiring them to develop their own systems or to
contract with another third party and integrate that party’s in-app payment system can
reduce the barriers to entry for developers, particularly small ones.
153. Developers also benefit because, by requiring use of IAP, Apple has superior means of
payment collection. For example, Apple can block a customer who commits fraud by
repeatedly asking for refunds of consumed purchases across different apps. A third-party
developer would not have comparable means. Improving consumer experience indirectly
benefits developers; more reliable payment collection benefits them directly.
154. For developers that offer in-app purchases of digital content, IAP provides an effective
means to collect its revenues from users.66 Moreover, the direct positive effects of the
smooth functioning of IAP are complemented by indirect network effects due to the two-
sidedness of the App Store. When consumers enjoy a better customer experience,
developers indirectly benefit as well due to increasing demand of their apps, and vice versa.
In contrast, if some developers deployed third-party payment processors instead of the App
Store’s facility, their customers’ purchase experiences could be less satisfactory, as noted
above. Increased friction is likely to make the App Store as a whole less attractive to
affected consumers, which, in turn, would make it a less profitable venue for developers.
66
In fact, the introduction of in-app purchases to the App Store was at least partly motivated
by developers’ request for additional monetization options. See DX-4192 at 003.

B. There Is Not Sufficient Demand for In-App Payment Processing as Facilitated by

IAP to Support a Market in Which It Is Efficient to Offer the Two Components
Separately
155. I understand that under Supreme Court precedent, to evaluate whether a tie exists here, one
must assess whether there is sufficient demand for in-app payment processing for digital
content, as facilitated by IAP, separate from the demand for app distribution to support a
product market in which it is efficient to offer these two components separately.
156. This analysis must first consider carefully the nature of the commissions collected via IAP.
Apple’s commissions are not a charge for the bundle of services that IAP offers. They are
the price that developers pay to take advantage of the iOS platform (including licensed
intellectual property rights) and distribution via the App Store, and Apple uses these
payments from developers as part of its overall monetization strategy. While Apple has
insisted on collecting commissions that developers are contractually obligated to pay, I
understand it has never charged separately for payment processing. And to my knowledge
Apple has never tried to market IAP or its technology to other online platforms.
157. Still, if one takes Epic’s tying claims at face value and narrowly focuses on the functions
performed by IAP, the but-for world would be one where Apple continues to earn its
commissions for digital content, but developers can rely on alternative systems to perform
those functions. As noted above, however, the App Store does not charge for the fees it
pays to third-party payment processors, and developers have a contractual obligation to pay
a commission to Apple for in-app purchases. Thus, even if developers were allowed to use
third-party payment processing systems, any developer using such a system instead of IAP
would have to pay that processor’s fees in addition to the commission that it is contractually
obligated to pay Apple.67
158. In a but-for world where Apple continues to earn a commission on in-app purchases, no
rational developer would forgo payment processing via IAP offered essentially free of
charge for an alternative—inferior in a number of respects—that would charge a positive
price for payment processing. Therefore, there would not be sufficient demand for in-app
payment processing for digital content for it to be efficient to offer in-app payment
processing for digital content as a stand-alone service.
159. This is illustrated by the Epic Games Store, where the majority of games still use Epic’s
payment system even though the Epic Games Store began to allow developers to use
payment mechanisms other than Epic’s. According to Mr. Allison, the General Manager of
the Epic Games Store, using a third-party payment processing system instead of Epic’s
payment processing mechanism may actually reduce demand in the platform, because users
would be required to set up a separate account for each third-party payment processor,
67
Apple’s commission rate is 30% of the price to the user. Apple absorbs the cost of payment
processing fees and does not charge developers an additional payment processing fee. See
DX-4623, at 005; DX-3256 at 003-004.

creating “eCommerce friction.”68 Instead, Mr. Allison argued that an integrated payment
service helps facilitate the payment process for users and developers.
160. It is not clear what Dr. Evans is assuming about a world in which the use of IAP is optional.
If switching away from IAP somehow enabled developers to avoid payment of the
commissions to which their contracts obligate them, it would not be surprising that “‘no one
will ever use it’” as said the Apple executive Dr. Evans quotes in his testimony.69 Since
third-party payment providers charge 2-3 percent, while large developers would pay
Apple’s 30 percent if they continued to use IAP, it is hard to see why any large developers
would not abandon IAP if doing so would enable them to avoid Apple’s commission. Even
small developers, who owe Apple a 15 percent commission on in-app sales, would have
strong incentives to switch from IAP to third-party processors if doing so would enable
them to avoid that commission.
161. Alternatively, in a scenario in which switching to a third-party processor would not relieve
developers of their contractual obligation, it is hard to imagine that many would switch. Dr.
Evans has admitted that IAP provides a convenient payment method for both developers and
consumers.70 He further acknowledges that IAP may be particularly attractive for smaller
developers “because it would allow them to avoid the fixed costs associated with managing
a payment solution.”71 It is my understanding that Apple does not charge developers for the
payment processing fees that it pays to third parties to handle payments that go through IAP.
If switching away from IAP obliged a developer both to pay the commissions it owed Apple
(assuming Apple had some way reliably to collect) and third-party payment processing fees,
its costs would have increased.
162. Although there are indeed plenty of third-party payment processors that serve a large
market, Dr. Evans’ suggestion that they could replace IAP72 is plainly false. IAP is not just
a payment processor. Importantly, it serves to collect Apple’s commission on paid apps and
in-app payments in an efficient fashion, and Apple’s commission, like the Epic Game Store
commission, is not a fee for payment processing. If one or more developers were to replace
IAP with a third-party payment processor, Apple would no longer automatically receive the
commission payments to which it is contractually entitled. Neither Epic nor Dr. Evans has
demonstrated that there is any practical alternative method to reliably collect commissions
from developers with apps in the App Store. Nor, indeed, has either made a serious attempt
to do so.
68
See Allison Deposition, Vol.1, at 244:2-244:20.
69
70
71
72

163. In light of all this, the fact that many developers, including Match Group, Facebook, and
Spotify—and, of course, Epic itself—have tried to circumvent IAP and use their own
payment processors instead does not provide any valid evidence that there is a separate
demand by these developers for such services. It simply documents the unsurprising fact
that developers would prefer not to pay a 30 percent commission for Apple’s services and
the use of its intellectual property and would prefer to pay a lower price or none at all. Dr.
Evans alleges that on one occasion, Microsoft expressed a willingness to pay the App
Store’s commission if it could use an alternative payment processing system, but Apple
refused to allow it.73 Whatever the value of this anecdote, Apple’s ability to collect its
commission from Microsoft says nothing about its ability to collect from many thousands of
other developers without using IAP.
C. Dr. Evans’ SSNIP Analysis of an “iOS In-App Payment Solution Market” Is Fatally
Flawed
164. To support his theory that Apple has tied use of IAP to iOS app distribution, Dr. Evans
claims that “the relevant market is the market for payment solutions for accepting and
processing payments for purchases of digital content made within an iOS app.”74 To reach
this conclusion, he implements a third SSNIP test—this one just as convoluted and
uninformative as the previous two.
165. First, Dr. Evans asserts that “the average transaction fee for a developer’s own payment
solution is 5%” based on data from third-party payment processors and Epic.75 He then
calculates that assuming 20 percent of in-app transactions in iOS would use a third-party
payment processor if they could, the average commission would be 23.2 percent. He finds
that “the App Store was able to raise the average commission from the competitive level by
at least 4.5 percentage points, from 23.2% to 27.7%” durably and profitably, which is 19.4
percent higher than 23.2 percent.76 As a result, Dr. Evans concludes that “[t]he App Store,
as the hypothetical monopolist, can […] raise the price of payment solutions to the targeted
group of developers by a substantial amount.”77 That is, he claims that payment handling
for digital content purchased within an iOS app has passed the hypothetical
monopolist/SSNIP test because Apple, “the actual monopolist”78 has been able to hold price
appreciably above the competitive level.
166. Dr. Evans’ third SSNIP test relies on the assumption that the App Store commission rate is
properly comparable to the payment processing fees charged by Paypal, Braintree and
73
74
75
76
77
78

others. As Apple’s witnesses will testify, that is nonsense. The App Store provides
numerous services to attract and retain both end-users and developers, and it is a critical part
of the broader iOS platform, from which users and developers benefit. The alleged
competitive level of payment processing fees cited by Dr. Evans covers only services
related to payment processing for in-app transactions, but the commission rate in the App
Store is not a fee for payment processing; it is the price developers pay for the full array of
benefits provided by the App Store and the iOS platform.
167. Like many other app store operators, Apple relies on third-party payment processing firms
to execute payments, and it pays the charges of those firms.79 Since platforms’ commission
rates include the cost of payment processing by a third party as well as serve as part of the
monetization strategy of the store itself, the commission rates charged by app stores,
including Epic’s Game Store’s 12 percent, are generally well above the rates charged by the
payment processing firms. As Mr. Sweeney testified, Epic’s 12 percent commission “is
intended to cover all of Epic’s variable operating costs associated with selling incremental
games to customers,” including payment processing, bandwidth, hosting servers, any losses
from transactions due to payment processing issues or fraud, and customer support. And
Mr. Allison, the store’s general manager, testified that he views the 12 percent commission
as a payment for granting third-party developers the privilege of accessing Epic’s
audience.80
168. And that is not the only faulty input to Dr. Evans’ IAP SSNIP test. Here, as throughout his
analysis, Dr. Evans miscalculates the App Store’s effective commission rate to be 27.7
percent.81 What Dr. Evans actually calculates is the App Store’s commission rate across
paid transactions, ignoring all free transactions. But the vast majority of apps in the App
Store are free to download and offer no in-app purchases. As Professor Hitt has shown,
taking those into account puts Apple’s effective commission in 2019 at less than 5 percent.
169. Numerical quibbles aside, however, Dr. Evans’ market definition has a more fundamental
flaw. He advances no reason why payment processing on iOS would be a market distinct
from online payment processing in general. In fact, many payment processors that currently
handle payment processing for purchases of physical goods or services in the App Store
(e.g., Square and Braintree) also handle other transactions online. Thus even if one were to
define a market for payment processing limited to iOS, no firm, and certainly not Apple,
could exercise market power in that market because entry would be very easy.
170. Some context about relative sales volumes is instructive. The total dollars transacted
through the App Store’s U.S. storefront in 2018 was , which is small relative to
both 1) the total payment volume reported by online payment processing companies such as
Braintree (a PayPal subsidiary), Square, and Adyen and 2) the total e-commerce payment
volume in the U.S. Total dollars transacted through the App Store’s U.S. storefront in 2018
79
See, e.g., DX-3122.
80
Allison Deposition, Vol.1, at 222:12-15.
81

was at most 3 percent of the total dollars processed in the U.S. by online payment
processing companies and less than 0.2 percent of the total e-commerce volume in the U.S.
in 2018. The notion that Apple is employing a tie to achieve market power is not an
economically credible one.
D. Apple Does Not Require Use of IAP as a Condition of App Distribution Through the
App Store
171. Lastly, even if app distribution and payment solutions are separate products, there is no “tie”
here because Apple does not even require that developers monetize their app by offering in-
app purchases of digital content or require payment for any app in order to take advantage
of app distribution through the App Store.
172. Offering in-app purchases of digital content or charging for an app are only two of the many
options offered to developers to monetize their apps in the App Store. No developer is
required by Apple to use either of these monetization strategies. The distinct business
models that involve in-app purchases are the Freemium model, offered by a little less than a
third of developers, and the Paymium model, offered by only 6 percent of developers.82 In
addition, 84 percent of apps are completely free, and developers pay no commission to
Apple for them.83 Apps that offer physical goods or services do not use IAP—in fact, Apple
requires that these apps use non-IAP payment processing.
173. Dr. Evans now refers to developers who choose to use IAP as a “targeted group,” comparing
them to “a group of customers that are isolated in a hard-to-reach geographic area.”84 This
label makes no sense. The developers who have chosen to use IAP were aware of the
existence of alternative monetization strategies when they made that choice, and they can
reverse that choice at any time. The comparison to isolated communities suggests ambush
of immobile customers, but developers are not immobile, and Apple has never increased
commissions to ambush anyone.
174. Even in instances where the app developer uses a Freemium model, such as Fortnite, iOS
users can make in-app purchases on other platforms and use those upgrades on iOS without
IAP. In particular, iOS users can download Fortnite for free from the App Store and use V-
Bucks purchased on other platforms to acquire in-game content while playing Fortnite on
iOS. By definition, the use of IAP cannot be said to be tied to app distribution through the
App Store.
82
DX-3512, at 010; DX-4614.
83
DX-3055.
84
Evans Testimony at ¶¶ 248, 253.

HIGHLY
Case CONFIDENTIAL –Document
4:20-cv-05640-YGR ATTORNEYS' EYES
490-4 ONLY
Filed SUBJECT
04/27/21 TO1 THE
Page of 19
PROTECTIVE ORDER
OAKLAND DIVISION

DANIEL L. RUBINFELD
vs.
Ex. Expert

05640-YGR-TSH Daniel L. Rubinfeld
HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO2 THE
Page of 19
PROTECTIVE ORDER
1. My testimony focuses on the technical and contractual restraints challenged by Epic. The
testimony is informed by my experience analyzing the role of intellectual property (“IP”) in
antitrust matters, including my service as chief economist for the Department of Justice
during the period in the late 1990s when DOJ brought a monopoly maintenance case against
Microsoft. My overarching conclusion is that the introduction of the iPhone and its App Store
has generated substantial consumer benefits. Furthermore, the restrictions at issue in this case
facilitated the introduction of these innovative products and were—and still are—crucially
responsible for enabling the growth of the iOS ecosystem and the benefits that flow from it.
Epic’s proposed remedies would weaken or eliminate the procompetitive benefits, depriving
consumers of the valuable option to choose a platform that protects their security and privacy
and the quality of their user experience.
2. Opinion 1. The technical features that Epic challenges, including the inability to sideload
apps, were procompetitive design choices that Apple made when it launched the iPhone, and
later, App Store—well before it could even be argued that Apple had monopoly or market
power. These design choices continue to enhance product differentiation and consumer
choice. (See section III.B, p. 5.)
3. Opinion 2. Apple’s App Store policies and rules, including the challenged restrictions, are
integral to a procompetitive licensing agreement—the Developer Program License
Agreement (“DPLA”). As an economist, my view—consistent with that of the antitrust
agencies—is that once an owner of IP decides to license its IP, that owner has significant
flexibility in choosing what rights it is willing to license and under what terms. (See
section III.C.1, pp 6–8.)
4. Opinion 3. The challenged restraints are procompetitive vertical restraints because, among
other purposes, they (a) assure the safety, security, and quality of experience for users,
(b) prevent free riding, and (c) foster interbrand competition. The DPLA does not harm any
horizontal competition that would have existed but for the DPLA. (See sections III.C.1,
pp. 6–8, and III.C.2, pp. 8–10.)
5. Opinion 4. There is no economic foundation for Epic’s position that Apple has a duty to
design its products and structure its policies, rules, and agreements in a manner that benefits
other firms, including Epic. It is well recognized in competition economics that imposing
such a duty would have adverse impacts on social welfare, including the fact it would erode
incentives for investment and innovation. (See sections IV, pp. 13–15, V, p. 15, and VI,
pp. 15–16.)
6. Opinion 5. Nothing in the testimony of Dr. Evans or Professor Athey changes my opinions
summarized above. (See sections VII, pp. 16–18, and VIII, p. 18.) Both Dr. Evans and
Professor Athey fail to acknowledge the importance of Apple’s IP related to the iOS
platform. Each fails to appropriately credit the procompetitive IP-driven benefits that flow
from Apple’s developer-related restrictions. The vague but-for worlds they describe would
diminish, rather than enhance, competition. (See section VIII, p. 18.)
7. Opinion 6. There are serious economic risks associated with wrongly condemning
procompetitive behavior, particularly in technology markets. Given the high volume and

HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO3 THE
Page of 19
PROTECTIVE ORDER
rapid growth of the app economy, an erroneous condemnation of Apple’s App Store policies
could lead to substantial foregone benefits. (See section IX, pp. 18–19.)

8. I am the Robert L. Bridges Professor of Law and Professor of Economics Emeritus at the
University of California Berkeley and Professor of Law at New York University. I served as
Deputy Assistant Attorney General for Antitrust at the DOJ from June 1997 through
December 1998. In that position, I was responsible for supervising a staff of approximately
70 Ph.D. economists, financial analysts, and research assistants with respect to a wide range
of antitrust matters, including monopolization, price fixing, and other restraints of trade. As
mentioned, of particular note was my role as chief economist in the monopoly maintenance
case brought against Microsoft.
9. I received my A.B. degree in mathematics from Princeton in 1967 and my Ph.D. in
economics from M.I.T. in 1972. I have previously taught at the University of Michigan and
have been a Visiting Professor at a wide range of law schools. My current teaching
experience at NYU Law includes antitrust law and economics and quantitative methods in
law; my antitrust seminar covers, among other things, the economic analysis of IP and
antitrust.
10. I am the coauthor of two textbooks, Microeconomics and Econometric Models and Economic
Forecasts, and have published or edited eight books and over 150 articles. I have received
fellowships from the National Bureau of Economic Research, the John M. Guggenheim
Foundation, and the Center for Advanced Studies in the Behavioral Sciences. I am a past
President of the American Law and Economics Association. I am a current Fellow of the
American Academy of Arts and Sciences and a Fellow of the National Bureau of Economic
Research.
11. I have consulted and testified extensively on antitrust, intellectual property, public regulation,
and damages issues, for private parties and for the DOJ and Federal Trade Commission.
III. The Challenged Technical and Contractual Restrictions Are Procompetitive
12. In my opinion, the restraints that Epic challenges are crucially responsible for enabling the
growth of the iOS ecosystem and the benefits that flow from it. To understand why, I offer a
brief review of the historical development of the iPhone and its App Store.
13. When the App Store was first announced in March 2008, the arrangement offered to
developers was simple. The developer would choose the price to charge the user to download
the app (which could be zero). The user would pay Apple, which hosts and delivers the app
and bears responsibility for the transaction. Apple would then pay the developer 70 percent
of the purchase price and keep 30 percent for itself. When acquiring the app is free, Apple
would receive no initial compensation.
14. This business model has remained essentially in place to this day, apart from refinements
and changes that have often benefited developers. For instance in March 2009, Apple
introduced post-sale “in-app purchases.” Initially limited to paid apps, in-app purchases
eventually allowed a developer to offer a free trial version of an app that could easily be
upgraded to full functionality with an in-app purchase. This advancement also gave more

HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO4 THE
Page of 19
PROTECTIVE ORDER
flexibility to apps with a subscription model: the initial download no longer required a
positive price. Apple has also cut its commission rate—or removed the commission
requirement altogether—for certain transactions.
15. A central component of this business model is that Apple grants developers access to its IP,
in order to allow them to create native iOS apps. This has allowed even the smallest
developers to reach millions of iOS users, while iOS users got access to more and better
apps—including many free apps. Without access to Apple’s IP, the app economy could not
have flourished as it has.
A. The iPhone and App Store have generated substantial consumer benefit
16. In the 14 years since the launch of the first iPhone, the resulting iOS ecosystem has generated
significant benefits for users of iOS devices, for developers of iOS apps, and for society more
broadly through the growth of the “app economy.”
17. The iOS ecosystem has undeniably had an important economic impact. Since the App Store’s
inception, Apple has distributed more than $155 billion in earnings to developers from paid
app downloads and in-app purchases. The economic impact is even larger when considering
other revenue streams earned by developers (e.g., advertising and the sale of physical goods).
One study estimates that the App Store ecosystem generated more than $138 billion in U.S.
commerce in 2019.
18. Each new product introduction—whether that product is altogether new or differentiated to
some degree from the products previously available—expands the set of products from which
consumers can choose. Such an expansion of consumers’ choices almost always benefits
consumers. A consumer for which the new product is not superior to those already available
can ignore the existence of the new product. But a consumer that values the new product
greater than those previously available sufficiently to purchase it clearly benefits from the
availability of the new product. Economic theory and practice tell us that the introduction of
new products is procompetitive because it is beneficial to some and neutral to others.
19. The introduction of the iPhone and the App Store is no exception. Indeed, it is widely
recognized by economists that the introduction of new products such as the iPhone and the
App Store has generated substantial increases in consumer welfare.
20. But, creating a new product involves creating a new design with a distinctive combination of
features and characteristics. Innovating firms are and should be given great deference
concerning how they design new products, subject of course to health, safety, environmental,
and other regulations. This is particularly true where, as here, the design features and
challenged restraints were implemented by Apple prior to it having the market/monopoly
power it is now accused of having.
21. Creating a new product is typically expensive, requiring a substantial investment of capital,
time, and other opportunity costs. Innovating firms have an incentive to innovate wisely—to
design products that will be perceived by customers as valued and superior to other available
choices. Ultimately, the market will be the best judge of whether the firm’s innovation
decisions were wise or a failure.

HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO5 THE
Page of 19
PROTECTIVE ORDER
22. Given the inherent risk involved in innovating to develop new products, innovation could be
substantially chilled if a prospective innovator believed that her legitimate product-design
decision would later be second guessed. Economists understand the importance of
encouraging product-design innovation. With this background in mind, I will explain why
the technical and contractual restrictions that Epic has challenged are pro- rather than anti-
competitive.
B. The challenged technical restrictions are procompetitive

23. The technical features that Epic challenges—principally the inability to sideload apps—were
original design features of the iPhone, iOS, and the App Store upon their respective product
introductions. When the iPhone was introduced in early 2007, it was not designed to enable
or facilitate sideloading. This was a design decision made before Apple expected to have an
App Store offering third-party apps. Once Apple made the decision to create the App Store,
it did not change this decision—for reasons explained at the time, including protecting the
reliability and security of the device and the user. Apple also made the further decision to
preinstall the App Store iOS app on the iPhone without an option to delete.
24. The challenged technical restrictions are procompetitive. The introduction of a new product
with distinct features serves a procompetitive purpose because it expands the scope of
consumer choice. Apple’s design choice to not facilitate sideloading, i.e., to create a “walled
garden,” was made before the first iPhone was sold and before Apple sold a single app
through the App Store, supporting my view that this design choice is procompetitive.
25. This conclusion flows from the general economic principle that a practice used by a firm
without antitrust market power is likely procompetitive. The economic reasoning is that a
firm without antitrust market power, or a dangerous probability of achieving that, is likely to
be motivated by a desire to be a more effective competitor. This may harm a competitor, but
it will not harm the competitive process. Dr. Evans himself recognized this principle when
he argued that, because tying occurs in competitive markets, it “presumptively occurs
because it is efficient.”1
C. The challenged contractual restrictions are procompetitive
26. To create a native iOS app for Apple-branded products, a developer uses Apple software and
tools and other IP provided under Apple’s Developer Program to develop and test the app.
To obtain a license to develop apps that use the Apple software, the developer must enter
into the Apple Developer Program Licensing Agreement (“DPLA”). The DPLA defines the
services provided to developers and crucially enables the use of Apple software, SDKs, APIs,
and other tools for third-party participation in the iOS app ecosystem.
27. The DPLA also requires an app developer to agree to various obligations, some of which are
at issue in this case. These obligations are clearly spelled out in the DPLA. In particular, the
DPLA provides that apps developed using Apple software may be distributed only through
1
David S. Evans and Michael Salinger, “Why Do Firms Bundle and Tie? Evidence from
Competitive Markets and Implications for Tying Law,” Yale Journal on Regulation, 22(1),
2005, 37–89 at 39.
HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO6 THE
Page of 19
PROTECTIVE ORDER
the App Store and in Apple’s discretion. Epic and Dr. Evans claim this restraint is
anticompetitive.
1. The challenged restraints are integral to procompetitive licensing
arrangements
28. In my view, the challenged contractual restraints make possible the combination of
complementary assets in a pro-competitive vertical relationship between Apple and
developers.
29. It is essential to view these restrictions with reference to the opinions that have been
promulgated by our competition authorities. The DOJ and FTC Antitrust Guidelines for the
Licensing of Intellectual Property (“IP Guidelines”) reflect the antitrust enforcement policy
of the U.S. agencies “with respect to the licensing of intellectual property protected by patent,
copyright, and trade secret law, and of know-how.”2 These Guidelines reflect the application
of sound economic analysis to these policy issues. Applying the IP Guidelines framework to
the DPLA shows that the DPLA is procompetitive. Importantly, the challenged restraints are
integral to this procompetitive licensing arrangement.
30. For background purposes and as a threshold matter, the IP Guidelines acknowledge that:3
• “Intellectual property law bestows on the owners of intellectual property certain rights
to exclude others. These rights help the owners to profit from the use of their property.”
• “The antitrust laws generally do not impose liability upon a firm for a unilateral refusal
to assist its competitors, in part because doing so may undermine incentives for
investment and innovation.”
31. In my view as an antitrust economist, I believe that licensing IP is generally procompetitive,
allowing firms to combine complementary factors of production. The IP Guidelines makes
this clear, stating as a general principle that “intellectual property licensing allows firms to
combine complementary factors of production and is generally procompetitive.”4 In this
case, Apple owns the relevant IP. This linkage of the procompetitive benefits of licensing to
combining complementary factors of production is exemplified by the license Apple grants
to developers through the DPLA. Third-party developers are “complementary factors of
production” and Apple, as the owner of its IP, “finds it most efficient to contract with others
for these [complementary] factors.”5
32. In other words, by virtue of its ownership of crucial IP, Apple could have chosen to be the
exclusive developer of iOS apps for the iOS ecosystem. Indeed, that was Apple’s original
decision when it launched the iPhone. But—as Apple quickly realized—such a choice would
have failed to maximize the potential of the iOS ecosystem. Third-party developers
collectively have a broader range of ideas, exposure to the needs of diverse (sometimes niche)
2
DOJ & FTC, “Antitrust Guidelines for the Licensing of Intellectual Property,” 2017 (hereafter
“IP Guidelines”), § 1.0.
3
IP Guidelines, § 2.1
4
IP Guidelines, § 2.0.
5
HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO7 THE
Page of 19
PROTECTIVE ORDER
groups of consumers, their own IP rights, and skills and talents than Apple could ever
assemble solely within its corporate boundary. By choosing to license its IP to third-party
developers, Apple has created an “integration [that] can lead to more efficient exploitation
of the intellectual property, benefiting consumers through the reduction of costs and the
introduction of new products. Such arrangements increase the value of intellectual property
to consumers and owners.”6
33. As an additional procompetitive benefit to licensing, according to the Guidelines,
“[l]icensing can allow an innovator to capture returns from its investment in making and
developing an invention through royalty payments from those that practice its invention, thus
providing an incentive to invest in innovative efforts.”7
34. Once an IP owner—here Apple—decides to license its IP, it has great flexibility in choosing
precisely what rights it is willing to license and under what terms. The IP Guidelines
acknowledge what antitrust economists see as the procompetitive role of restraints on
licensees embodied in IP licenses, including limitations on fields of use and distribution.
Such restraints “may also increase the licensor’s incentive to license, for example, by
protecting the licensor from competition in the licensor’s own technology in a market niche
that it prefers to keep to itself.” 8
35. Thus, the IP Guidelines explicitly recognize that licensing IP need not be an “all or nothing”
proposition but instead is a matter amenable to considerable refinement and nuance.
36. The fact that an IP owner attaches conditions to a license—a ubiquitous practice blessed by
the IP Guidelines—does not suggest that somehow the IP owner is conducting some kind of
exclusionary “conditional refusal to deal” simply because the IP owner is only conditionally
willing to license IP that it is under no obligation to license in the first place.
37. Although licensing IP is “generally procompetitive,” it is not immune from antitrust scrutiny.
Indeed, when investigating Microsoft in the late 1990s, the Antitrust Division of the DOJ
found (and the Court eventually agreed) that certain OEM (original equipment manufacturer)
license arrangements were anticompetitive. According to the IP Guidelines a licensing
arrangement may raise antitrust concerns when it: “harms competition among entities that
would have been actual or potential competitors in a relevant market in the absence of the
license”—i.e., entities in a “horizontal relationship.”9 Typically, the relationship between
Apple and a developer is vertical, not horizontal, because the typical third-party app does not
compete with any specific Apple app. For developers in a purely vertical relationship with
Apple, the DPLA cannot restrain any competition that would occur in the absence of the
DPLA.
38. Of course, it is possible that a developer’s native iOS app could compete with an Apple app—
as Apple acknowledges that some third-party apps do. But, the developer must still enter into
the DPLA to develop, distribute, and sell its native iOS app. Entering the DPLA is a
6
7
8
9
HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO8 THE
Page of 19
PROTECTIVE ORDER
necessary condition for the developer’s iOS app to compete with Apple’s. Apple and such
third-party developers of native iOS apps would not “have been actual or potential
competitors in a relevant market in the absence of the license.” Put another way, Apple’s
decision to license its IP in the first place pursuant to the DPLA creates competition that
would not exist but for that licensing arrangement. Thus, for these less-typical developers
whose apps compete with Apple apps or who claim to be potential distributors of iOS apps
developed using Apple’s software, the DPLA also does not restrain any competition that
would occur in its absence.
39. This argument—that there is no competitive concern even when a developer’s app competes
with an Apple app—applies to Epic’s desire to compete with Apple with its own app-store
app distributed on the App Store, because an app-store app is itself an app.
40. I do not understand Epic to be alleging that Apple does not hold relevant IP or that Epic does
not require a license to Apple’s IP in order to develop and distribute iOS apps. Indeed, Epic
has stated that it has no right to use Apple’s services independent of the rights granted by the
DPLA. Of course, if Epic (or any other developer) did not require a license from Apple to
use Apple’s IP, or did not need to use Apple software to develop, test, and run their iOS apps,
then the terms of the DPLA would not restrain its actions.
41. In their reports submitted to this Court, both Dr. Evans and Professor Athey criticize certain
provisions of the DPLA. However, their reports fail to consider that Apple’s IP gives Apple
the right to exclude others from use of that IP absent an Apple license. Furthermore, neither
Dr. Evans nor Professor Athey has performed an antitrust analysis of Apple’s DPLA that
takes Apple’s IP into account. Neither asserts that the DPLA harmed horizontal competition
that would have existed in the absence of the license.
2. The restraints are procompetitive vertical restraints that prevent opportunism
and free riding and foster competition between platforms
42. The license terms that affect how third-party apps are distributed on iOS devices are
appropriately described as vertical restraints because Apple and the typical developer provide
complements that are combined via the iOS ecosystem. On the one hand, Apple provides the
“infrastructure”—the device, the operating system, the App Store, and many related services.
On the other, the developer provides an app with features and characteristics that work in
tandem with the other components of Apple’s ecosystem.
43. Vertical restraints are common in contractual arrangements without regard to whether parties
have market power. This by itself indicates that vertical restraints often serve economic
efficiency and often are procompetitive. Although the term “vertical restraint” emphasizes
that the provision restrains a party from doing something it might otherwise do if not so
restrained, the crucial flip side is that the contractual relationship thereby enables a mutually
beneficial arrangement—one where the party that voluntarily accepts the restraint is made
better off by the arrangement than it would be without the arrangement and the restraint. For
example, a restraint that requires downstream entities to maintain high quality standards is
costly to each downstream entity. However, the fact that all other downstream entities
maintain high quality could increase the value of the platform so much that all downstream

HIGHLY
490-4 ONLY
Filed SUBJECT
04/27/21 TO9 THE
Page of 19
PROTECTIVE ORDER
entities are benefitted even net of each downstream entity’s costs of compliance with the
restraint.
44. A frequent goal of a vertical restraint is to inhibit what economists refer to as opportunistic
behavior, by which one party would otherwise—but for the vertical restraint—take actions
that are privately beneficial to the performer but harmful on net to the joint enterprise—so
much so that allowing the opportunistic behavior is inefficient because it reduces the joint
surplus.
45. One common and well-studied form of opportunistic behavior is “free riding.” A canonical
example is a pair of retailers, one with a high level of service to consumers and the other
with a low level of service. Customers could take advantage of information available from
the high-service retailer to determine what brand of equipment to purchase, but consummate
the purchase at the lower price available at the low-service retailer. The low-service retailer
would thus free ride on the services of the high-service retailer by benefitting from them
without paying for them. The prevention of free riding is widely recognized by economists
and courts as a procompetitive justification for vertical restraints.
46. The vertical restraints challenged by Epic serve that very purpose. Suppose that Apple could
no longer enforce its policies that all native iOS apps written using Apple-licensed software
and tools be available only through the App Store. In this but-for world, there would be an
incentive for a rogue developer to create and distribute outside of the App Store an alternative
app store that exercised little or no oversight, permitting the distribution and installation of
apps that are buggy, are insecure, install malware, harvest users’ private data, cause excessive
battery drain, hog bandwidth, or any combination of these. This rogue app store likely would
exercise too little oversight for several reasons. For one thing, oversight is costly to perform,
and a rogue app store would not internalize in its private incentives the broad array of benefit
and costs across the entire iOS ecosystem, as Apple rationally does. For another, the rogue
app store would lack the specialized expertise Apple has to exercise such oversight
effectively. These rogue app stores are analogous to the low-service retailer I just described.
47. Users that installed the rogue app, and were harmed or otherwise had a poor user experience,
could not necessarily trace their displeasure back to the rogue app specifically, or to the rogue
app store, but may attribute the problems more generally to Apple. Going forward, these
users would attach a lower level of expected quality and safety when, among other things,
considering whether to install new apps from the App Store, whether to invest in new iOS
devices or even whether to stay within iOS at all.
48. There are two senses in which this rogue developer would be free riding—both occurring
simultaneously and having the same adverse effect on Apple as well as other developers:
• The rogue developer would be free-riding on Apple’s innovation and investments that
created, continually improves, and maintains its iOS devices, its tools for the iOS
platform, the iOS operating system, and the App Store (along with its policies and rules)
that creates a large installed base of iOS devices and users and generate a high demand
from iOS users for iOS apps.

HIGHLY
Case CONFIDENTIAL Document
4:20-cv-05640-YGR – ATTORNEYS'
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
10THE
of 19
PROTECTIVE ORDER
• The rogue developer would be free-riding on other developers that create safe, secure,
high quality apps and otherwise conform with developers’ responsibilities under the
App Store policies and rules, because the costly actions of conforming developers also
help establish and maintain the valuable reputation of the iOS ecosystem.
49. The effects of the rogue developer’s behavior include:
• The rogue developer would receive as a private benefit the incremental lift in demand
for its app that would flow from the large installed base and high reputation of the iOS
ecosystem.
• The reputation of the entire iOS ecosystem, and of all the apps distributed on it, would
be harmed by the bad experiences by users of the rogue app. This harms the value of the
iOS ecosystem to Apple, all developers, and users.
50. Alternatively, suppose the rogue app would have passed App Store review but was instead
distributed outside the App Store. In that case, the rogue app free rides on the large installed
base and high reputation of the iOS ecosystem caused both by Apple’s innovations and the
costly actions of conforming developers. The rogue developer gains the private benefits of
that free riding but deprives Apple of the commission it charges unless Apple is permitted to
charge a commission on transactions that occur outside of the App Store.
51. Free riding is of particular concern when the agreements at issue involve intellectual rather
than contractual property rights. The problem is that IP rights can be appropriated relatively
easily. And, if the innovator cannot effectively exclude others from copying or otherwise
making use of the invention for free, some of the returns from the innovation will be lost,
with the expectation that there will be less innovation.
52. Apple’s in-app purchase mechanism (“IAP”) also prevents developers from free-riding on
Apple’s IP. Developers bypassing IAP would in effect exploit Apple’s historical innovations
and investments while avoiding paying the remuneration that an IP holder is entitled to
collect. This would chill Apple’s incentive to make similar investments and innovation in
the future—to the detriment of consumers and developers.
53. Vertical restraints tend to foster interbrand competition, which in the present context might
be more clearly named inter-platform competition. Promoting interbrand competition is a
primary concern of antitrust economists such as myself.
D. The challenged technical and contractual restrictions work together to serve
procompetitive goals by protecting iPhone users’ security, privacy, and quality of
experience
54. The challenged technical restrictions are complementary to—and work together with—the
challenged contractual restrictions to protect iPhone users’ security, privacy, and quality of
experience. First, the challenged contractual restrictions assure the safety and quality of the
apps available from the App Store. Second—and relatedly—the challenged technical
restrictions inhibit developers from circumventing the protections of the App Store.

HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
11THE
of 19
PROTECTIVE ORDER
55. The app-review process and the App Store Review Guidelines benefit Apple, iPhone users,
and developers. They benefit users by ensuring that users have a good experience with any
apps they download from the Store. And they benefit developers because each developer
benefits from the review of all other developers’ apps. By improving the overall quality of
available apps, the App Store performs a certification function. It assures users that any
particular developer’s app will likely be compatible with the user’s device and version of the
operating system and likely will not be hazardous, be unpleasant, harm the performance of
the user’s device, or create a threat to the user’s privacy. This App Store review and approval
process makes users more likely to purchase the developer’s app.
56. This instills trust: iOS users know that when they transact on the App Store, they are using a
reliable and secure platform and will receive a product that has been vetted to meet Apple’s
high standards. This consumer confidence in turn enriches the App Store ecosystem (and
developers) as they are more prone to download, use, and pay for developers’ apps.
57. A corollary of this principle is the converse: Any developer that would be allowed to
circumvent Apple’s app review process potentially harms all other developers (and therefore
Apple itself) because users’ confidence in all iOS apps can be undermined by bad
experiences with apps from the circumventing developer.
58. Lastly, the app-review process and the App Store Review Guidelines benefit Apple by
fostering positive perceptions of the iOS ecosystem that likely result in greater sales of
iPhones and iPads as well as related services.
59. The safety, security, privacy, and usability delivered by the App Store—including as a result
of its policies and rules—provide consumers a valuable, differentiated option from the
Android platform. This drives competition between iOS and Android platforms.
E. Dr. Evans wrongly believes that Apple’s restrictions and policies are unnecessary
and not procompetitive
60. I understand Dr. Evans may opine that Apple’s restrictions and policies regarding the App
Store are unnecessary and not procompetitive. Such opinions would conflict with his prior
writings.
61. For example, Dr. Evans previously acknowledged the need for digital platforms to “have
platform rules that prohibit or require certain behavior” to deter participants “from
[b]ehaving [b]adly” in ways that reduce the value of the platform to its members. In
particular, Dr. Evans has stated that such rules allow the platform to “impose penalties for
breaking the rules as well as screening methods for keeping bad actors off the platform.”10
62. Indeed, less than a year before he filed his report, Dr. Evans praised Apple’s rules governing
distribution of iOS apps and the app-review process in an article he wrote, stating:11
10
David S. Evans, “Vertical Restraints in a Digital World,” Competition Policy International,
December 2020, at 5–6.
11
Evans, “Vertical Restraints in a Digital World,” at 13.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
12THE
of 19
PROTECTIVE ORDER
There isn’t much controversy that Apple’s rules have enabled it to create a high-
quality app ecosystem for the iPhone.
63. Dr. Evans acknowledges that Apple policies and rules he criticizes were design decisions
integral to the iPhone and iOS platform.12 But, he says, Apple “did not have to block all
channels of distribution, other than its own to create value for iPhone users and app
developers.”13
64. Dr. Evans is wrong. An alternative iOS app store would allow iOS users to download and
install native iOS apps that were not subject to Apple’s app review and therefore did not pass
Apple’s criteria that exist to ensure users’ safety, privacy, and quality of experience on iOS.
In other words, an alternative iOS app store would allow developers to circumvent Apple’s
procompetitive role in assuring safe, high-quality apps could be installed on users’ devices.
Alternative app stores would be likely to exercise too little oversight for several reasons,
including that such oversight is costly to perform while only Apple has the incentives to
internalize the broad array of benefit and costs, across the entire iOS ecosystem, to screen
apps with sufficient vigilance. Indeed, Apple would possibly need to resolve various disputes
with third-party app stores and may even need to revoke their licenses and engage in litigation
(as it is doing with Epic). These would be additional costs imposed on Apple.
65. Apple’s control over what types of apps can and cannot be installed on users’ iOS devices
contributes to the value of the iOS ecosystem to consumers and hence is procompetitive.
With respect to this quality concern, I have found the testimony of Dr. Aviel Rubin to be
informative. Dr. Rubin explains that purveyors of malware and pirated applications take
various measures to evade Apple’s app review. For example, if an app is rejected, the
developer may try to submit that app in another language, or modify some behavioral aspect
of the app. An alternative iOS app store would create a whole new outlet for the distribution
of rejected apps. It is the design of the iPhone and the App Store, and the licensing terms of
the DPLA, and Apple’s enforcement of its policies and rules thereunder that create the benefit
of a high quality, secure, safe, and privacy respecting experience for users. The alleged
restrictions are procompetitive because they are integral to the delivery of the procompetitive
benefits to users and developers.
66. That similar policies have been adopted on several other “closed” platforms further suggests
that those policies are procompetitive. As the Court previously noted, Sony, Nintendo and
Microsoft all operate similar walled gardens or closed platform models as Apple, whereby
the hardware, operating system, digital marketplace, and payment system are controlled by
the platform owner.
67. Dr. Evans also errs in claiming that, even absent Apple’s App Store rules and policies, there
would be no free riding on Apple’s innovation because Apple benefits from having third-
party apps on its platform by selling more iPhones. Dr. Evans appears to assert that the
unquantified contribution of third-party apps to Apple’s iPhone sales is necessarily sufficient
compensation to Apple to provide Apple with optimal incentives for investment. But the fact
12
Evans Direct Testimony ¶¶ 94-101.
13
Evans Direct Testimony ¶ 299.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
13THE
of 19
PROTECTIVE ORDER
that two parties jointly contribute to the creation of a surplus does not imply that their
individual contributions are necessarily equal and offsetting.
68. As the owner of its IP, it is Apple’s decision how to create incentives for developer
engagement that optimize the value of the platform. Apple could have chosen to call the
developers’ contribution to the platform sufficient and charge no commission to anyone; it
did not. Apple could have decided that developers’ contributions to the value of the platform
were so great that Apple would pay developers to create apps; it did not. Apple could have
chosen to assess a commission on a broader range of apps; it did not. Instead, Apple chose
to set a commission structure —which has been in place largely the same since the App Store
was founded, with the exception of changes that reduced commission rates in some cases. It
was Apple’s choice to have made and Apple’s choice to continue to refine. Apple has the
greatest stake in the consequences of that choice and the greatest incentive to get it right.
69. Dr. Evans also suggests that Apple is obligated to allow distribution of apps on the iPhone
in all the same ways that apps can be distributed on the Mac or Windows, platforms that
launched decades earlier on very different devices. Although Dr. Evans describes how other
firms have made different business model choices in different circumstances than Apple, the
choice of business model is an important dimension of competition and there should be no
presumption that the old ways are the best ways or that a firm needs to justify choosing
something new. The fact that Apple’s choice to create and adopt a new business model (the
App Store) has been so successful illustrates the value of business model innovation.
70. In any case, the way third-party apps have been distributed on the Mac is not a benchmark
for how Apple should allow third-party apps to be distributed on iOS. There are significant
differences between the macOS and iOS platforms, including their history (Mac launched in
1984, the iPhone in 2007), how the devices are used, the range of sophistication of the users,
the sensitivity of the information carried on them coupled with the relative vulnerability to
theft and loss, and the number of apps users download. It would be chilling if Apple were
prohibited from learning, adapting, and changing in response to two decades of experience,
to the evolution of computing, cybersecurity, and telecommunications since that time, and to
the different ways in which users utilize different devices.
IV. There Is No Economic Basis to Impose a Duty upon Apple to Redesign the iPhone
and App Store
71. An inherent premise of Epic’s challenges to the alleged technical and contractual restrictions
is that Apple has a duty to design its products and to structure its policies, rules and
agreements in a manner that benefits other firms, including Apple’s competitors (assuming
Epic to be a competitor) and that such conduct can be considered anticompetitive rather than
beneficial to competition. In my opinion, imposing a duty to redesign the iOS ecosystem
does not make economic sense.
72. The economic literature about a duty to deal warns about the dangers of imposing on one
firm a duty to aid its competitors. For example, Professor Dennis Carlton (a former Deputy
Assistant Attorney General for Economics at DOJ) highlights the fundamental economic
tension between the goals of antitrust and a duty to deal: “[A]ntitrust laws are premised on
the simple notion that rivalry among firms benefits consumers, yet a doctrine of a duty to

HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
14THE
of 19
PROTECTIVE ORDER
deal clearly limits that rivalry.” Because, Professor Carlton explains, it is difficult to
distinguish exclusionary behavior from beneficial competition (as both can involve one
firm’s sales increasing at the expense of another’s), “[o]veraggressive use of antitrust claims
against exclusionary conduct” will “inevitably reduce the incentive for beneficial
competition.”14
73. From my perspective the but-for worlds of both Dr. Evans and Professor Athey in essence
argue for imposing a duty to deal upon Apple because each would impose a duty upon Apple
to admit developers, third-party app stores, and third-party payment processors into the App
Store, the iOS platform, or both, on terms not chosen by Apple. Dr. Evans acknowledges that
this case involves “Apple’s decision to make the App Store the only way to distribute apps.”
The duty Epic and its experts would impose upon Apple is more than the usual duty to deal;
it would include a duty to redesign the iPhone—both the hardware and software of which are
covered by Apple’s IP—to make the iPhone interoperable with alternative app stores and
with apps that would not qualify under Apple’s app-review guidelines for distribution
through the App Store.
74. It is well recognized by competition economists (often in reacting to antitrust cases) that
imposing a duty to aid one’s competitors would have adverse impacts on social welfare. For
one thing, it would erode ex ante incentives for investment and innovation. For another, it
could compel cooperation between rivals (which could shade into collusion) and it could
entail ongoing oversight and regulation by courts.
75. From an economic perspective, second-guessing a firm’s product design should be viewed
with suspicion, even if it happens to render rival products (or services) incompatible.
Requiring the innovator to alter the product design—for example, for interoperability—risks
imposing substantial costs on existing customers and is likely to discourage future design
innovation.
76. These are all concerns that arise from Epic’s claims here. Imposing a requirement that Apple
allow software developers that use Apple software to distribute native iOS apps outside the
App Store would reduce Apple’s incentive to invest in developing its App Store. As
explained, that would allow free riding and, more broadly, would reduce Apple’s incentive
to engage in innovation-driven research and development. Imposing a duty to deal on Apple
would also leave open the possibility of costly disputes over the exact terms of the dealing
requirement. In the end, such a duty would likely require some form of costly continuing
oversight.
77. It is appropriate for an innovator, particularly one that seeks to create an entirely new device
(the iPhone) and platform (the App Store), to make the design decisions it thinks best for the
success of that new device and platform. Any distortion in those design decisions
necessitated by a duty to deal would represent a lost benefit for Apple (or Apple would have
chosen such a path in the first instance).
14
Dennis W. Carlton, “A General Analysis of Exclusionary Conduct and Refusal to Deal—Why
Aspen and Kodak Are Misguided,” Antitrust Law Journal, 68(3), February 2001, 659–83,
at 659–60.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
15THE
of 19
PROTECTIVE ORDER
78. Thus, imposing such a duty on Apple to redesign its successful iPhone and App Store would
both (a) risk harming current and future iPhone users by virtue of having a less safe and
secure iOS ecosystem and (b) harm incentives for innovating firms that are contemplating
the design of new products.
79. I offer no legal opinions here, but as a competition economist I conclude that Apple’s
decisions not to design the iPhone and iOS with a sideloading option and not to license
developers who use Apple software to bypass the App Store are not a reversal by Apple of
some conduct in which Apple had earlier voluntarily and profitably engaged.
V. There Is No Economic Basis to Impose Compulsory Licensing on Apple
80. A particular consequence of both Dr. Evans’s and Professor Athey’s but-for worlds is that
Apple would be compelled to license its IP in ways it otherwise would not. From an
economics perspective, however, imposing duties on Apple to design its innovations and to
license its IP in specified ways is anticompetitive.
81. Dr. Evans’s prior writings on this subject are informative. Dr. Evans has opined that it is
“close to impossible” to accurately balance the welfare-increasing and welfare-decreasing
effects of compulsory licensing, such that “[i]n general, in the absence of any positive effects
on innovation, compulsory licensing is likely to have an overall negative impact on
welfare.”15
82. And Dr. Evans is not alone in fearing that compulsory licensing would reduce economic
welfare. Two prominent antitrust economists, Professors Richard Gilbert and Carl Shapiro—
who, like me, are past Deputy Assistant Attorneys General for Antitrust at the DOJ—have
explained that compulsory licensing “can also reduce welfare in the long run by reducing
incentives for innovation.” They describe the long-run effects of an obligation to deal as
“profound adverse incentives for investment and for the creation of intellectual property”
and that this “is reason enough to justify skepticism toward policies that call for compulsory
licensing.”16
83. Finally, from my perspective, Professor Athey’s discussion of counterfactual alternative
distribution structures is no more than an academic exercise because she fails to acknowledge
that Apple’s IP gives Apple control over how apps created with Apple-licensed software may
be distributed.
VI. The Claim that Apple Controls an “Essential Facility” Would Raise Similar
Concerns to Duty to Deal and Compulsory Licensing
84. It is unclear whether either Dr. Evans or Professor Athey believe that Apple is denying Epic
access to an “essential facility,” as Epic asserts in its Complaint. Dr. Evans does not mention
15
Christian Ahlborn, David S. Evans, and A. Jorge Padilla, “The Logic & Limits of the
Exceptional Circumstances Test in Magill and IMS Health,” Fordham International Law
Journal, 28(4), 2004, 1109–1156, at 1137–1138.
16
Richard J. Gilbert and Carl Shapiro, “An economic analysis of unilateral refusals to license
intellectual property,” Proceedings of the National Academy of Sciences, 93(23),
November 12, 1996, 12749–55, at 12753–54.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
16THE
of 19
PROTECTIVE ORDER
“essential facility” in his testimony and appears not to adopt this claim. Dr. Evans also does
not claim that Apple has a monopoly over mobile operating systems or mobile devices, which
I understand to be the alleged essential facility, according to Epic.
85. Just to be clear, iOS is the operating system for Apple’s iOS devices and is subject to IP
protection. If the essential facility doctrine is governed by the same principles that govern
the creation of duties to aid competitors, its application here would be subject to the same
concerns and economic objections I have identified above. And if the doctrine goes beyond
those principles, from the perspective of competition economics, the application of the
concept of an essential facility would be at best suggestive of the imposition of a regulatory
regime. It would be inappropriate to apply such an “‘essential facility’ doctrine” to IP.
86. An antitrust law and economics commentator, Herbert Hovenkamp, has described the so-
called essential facility doctrine as “one of the most troublesome, incoherent and
unmanageable of bases for Sherman § 2 liability.”17 Professor Hovenkamp concludes that
“the essential facility doctrine discourages competitive investment and is best left to
regulatory policy rather than antitrust law.”18
87. In this regard, it is useful to recollect that despite Microsoft’s monopoly over its Intel-based
operating system (“OS”) market, in its 1998 filing, the DOJ did not claim that the Microsoft
OS was an essential facility. Since then, observers such as Gregory Sidak and Abbott Lipsky
have concluded that it would have been inappropriate to characterize the Microsoft OS as an
essential facility. The essential facilities doctrine can be seen as an equivalent to the
economic concept of a natural monopoly. The implication is that managing the facility would
involve administrative complexity, while attempts to restrain Microsoft through regulation
would likely be ineffective.
VII. Professor Athey’s Analysis Does Not Advance the Economic Analysis of This Case
88. Professor Athey posits an alternative world with one or more “Multi-platform App Stores”
that she asserts is advantageous because of allegedly lower switching costs for consumers
and developers.
89. Indeed, Professor Athey claims her Multi-Platform App Stores would lead to an
improvement in welfare for users, but she does not acknowledge or analyze the myriad of
other consequences that would flow from the assumed change in Apple’s policies and rules
concerning the App Store. For example, Professor Athey does not refer in her written
testimony to either security or privacy. She also ignores the role of Apple’s IP.
90. In this section, I explain why Professor Athey’s focus on reducing switching costs—while
ignoring Apple’s IP rights and the procompetitive benefits of Apple’s design of the iPhone
and App Store policies—should be given little, if any, weight.
17
Herbert Hovenkamp, Federal Antitrust Policy, The Law of Competition and Its Practice (6th
Edition), 2020 (hereafter “Federal Antitrust Policy”), at 401.
18
Federal Antitrust Policy, at 382–383.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
17THE
of 19
PROTECTIVE ORDER
A. Switching costs are consistent with, can intensify, and can be manifestations of
competition
91. Although Professor Athey acknowledges the existence of a substantial economic literature
regarding switching costs, she did not rely upon that literature when forming her opinions in
this case and ignores that literature in claiming—without basis—that lower switching costs
necessarily result in more competition.
92. In fact, switching costs can intensify competition—even when switching costs are
significant. The intuition is that a customer, once acquired, is particularly valuable when
there are switching costs. This causes firms to compete more vigorously to acquire new
customers by making more-enticing offers. To give an example: Not long ago, mobile-phone
customers commonly signed a two-year contract with a carrier. This created a switching cost
if the consumer wanted to switch carriers before the contract term expired. Mobile carriers
intensely competed to win customers, for example, offering free or heavily subsidized
phones. They did this precisely because a customer with a signed contract, and therefore
significant switching costs, was a particularly valuable convert.
93. Professor Athey has performed no analysis to show that the switching costs are sufficiently
high that their reduction would increase competition; indeed, she has not acknowledged that
this question is critical to her opinion. The impact of switching costs on equilibrium prices
is an empirical question, but Dr. Athey has not conducted that empirical analysis.
B. Professor Athey fails to show that claimed reductions in developer and other
switching costs would increase competition between existing platforms or lead to
the emergence of new ones
94. Professor Athey also fails to show that the decrease in switching costs she posits would result
from the existence of more “multiplatform app stores” would increase competition between
existing or new platforms.
95. The one concrete developer cost Professor Athey identifies as a “friction” is the fixed cost
required “invest in the infrastructure to offer account management that would recognize a
given user across platforms.”19 Professor Athey offers no quantification of such costs and
fails to show they are large or material to a developer’s decision as to whether to launch an
app on a platform.
96. The reality suggests they are not. If these cross-platform developer costs were as significant
to developers’ decisions on what platforms to launch their apps as Professor Athey seems to
suggest, one would expect to see many important apps developed for only iOS or only
Android for that reason. This is not the case.
97. Professor Athey also does not explain how a multi-platform app store would materially
decrease cross-platform developer costs. The costs of developing across platforms arise
largely from different platforms having different operating systems; software for those
platforms must be developed using different Software Development Kits and perhaps a
different programming language. A multi-platform app store that permits a user to buy either
19
Athey written direct, ¶ 40.
HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
18THE
of 19
PROTECTIVE ORDER
an iOS or PC version of an app does not obviate the need for the developer of that software
to create an iOS and PC version of that software.
VIII. The But-For Worlds of Dr. Evans and Professor Athey Would Harm Competition
between iOS and Other Platforms by Reducing Product Differentiation That Drives
Competition
98. Regardless of the particular relevant market found to be appropriate for analyzing the
particular issues in this case, consumers benefit from the rivalry between iOS and Android
mobile operating systems. From the perspective of my analysis, perhaps the most
fundamental flaw in Professor Athey’s analysis—and Dr. Evans’s, for that matter—is their
failure to recognize that it is the differentiation between the iOS and the Android platforms
that drives the competition between them. The choices thereby offered to consumers are
extremely valuable.
99. A key differentiator between these two platforms is Apple’s approach, via the App Store and
its policies and rules, to create a safe, secure, privacy-respecting, and highly usable
experience for iOS users. Only Apple internalizes the health of the entire iOS platform,
giving Apple uniquely strong incentives to manage the platform to maximize its value.
100. Professor Athey sees the differentiations between iOS and Android not as a source of value
to consumers, but rather as creating switching costs that should be minimized. She proposes
to reduce switching costs by removing Apple’s power to protect the security, safety, privacy,
and quality of experience of iOS users through its curation of the iOS apps that can be
installed on the iPhone. Professor Athey’s proposal to remove this key differentiator would
reduce product differentiation between iOS and other platforms, and that product
differentiation is itself a manifestation of the intense product-design competition between
platforms. Consumers would be deprived of being able to choose a platform in which they
are so protected. Thus, Professor Athey’s proposal would reduce existing competition in the
speculative hope of increasing competition in the future.
101. Similarly, Dr. Evans’ proposals to allow sideloading and alternative third-party app stores—
all of which are features of the Android platform—would also reduce the product
differentiation between the iOS ecosystem and other platforms, thus reducing the
competition between iOS and other platforms on a key dimension.
102. Consumers have the choice to select a mobile smartphone platform that takes a different
approach than Apple. Many consumers exercise that option; many others choose Apple’s
platform. Apple’s policies and rules for its App Store do nothing to dampen the viability of
the alternative Android platform or deprive consumers of their option to choose that
alternative platform. The but-for worlds of Dr. Evans and Professor Athey, on the other hand,
would deprive all consumers of the option of choosing the platform where Apple takes
responsibility for the safety, security, and privacy preservation of users.
IX. There Are Serious Risks from Wrongly Condemning Procompetitive Behavior,
Particularly in Technology Markets
103. I would be the first to admit that competition law and economics is an inherently difficult
subject, especially when it requires predictions about likely future behavior. Antitrust law

HIGHLY
490-4EYES
FiledONLY SUBJECT
04/27/21 PageTO
19THE
of 19
PROTECTIVE ORDER
and policy often require predicting the implications of several alternative courses of actions,
which can be prone to error for reasons including the lack of good data about the past and
present and the absence of a reliable framework for evaluating the causal impact of legal
rulings or policies. The latter is a particular risk when complicated conduct in technology
markets is being evaluated.
104. There are serious economic risks from wrongly condemning procompetitive behavior,
particularly in technology markets. Innovative business models can often raise antitrust
concerns because (1) they are less well understood than long-standing business practices in
fully mature industries, (2) we have less experience with them, there is a shorter track record
of data about how they operate, and (3) their evaluation may require modes of economic
analysis that are comparably novel and thus less mature and settled.
105. As a result, the harm caused by an erroneous condemnation of procompetitive behavior will,
in many instances, be greater than the injury that may result from failing to condemn harmful
behavior. This risk is especially concerning in dynamic technology markets, where erroneous
condemnations can have long-lasting effects.
106. Apple's App Store policies and rules now have a 12-year track record of success and positive
impact; there is little uncertainty about the general direction of their trajectory. However,
there would necessarily be substantial uncertainty about the effects o f Epic's desired
injunctive relief. Given the high volumes and rapid growth of the app economy, an erroneous
condemnation of Apple's App Store policies, regardless of the extent of the impact, could
lead to significant foregone benefits.
107. Indeed, condemning Apple's platform policies could have a ripple effect and lead to
condemnation of common and corresponding policies on other platforms, thereby amplifying
the economic impact of a condemnation here. In addition, setting an erroneous precedent for
the evaluation of platforms generally could chill innovation and investment in new platforms
whose policies may also be called into question.
108. The success of the iOS ecosystem shows clearly that it is working. Epic nevertheless is asking
the court to fix something that is not broken. Granting Epic's request would be risky for the
entire app economy and the wider economy more generally.
X. O a t h
I declare under penalty of perjury under the laws of the United States that the foregoing is true
and correct.
Respectfully submitted.
Daniel L. Rubinfeld, Ph.D.

April 23, 2021
Word count: 9168.
Epic v. Apple, No. 4:20-CV- 1 9 W r i t t e n Direct Testimony of

05640-YGR-TSH D a n i e l L. Rubinfeld
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO1 THE
Page of 30
PROTECTIVE ORDER
OAKLAND DIVISION

DOMINIQUE HANSSENS, PH.D.
vs.
APPLE INC., Trial Date: May 3, 2021

Time: 8:00 a.m.
Defendant, Counterclaimant.
Judge: Hon. Yvonne Gonzalez Rogers Ex.
Ex. Expert 10
Epic v. Apple, No. 4:20-CV- Written Direct Testimony of

05640-YGR-TSH Dominique Hanssens, Ph.D.
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO2 THE
Page of 30
PROTECTIVE ORDER
1. Device Usage Opinions: I conducted two surveys following standard survey research
practices. In both of my surveys, the vast majority of respondents—who were all iPhone and
iPad users in the U.S.—regularly used or could have regularly used at least one of the following
types of devices in the last 12 months: smartphones and tablets not manufactured by Apple,
desktops and laptops manufactured by Apple and others, gaming consoles, and handheld gaming
devices. These are all devices on which a user can access digital gaming content. ¶¶ 5–20.
a. Opinion 1. Results of my first survey show that 92 percent of respondents who
downloaded apps from the App Store had regularly used at least one other type of
device (i.e., devices other than iPhones and iPads) with which they could access
digital gaming content, in the last 12 months. Further, 99 percent of respondents
in the first survey had regularly used or could have regularly used at least one
other type of device (i.e., devices other than iPhones and iPads) with which they
could access digital gaming content, in the last 12 months. ¶ 16
b. Opinion 2. Similarly, results of my second survey show that 97 percent of
respondents who played Fortnite on their iPhones and iPads had regularly used at
least one other type of device (i.e., devices other than iPhones and iPads) with
which they could access digital gaming content, in the last 12 months. Further, 99
percent of respondents in the second survey had regularly used or could have
regularly used at least at least one other type of device (i.e., devices other than
iPhones and iPads) with which they could access digital gaming content, in the
last 12 months. ¶ 17
c. Opinion 3. The vast majority of respondents who played Fortnite on their
iPhones and/or iPads (94 percent) played digital games on at least one other type
of device (i.e., devices other than iPhones and iPads), in the last 12 months. ¶ 17.
2. Rossi Survey/Results Opinions: Professor Rossi conducted a survey of U.S. consumers
to measure how they would respond to a hypothetical five percent price increase to “in-app
purchases and subscriptions when purchased from within iOS apps.” 1 In my review of Professor
Rossi’s survey, I found that he failed to follow established practices for survey research, and his
questionnaire design choices introduced known sources of response bias. The results of his
survey are therefore biased and unreliable, and analyses (such as Professor Rossi’s price
elasticity calculations) that rely on these biased data are themselves biased and unreliable.
¶¶ 21–60. Specifically, Professor Rossi:
a. Opinion 4. Excluded from his survey an important segment of digital game
consumers (those aged 13 to 16), and an important segment of in-app purchase and
subscription consumers (those who had made in-app purchases or bought
subscriptions outside the 30-day window prior to the survey). ¶ 29.
1 Written Direct Testimony of Peter E. Rossi, Ph.D. (“Rossi”), ¶ 2.

Page 1
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO3 THE
Page of 30
PROTECTIVE ORDER
b. Opinion 5. Created a hypothetical price increase scenario that was confusing and
likely biased respondents toward keeping the same purchases under the hypothetical
price increase. ¶ 37.
c. Opinion 6. Made multiple, significant revisions to resolve respondent confusion
about the hypothetical price increase scenario, but did not interview a single survey
taker to determine if those sources of confusion were actually resolved (i.e., he did
not pretest his final survey instrument). ¶ 34.
d. Opinion 7. Failed to provide a clear and complete set of choices respondents could
make to shift payments from the App Store, and likely understated the proportion of
real-world consumers who would switch to paying for in-app purchases and
subscriptions through means other than the App Store to reduce spending. ¶ 54.
3. I am a Distinguished Research Professor of Marketing at the UCLA Anderson School of
Management. My research focuses on strategic marketing problems, to which I apply expertise
in data-analytical methods, such as surveys, econometrics, and time-series analysis. From 2005
to 2007, I served as the Executive Director of the Marketing Science Institute in Cambridge,
Massachusetts. The American Marketing Association awarded me the Churchill Award (2007)
and the Mahajan Award (2013) for Career Contributions to Marketing Research and Marketing
Strategy, respectively. The INFORMS Society for Marketing Science elected me as a Fellow
(2010) and awarded me the Buck Weaver Award (2015) for lifetime contributions to the theory
and practice of marketing.
4. My assignments and opinions in this matter were described in an opening expert report
dated February 16, 2021, and a rebuttal report, dated March 15, 2021. I designed and conducted
two device usage surveys. I also reviewed and evaluated the reliability of the survey instrument
designed and implemented by Epic’s survey expert, Professor Rossi, and the validity and
reliability of his survey results. This is my direct testimony as if I were in court testifying in
person, and is given under penalty of perjury.
III. iOS App Survey and iOS Fortnite Survey
5. I was asked to assess the use of or access to devices other than Apple smartphones and
tablets (“iOS Devices”) on which users can access digital gaming content (“Other Electronic
Devices”) 2 by two specific populations: iOS Device users in the United States who visited the
2 “Other Electronic Devices” or “Other Electronic Device” include smartphones and tablets other
than iOS Devices, personal computers (desktops and laptops, including such products
manufactured by Apple), gaming consoles, and handheld gaming devices. See Complaint for
Injunctive Relief, Epic Games, Inc. v. Apple Inc., U.S. District Court for the Northern District of
California, Civil Action No. 4:20-CV-05640-YGR, August 13, 2020, ¶ 23. (“Epic has developed
several popular entertainment software products that can be played on an array of platforms—
such as personal computers, gaming consoles, and mobile devices.”).

Page 2
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO4 THE
Page of 30
PROTECTIVE ORDER
App Store and downloaded apps (“iOS App Store Users”), and iOS Device users in the United
States who played Fortnite on their iOS Devices (“iOS Fortnite Players”). 3 I was also asked to
assess whether iOS Fortnite Players use Other Electronic Devices to play digital games. To
carry out this assignment, I conducted two surveys targeting the two different populations
identified above (“iOS App Survey” and “iOS Fortnite Survey,” respectively).
A. Surveys’ Design and Implementation
6. In designing and implementing my surveys, I followed standard scientific methods to

ensure the reliability of the results, including (but not limited to): identifying the population of
interest and generating a sample that is representative of that population; minimizing or
eliminating bias in questionnaire design (e.g., potential biases stemming from question wording,
ordering of questions, or answer choices); ensuring high-quality answers (e.g., by avoiding
confusion or guessing by respondents); and using measures to ensure data integrity (e.g.,
attention check questions and decoys). I also conducted rigorous pretests of both surveys, which
involved interviews of each pretest participant about the survey they took by a professional
survey interviewer who was not privy to the purpose of the survey. The results of the pretests
showed that respondents understood the questions as I intended and had no difficulty providing
answers.
7. I used an online panel to field the two surveys and applied measures to ensure that each
of my samples was representative of the relevant target populations. For the iOS App Survey, I
required that the demographics of respondents who started the survey were representative of the
U.S. Census demographics in terms of age, gender, and geographic region (commonly referred to
by survey practitioners as “click balancing”). For the iOS Fortnite Survey, I imposed
demographic quotas for age and gender to ensure that the demographics of respondents were
representative of the Fortnite players’ known demographics, and applied geographic quotas
representative of the U.S. Census demographics. The sample for both surveys included
respondents aged 13 or older. People younger than 18 commonly access digital gaming content
on a variety of devices. Further, Fortnite is deemed suitable for users aged 13 and older, and the
13 to 16 year old demographic comprises a significant part of the population of Fortnite players.
Therefore, it was important to include this younger demographic group in the sample. Parental
consent was obtained for all survey participants under 18 years old.
8. In the screening section of the surveys, I asked respondents a series of questions to
identify those who met the criteria for the target population. In particular, I asked respondents to
identify the types of devices they regularly used in the last 12 months (the list included
smartphones, tablets, and “decoy” electronic devices). I then asked respondents who indicated
they regularly used a smartphone or tablet in the last 12 months to identify the operating systems
of these devices from a list. I asked the question separately for smartphones and tablets, but the
3 Further, for iOS App Store Users, I was asked to collect information about paid app purchases
and in-app purchase behavior on iOS Devices. For iOS Fortnite Players, I was asked to collect
information about their in-app purchase behavior when playing Fortnite on their iOS Devices.
Page 3
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO5 THE
Page of 30
PROTECTIVE ORDER
choices of operating systems were identical. To ensure respondents could respond accurately, I
provided examples of popular smartphone and tablet models for each operating system. Exhibit
1 shows the question posed to respondents who indicated that they had regularly used a
smartphone in the last 12 months. Respondents who did not select “Apple iOS” as the operating
system for either a smartphone or tablet were terminated.
Exhibit 1
Source: DX-4871.012; DX-4875.012

Page 4
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO6 THE
Page of 30
PROTECTIVE ORDER
9. In the iOS App Survey, I asked respondents to indicate whether they had visited the App
Store to download apps on the iOS Devices they regularly used in the last 12 months. To ensure
that they could not guess the criterion for continuing with the survey, I included several “decoy”
activities, such as listening to music, editing digital photographs or images, and streaming and
watching videos, as seen in Exhibit 2 below. Respondents who did not indicate that they had
downloaded apps on the iOS Devices they regularly used in the last 12 months were terminated.
Exhibit 2
Source: DX-4871.014

Page 5
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO7 THE
Page of 30
PROTECTIVE ORDER
10. In the iOS Fortnite Survey, I asked respondents to identify which games they had played
on the iPhones and/or iPads they had regularly used in the last 12 months. As seen in Exhibit 3,
I provided respondents a list of games based on lists of popular games available on iOS Devices
that covered a mix of gaming genres (including Fortnite). As an additional quality assurance
measure, I included three fake “decoy” games (“Sands of Alethkar,” “Guard Dogs,” and “Sword
of Radch”) to catch respondents who were either not paying attention or were trying to guess the
criterion for continuing with the survey. Respondents who did not select Fortnite, as well as
those that selected any of the “decoy” games, were terminated.
Exhibit 3
Source: DX-4875.015

Page 6
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO8 THE
Page of 30
PROTECTIVE ORDER
11. Across both surveys, for respondents who qualified for the main questionnaires and had
regularly used both a smartphone and a tablet in the last 12 months, I asked about other
smartphones and tablets they indicated they had not regularly used in this time frame. In these
questions, respondents were shown only types of smartphones and tablets they did not indicate
regularly using in the screener section. Exhibit 4 below provides an example of the question
posed for iPhone users. In this question, I asked respondents to indicate whether (1) this type of
device was available for them to regularly use, but they did not regularly use it in the last 12
months; or (2) this type of device was not available for them to regularly use in the last 12
months. I also provided an example for devices available to regularly use, as seen in Exhibit 4.
Exhibit 4
Source: DX-4871.018; DX-4875.021

Page 7
HIGHLY
490-5 ONLY
Filed SUBJECT
04/27/21 TO9 THE
Page of 30
PROTECTIVE ORDER
12. The next question in both surveys probed respondents about a larger set of Other
Electronic Devices they regularly used or could have regularly used in the last 12 months. I
provided respondents five different types of electronic devices, and included examples of the
most popular brands for each device type. As seen below in Exhibit 5, I asked respondents to
indicate whether (1) they regularly used this type of device in the last 12 months; (2) this type of
device was available for them to regularly use, but they did not regularly use it in the last 12
months; or (3) this type of device was not available for them to regularly use in the last 12
months. I also included a “Don’t know / not sure” option for each type of device.
Exhibit 5
Source: DX-4871.020; DX-4875.023

13. I posed questions about devices that respondents used or could have used “in the last 12
months” in both of my surveys to reduce the impact of seasonality and other idiosyncratic
circumstances on device usage patterns. I asked respondents about “regular use” to narrow the
relevant universe of devices to a set that respondents could easily identify as those they actually
used or those they could have used on a regular basis (and exclude devices that they would rarely
or occasionally use). Due to the natural variation in the frequency of device use between

Page 8
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
10THE
of 30
PROTECTIVE ORDER
respondents and/or devices, this framing also allowed flexibility for respondents to be able to
identify devices appropriately based on what regular use would mean for each of them and for
each of their relevant devices.
14. Similarly, I used the term “available” when I asked about devices respondents did not
regularly use but could have regularly used if they wanted to. In all questions using this term, I
provided respondents with a simple example to further ensure what I was asking was clear. I
pretested both surveys and confirmed that pretest participants did not have difficulty identifying
the devices they had regularly used or could have regularly used in the last 12 months.
15. In the iOS Fortnite Survey, I asked respondents to identify whether they played games on
the Other Electronic Devices they identified as having regularly used in the last 12 months (see
Exhibit 6). I did not include iPhones and/or iPads in the list since I already asked the
respondents in the screening questions about their gaming behavior on their iOS Devices (see
Exhibit 3). I also did not include gaming consoles and/or handheld gaming devices in the list to
avoid confusion among respondents as to why they were being asked whether they played games
on gaming devices. When I analyzed the share of respondents who used Other Electronic
Devices to play games, I implicitly assumed that respondents who regularly used a gaming
device, used the device to play games. While consumers may use these devices for multiple
purposes in addition to playing games (such as watching streamed shows and movies), these
devices are marketed as designed specifically for gaming activity, so it is reasonable to assume
that the vast majority of, if not all, users of these devices would use them to play games.
Exhibit 6
Source: DX-4875.026
B. Survey Results
16. In my iOS App Survey, I found the vast majority of iOS App Store Users regularly used
Other Electronic Devices in the last 12 months (see Exhibit 7 below). In particular:

Page 9
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
11THE
of 30
PROTECTIVE ORDER
• 92 percent of iOS App Store Users had regularly used at least one type of Other
Electronic Devices in the last 12 months;
• 81 percent of iOS App Store Users had regularly used at least one type of Other
Electronic Devices not manufactured by Apple in the last 12 months;
• 99 percent of iOS App Store Users had regularly used or could have regularly
used at least one type of Other Electronic Devices in the last 12 months;
• 95 percent of iOS App Store Users had regularly used or could have regularly
used at least one type of Other Electronic Devices not manufactured by Apple in
the last 12 months;
• 41 percent of iOS App Store Users indicated that they had regularly used gaming
consoles and/or handheld gaming devices in the last 12 months; and
• 61 percent of iOS App Store Users indicated that they had regularly used or could
have regularly used gaming consoles and/or handheld gaming devices in the last
12 months.
Exhibit 7
Other Electronic Devices Used or Available to Use by iOS App
Store Users (Share of Respondents)
Regularly Used This Device or

Device Was Available
Regularly Used This Device to Regularly Use
Device in the Last 12 Months in the Last 12 Months
Smartphones with Non-iOS Operating Systems 27% 56%

Android 22% 49%
Microsoft 13% 30%
Tablets with Non-iOS Operating Systems 23% 48%
Android 18% 41%
Microsoft 11% 31%
Laptops 71% 86%
Apple 36% 47%
Brands Other Than Apple 50% 68%
Desktops 48% 64%
Apple 22% 35%
Brands Other Than Apple 36% 53%
Gaming Consoles and/or Handheld Gaming Devices 41% 61%
Nintendo Switch (including Nintendo Switch Lite) 21% –
PlayStation (PS Series Console) 25% –
Xbox 21% –
Nintendo DS Series 10% –
PlayStation Vita 6% –
GPD XD Plus 1% –
Other Electronic Devices 92% 99%

Other Electronic Devices (Non-Apple) 81% 95%
Source: DX-4714

Page 10
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
12THE
of 30
PROTECTIVE ORDER
17. In the iOS Fortnite Survey, I found that the vast majority of iOS Fortnite Players
regularly used Other Electronic Devices in the last 12 months and regularly used these devices to
play games (see Exhibit 8 below). In particular:
• 97 percent of iOS Fortnite Players regularly used at least one type of Other
Electronic Devices in the last 12 months;
• 94 percent of iOS Fortnite Players regularly used at least one type of Other
Electronic Devices not manufactured by Apple in the last 12 months;
• 99 percent of iOS Fortnite Players regularly used or could have regularly used at
least one type of Other Electronic Devices in the last 12 months;
• 99 percent of iOS Fortnite Players regularly used or could have regularly used at
least one type of Other Electronic Devices not manufactured by Apple in the last
12 months;
• 79 percent of iOS Fortnite Players indicated that they regularly used gaming
consoles and/or handheld gaming devices in the last 12 months;
• 90 percent of iOS Fortnite Players indicated that they regularly used or could
have regularly used gaming consoles and/or handheld gaming devices in the last
12 months; and
• 94 percent of iOS Fortnite Players played games on at least one type of Other
Electronic Device in the last 12 months.

Page 11
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
13THE
of 30
PROTECTIVE ORDER
Exhibit 8
Other Electronic Devices Used or Available to Use and Other Electronic Devices Used to Play Games
by iOS Fortnite Players (Share of Respondents)
Regularly Used This Device or

Device Was Available to
Regularly Used This Device Regularly Use in the Used This Device to Play
Device in the Last 12 Months Last 12 Months Games in The Last 12 Months
Smartphones with Non-iOS Operating Systems 38% 72% 27%

Android 28% 62% 22%
Microsoft 19% 43% 8%
Tablets with Non-iOS Operating Systems 33% 67% 18%
Android 25% 55% 14%
Microsoft 16% 40% 7%
Laptops 80% 91% 57%
Apple 44% 56% 28%
Brands Other Than Apple 59% 78% 37%
Desktops 59% 76% 41%
Apple 30% 46% 16%
Brands Other Than Apple 48% 66% 31%
Gaming Consoles and/or Handheld Gaming Devices 79% 90% 79%
Nintendo Switch (including Nintendo Switch Lite) 42% – 42%
PlayStation (PS Series Console) 55% – 55%
Xbox 49% – 49%
Nintendo DS Series 20% – 20%
PlayStation Vita 11% – 11%
GPD XD Plus 2% – 2%
Other Electronic Devices 97% 99% 94%

Other Electronic Devices (Non-Apple) 94% 99% 90%
Source: DX-4663; DX-4754
18. I performed sensitivity analyses and robustness checks of my results in Exhibits 7 and 8
above. In particular:
a. I recalculated the results for three mutually exclusive subsets of respondents:
those who regularly use both an iPhone and an iPad, those who regularly use an
iPhone only, and those who regularly use an iPad only.
b. I recalculated the results after I excluded sets of respondents whose responses
might indicate they were inattentive: respondents who completed the survey too
quickly or too slowly, 4 and respondents who indicated they regularly used or
could have regularly used smartphones operated by a Windows Microsoft
operating system.
4 I used two different methods for the robustness test with respect to time spent completing the
survey. First, I excluded respondents who were outliers in terms of overall survey completion
time between all respondents (for the iOS App Survey, this was defined as those who completed
the survey in under two minutes or in fourteen minutes or more; for the iOS Fortnite Survey, this
was defined as those who completed the survey in under two minutes or in sixteen minutes or
more). Second, I excluded respondents who completed the survey in under three minutes.
Page 12
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
14THE
of 30
PROTECTIVE ORDER
19. These changes had no material impact on the results of my surveys, as I still found that
the vast majority of respondents regularly used or could have regularly used Other Electronic
Devices in the last 12 months, and that the vast majority of iOS Fortnite Players regularly used
Other Electronic Devices to play games. Even if certain survey respondents are removed from
the analysis, my overall results and conclusions do not change in a meaningful way.
20. Professor Rossi’s survey data about device usage corroborate the main results of my
surveys. Professor Rossi asked his survey respondents to indicate which (if any) of the following
types of devices they used: smartphones (Q7), tablets (Q8), computers / laptops (Q12), and
gaming consoles (Q13). 5 Analysis of these responses shows that the vast majority of Professor
Rossi’s respondents used Other Electronic Devices (see Exhibit 9). 6
Exhibit 9
Device Usage by Professor Rossi's
Survey Respondents
Professor Rossi's
Device Respondents
Any Other Electronic Devices 98%

Any Other Electronic Devices (Non-Apple) 93%
Source: Summarizing PX-2545
IV. Professor Rossi’s Survey Methodology Was Flawed and His Results Are Unreliable
21. I was also asked to review and evaluate the reliability of Professor Rossi’s survey
instrument and the validity and reliability of his survey results.
22. Professor Rossi designed and fielded a survey of U.S. residents aged 17 and older using a
sample limited to respondents who had used an iOS Device and made in-app purchases or
purchased subscriptions through the App Store in the 30 days before taking the survey. 7 The
main goal of Professor Rossi’s survey was to measure how consumers would react to a
hypothetical price increase scenario. 8 To achieve this, Professor Rossi asked respondents how
they might have reacted had there been a five percent increase hypothetically applied to the
prices of the in-app purchases and subscriptions respondents had made in the past 30 days on
their iOS Devices.
23. Academic research demonstrates that survey respondents’ predictions about how they
would have changed their behavior in a hypothetical situation (especially in relation to prices)
5 PX-2547.8–9; 15–16.
6 While Professor Rossi’s target population and his questions are different from mine, these
survey responses provide relevant insights into the device usage question my surveys address.
7 Rossi, ¶ 4.
8 Rossi, ¶ 2.

Page 13
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
15THE
of 30
PROTECTIVE ORDER
are not reliable indicators of their actual behavior, and that such survey methods generate
unreliable responses. Further, as I describe below, Professor Rossi’s survey generated biased
and unreliable responses because he failed to follow standard survey practices in conducting his
survey, including not pretesting his final survey instrument. Even if Professor Rossi’s survey
results were reliable, which they are not, they are not generalized to the broader population he
was asked to target in his assignment. Furthermore, the results cannot be generalized to any
hypothetical price increase scenario other than the five percent price increase he tested.
A. The Stated Preference Survey Method that Professor Rossi Used Is Widely
Regarded as Unreliable
24. Professor Rossi asked respondents to predict how they might have changed their
purchasing behavior for “at-issue purchases” 9 made in the last 30 days on the App Store in
response to a hypothetical price increase of five percent.
25. Academic research documents the gap between actual behavior and survey respondents’
statements as to how they would behave (i.e., their stated preferences). One explanation for this
well-documented disconnect is known as “hypothetical bias,” or respondents’ inability to reliably
predict their behavior in hypothetical situations for which they do not have enough real-world
context to tie their responses to their actual behavior. As Professor Jerry Hausman of MIT noted,
“put simply, what people say is different from what they do.” 10
26. Professor Rossi has also acknowledged that simply asking respondents to predict their
behavior yields unreliable results 11 and has advocated for the use of other survey methods, such
as choice-based conjoint, 12 to avoid the use of stated preference data. Professor Rossi states in
one publication that “[v]irtually all researchers in marketing accept the premise that choice-based
conjoint studies offer superior recovery of consumer preferences than a pure stated preference
method in which direct elicitation of preferences is attempted.” 13
27. However, here Professor Rossi used a “pure stated preference method,” asking
respondents to predict how they might have changed their behavior in a hypothetical situation
involving a price increase.
9 Professor Rossi defines “at-issue purchases” as “in-app purchases and subscriptions when
purchased from within iOS apps.” Rossi, ¶ 2.
10 J. Hausman (2012), “Contingent Valuation: From Dubious to Hopeless,” Journal of Economic
Perspectives, 26, 4, 43–56 at p. 44.

11 See, for example, Rossi, P. E., G. Allenby, and N. Hardy (2019), “Economic Foundations of
Conjoint Analysis,” in Handbook of the Economics of Marketing, Rossi, P.E. and J. P. Dube, eds.
Oxford, United Kingdom: Elsevier, 151–192 (“Rossi et al.”) at pp. 152–154; DX-4942.
12 A conjoint survey is a survey method pursuant to which the researcher analyzes trade-offs
respondents make when choosing between different products given various choices to infer their
preferences.
13 Rossi et al. at pp. 152–153. Emphasis added; DX-4942.

Page 14
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
16THE
of 30
PROTECTIVE ORDER
B. Professor Rossi’s Survey Results Are Not Generalizable to the Broader

Population He Was Asked to Study
28. Even if the results of Professor Rossi’s survey were reliable, which they are not, they are
not generalizable either to iOS Device users younger than 17 years old or to iOS Device users
who made in-app purchases or subscriptions outside the 30-day window of Professor Rossi’s
survey.
29. Professor Rossi’s target population (U.S. residents aged 17 and older who have used an
iOS device and have made at-issue purchases in the past 30 days 14) does not match the
population of U.S. consumers who make in-app purchases or buy subscriptions on iOS Devices
because:
• It excludes the important segment of 13 to 16 year-olds, despite the fact that this
demographic group comprises a significant part of the population of Fortnite
players. 15 Younger buyers may have a different price elasticity than those included in
Professor Rossi’s survey. Further, Professor Rossi provides no evidence that younger
buyers do not make the purchase decisions for their at-issue purchases.
• It excludes individuals who make in-app purchases and subscriptions but who did not
do so in the 30 days before Professor Rossi’s survey. Infrequent buyers may have a
different price elasticity than frequent consumers.
30. Additionally, the results of Professor Rossi’s survey are not generalizable to any
hypothetical price increase scenarios other than the five percent price increase he tested.
Professor Rossi’s estimated price elasticity of demand therefore cannot be generalized.
C. Professor Rossi’s Failure to Pretest His Final Survey Instrument Violates

Standard Survey Practices
31. Professor Rossi generated four versions of his survey instrument: Initial Draft Survey,
Version 1 (“V1”), Version 2 (“V2”), and Version 3 (“V3”). 16 In each version, Professor Rossi
posed three main questions (Q16, Q17, and Q18) that attempted to elicit respondents’ predictions
about how they might react to his hypothetical price increase scenario. The wording of the
questions and answer options changed substantially across the different versions of the survey
instrument.
32. It is standard survey practice to conduct pretests, which involve interviewing a small
group of subjects, to identify “questions that respondents have difficulty understanding or
interpret differently than the researcher intended” before administering the survey instrument to a
14 Rossi, ¶ 15.
15 DX-3233.
16 See DX-4667, DX-4668.

Page 15
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
17THE
of 30
PROTECTIVE ORDER
full sample. 17 Depending on the findings of pretests, changes can be made to the survey
instrument. In some cases, these changes are minor (e.g., a word in the question is underlined to
ensure respondents read the question accurately), but in others they are extensive (e.g., wording
of questions or answer options is significantly altered or new questions or answer options are
added to the survey). In the latter scenario, additional pretests, namely interviews of subjects,
must be conducted to ensure the alterations to the survey instrument provide the intended
clarification and respondents understand and interpret the new questions and answer options
appropriately.
33. Professor Rossi pretested his Initial Draft Survey (or conducted an “unstructured pre-test”
of this initial version, per Professor Rossi’s non-standard terminology). 18 This included
interviews with eight respondents. Based on these interviews, Professor Rossi identified
multiple issues with the survey instrument, including respondent confusion about descriptions of
the at-issue purchases and the hypothetical price increase scenario. 19 He then made three rounds
of substantive revisions to his main questions to address these issues, reflected in survey
instruments V1, V2, and V3, respectively. Exhibit 10 lists some of the substantial changes that
Professor Rossi made, the version in which he made the changes, and whether respondents were
interviewed as part of any pretests after these changes were made. But none of these changes
were pretested. 20
Exhibit 10
Professor Rossi’s Survey Changes
Change to Survey Instrument First Introduced In Final Survey Pre-tested
Added details to explain the hypothetical price increase V1 No
Changed spending level subject to hypothetical price

V1 No
increase from 12 months to 30 days
Changed hypothetical price increase scenario from
V3 No
forward-looking to backward-looking
Changed the answer option associated with “Stickers”

V3 No
from "No" to "Yes" in question Q16
Source: DX-4667, DX-4668, DX-4670
17 Krosnick, J. A. (1999), “Survey Research,” Annual Review of Psychology, 50, 537–567

(“Krosnick”) at p. 541.
18 Rossi, ¶¶ 18–19; DX-4665.
19 Rossi, ¶ 20.
20 In addition to the changes noted in Exhibit 10, Professor Rossi also failed to pretest a number
of other substantive changes to his survey.

Page 16
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
18THE
of 30
PROTECTIVE ORDER
34. Professor Rossi conducted a separate pilot study for each revised version of his survey
instrument (see Exhibit 11). He uses the misleading, non-standard phrase “structured pre-test”
to present these pilot studies as a type of pretest. While pretests involve interviewing
respondents about the survey in an open-ended manner by a professional interviewer, pilot
studies do not involve interviews or any direct interaction with respondents, and entail nothing
more than administering the survey to a group of respondents and evaluating their collected
answers. 21 Indeed, Professor Rossi affirmed this point in his testimony in another matter, noting
that “[p]ilot testing is not the same thing as a pre-test.” 22
Exhibit 11
Professor Rossi’s Survey Revision Process

Type of Study Survey Number of Respondent
Fielding Period As Defined by Professor Rossi Instrument Respondents Interviews
12/23/20 – 1/2/21 Unstructured Pretest Initial Draft Survey 8 Yes
1/7/21 – 1/8/21 Structured Pretest Version 1 Version 1 (V1) 36 No
Source: DX-4665
35. Because Professor Rossi failed to pretest his final survey instrument, he has no assurance
that the questions in his final survey instrument were understood and appropriately interpreted by
respondents. In fact, the data generated across Professor Rossi’s pilot studies do not support the
inference that respondents were not confused; as I show below in Exhibit 13, the data suggest
that he did not solve the problems he identified, and his finalized survey instrument was still
susceptible to respondent confusion and generated unreliable responses. As Professor Rossi
stated in another matter, it is “incumbent upon any market survey professional” to “provide
proper evidence of a valid pre-test” and, absent this, “the results [from a survey] are not
reliable.” 23 Professor Rossi failed to do so here. His survey results are therefore unreliable.
21 Ruel, E., Wagner, W. E., and B. J. Gillespie (2016), The Practice of Survey Research: Theory
and Applications, Thousand Oaks, CA: SAGE Publications, Inc., pp. 114–115.
22 Contingent Rebuttal Witness Statement of Peter E. Rossi, Certain Gaming and Entertainment
Consoles, Related Software, and Components Thereof, U.S. International Trade Comm., Wash.,
D.C., Investigation No. 337-TA-752, Nov. 13, 2012 (“Rossi ITC Rebuttal Witness Statement”),
¶ 38; DX-4944.
23 Rossi ITC Rebuttal Witness Statement, ¶ 57; DX-4944.

Page 17
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
19THE
of 30
PROTECTIVE ORDER
D. Question Q16
1. Professor Rossi’s Confusing Hypothetical in Question Q16

36. When designing surveys based on hypothetical scenarios, it is necessary to describe the
hypothetical in a manner that is consistent with how respondents would experience it in the real
world. 24 Failing to do so results in respondent confusion and unreliable responses. 25 As
Professor Rossi has himself explained, using questions untethered from the real world “violates
everything that anyone has taught or written about survey research; namely, you have to expose
people to stimuli that they can understand and appreciate in the way in which they come to terms
with the real world on a day-to-day basis.” 26
37. In question Q16, Professor Rossi asked respondents whether they would have kept or
changed in-app purchases or subscriptions made on the App Store in the past 30 days in response
to a hypothetical price increase (see Exhibit 12). However, Professor Rossi did not provide the
necessary context for respondents to properly answer this question, and he employed a
backward-looking hypothetical that resulted in respondent confusion.
24 See Diamond, S. S., and J. B. Swann, eds. (2012), Trademark and Deceptive Advertising
Surveys: Law, Science, and Design, Chicago, IL: ABA Publishing (“Diamond and Swann”), at p.
249, (“[T]he suitable method and the determination of whether a survey creates experimental
demand effects and biased results, relative to what is likely to occur in reality, depends largely on
the pertinent marketplace conditions. That is, whether a survey creates responses that deviate
from what naturally occurs in reality depends on whether the manner in which survey
respondents are exposed to the stimuli at issue is fundamentally different from what consumers
encounter in the marketplace.”).
25 See Diamond and Swann, at p. 259 (“[A] primary criterion for assessing the reliability of
surveys and the severity of demand effects is based on the degree to which they alter the
fundamental conditions in which the marks or terms at issue are encountered by buyers in
reality.”).
26 DX-4946, 258:22–259:2.

Page 18
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
20THE
of 30
PROTECTIVE ORDER
Exhibit 12
Source: PX-2547.35
38. First, Professor Rossi told his respondents that prices in the App Store increased by five
percent without specifying the reason for the price increase or providing any information
regarding the prices of the at-issue purchases on non-iOS Devices. Based on feedback from his
pretest participants, in the second version (V1) of his survey, Professor Rossi added the
following language “[n]othing else about your apps or IAPs/subscriptions has changed.” 27 This
change addresses only one potential reason for the price increase (i.e., a change to the app, in-app
or subscription that is being purchased) and the question still lacks context for the reason for the
price increase. Because Professor Rossi did not interview and ask a single respondent whether
the additional language did provide sufficient context, he has no evidence that this source of
respondent confusion was properly remedied.
39. Second, Professor Rossi changed the hypothetical scenario posed in question Q16 from a
forward-looking exercise (a scenario where the prices of the at-issue purchases respondents had
made over the past 30 days would hypothetically increase by five percent in the future) to a
backward-looking exercise (a scenario where the prices of the at-issue purchases respondents had
made over the past 30 days hypothetically increased by five percent 30 days ago). 28 The
backward-looking hypothetical in question Q16 does not resemble any decision respondents
would make in the real world. In the real world, consumers make decisions on purchases for
goods or services they intend to consume immediately or in the future. In contrast, Professor
27 DX-4667.
28 DX-4668.
Page 19
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
21THE
of 30
PROTECTIVE ORDER
Rossi’s respondents were asked to make decisions for purchases they had already paid for at a
price level that was not a part of their decision-making process at the time.
40. I analyzed the data collected by Professor Rossi in his pilot studies and final survey and
concluded that the backward-looking scenario in Professor Rossi’s final survey was indeed more
confusing to respondents. Exhibit 13 shows that the share of respondents who chose the don’t
know / unsure option (a potential indication of confusion 29) was substantially higher in the
backward-looking scenario (10 percent) than in the forward-looking scenario (4 percent). 30
Exhibit 13
Percentage of Non-“Deciders” Identified in Question Q16 Under the
Forward-Looking and Backward-Looking Hypothetical Scenarios
Percentage of
Respondents
12%
Backward-Looking
Hypothetical
10%
10%
8%
6%
Forward-Looking
Hypothetical
4%
4%
2%
0%
Q16. “Would the 5% price increase cause you to make fewer Q16. “Thinking about the same 30-day period, would you have
purchases in the future from the Apple App Store?” made the same purchases of IAPs/subscriptions from the
Apple App Store with the higher prices?”
A. “Don't know”
A. “Not sure what I'd have done”
Out of 74 combined responses for pilot study for V1 and V2 Out of 2,662 combined responses for pilot study for V3 and the final survey
Source: DX-4669
29 Literature on surveys shows that respondents are “particularly likely to be attracted to a ‘don’t
know’ option when the question is difficult to understand or the respondent is not strongly
motivated to carefully report an opinion.” Diamond, S. S. (2011), “Reference Guide on Survey
Research,” in Reference Manual on Scientific Evidence, Washington, D. C.: The National
Academies Press, 359–423 (“Diamond”), at p. 391.
30 This difference is statistically significant at the five percent level.

Page 20
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
22THE
of 30
PROTECTIVE ORDER
41. This increase in no-opinion responses suggests respondents had more difficulty
understanding the backward-looking hypothetical compared to the forward-looking hypothetical.
2. Framing of the Hypothetical Scenario and Answer Choices Artificially
Inflated the Number of “Stickers” in Question Q16
42. Professor Rossi’s framing of the hypothetical scenario and answer choices in question
Q16 are likely to induce many respondents to answer that they would make “the same
purchases” and thus inflate the number of “Stickers” in Professor Rossi’s survey. Analysis of
data from Professor Rossi’s pilot studies and his final survey instrument demonstrates that his
questionnaire design choices caused the percentage of “Stickers” to increase by more than 40
percent in proportional terms, from 51 percent to 73 percent (see Exhibit 15 below).
a) Acquiescence Bias
43. Academic literature has shown that respondents are more likely to answer “yes” than
“no,” or otherwise agree with the question posed, all else equal. This tendency is sometimes
referred to as acquiescence bias or the yea-saying response tendency. 31 In fact, one of the
authorities that Professor Rossi has referenced in this matter warns that “acquiescence, ‘the
tendency to endorse any assertion made in a question, regardless of its content,’ is a systematic
source of bias.” 32
44. Professor Rossi framed question Q16 as a yes/no question. In his initial survey, “No”
indicated that a respondent would not change past purchases from the App Store, while “Yes”
indicated that a respondent would change past purchases. In his third round of revisions to his
survey instrument, Professor Rossi reversed the meaning of the “Yes” and “No” options (see
Exhibit 14), so that leaving past purchases from the App Store unchanged was associated with a
“Yes” answer. 33 Acquiescence bias suggests the percentage of “Stickers” should increase as a
result, which was indeed the case.
31 Hurd, M. (1999), “Anchoring and Acquiescence Bias in Measuring Assets in Household

Surveys,” Journal or Risk and Uncertainty, 19, 1, 111–136 at pp. 116–117; Diamond and
Swann, at p. 274, (“The reason a question that may be answered by a mere ‘yes’ or ‘no’ is likely
to be leading is because, all other things being equal, respondents – generally agreeable people
who have agreed to participate in the first place – are more inclined to be agreeable and answer
‘yes’ than to answer ‘no.’ Given this ‘yea-saying response tendency,’ all other things being
equal, yes/no questions are much more likely to garner answers of yes rather than no. […] [T]he
use of yes/no questions can pose serious problems.”).
32 Diamond, at p. 394.
33 Professor Rossi defines respondents who said that they would have changed their purchases in
response to the hypothetical price increase as “Decrementers.” Rossi, ¶ 11.

Page 21
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
23THE
of 30
PROTECTIVE ORDER
Exhibit 14
Reversal of Answer Choice Identifying "Stickers" in Question Q16

Professor Rossi's Answer Choices in Initial Draft Survey,
Classification V1, and V2 Answer Choices in V3 and Final Survey
“Sticker” No, the price increase would not cause me to Yes, I would have made the same purchases
make fewer purchases from the Apple App and spent $16.78
Store
“Decrementer” Yes, the price increase would cause me to No, I would have changed my purchases and
make fewer purchases from the Apple App spent less than $16.78
Store
Source: DX-4670
45. Professor Rossi explained that he reversed the “Yes” and “No” options to correct a
“double negative,” 34 but his pretests of the Initial Draft Survey provided no evidence that
respondents were confused by this phrasing, and he did not implement this change until the
fourth version of his survey instrument (i.e., V3). Nor did he interview any respondents (or
conduct a pretest) after he changed these answer options.
b) Endowment Effect
46. It is a well-established finding in consumer behavior research that people often demand
more to give up something they already own than they would pay to acquire it. This
phenomenon is known as the endowment effect. 35 Pursuant to this, people will consider
foregone gains less painful than perceived losses.
47. When Professor Rossi asked respondents in question Q16 whether they would have
changed their past purchases (or changed purchases that the respondents were already
“endowed” with), respondents were in effect asked to second-guess a decision they had already
made and benefited from. Thus, respondents were asked whether they would have given up a
purchase whose benefits they had already enjoyed. The endowment effect highlights that
respondents would stick to the purchases they had already made.
c) Impact of Framing of the Hypothetical Scenario and Answer

Choices
48. To assess the impact of Professor Rossi’s changes to question Q16, I analyzed the data
collected by Professor Rossi and compared (a) the percentage of “Stickers” in pilot studies for
34Rossi, ¶ 20.
35See Kahneman, D., J. Knetsch, and R. H. Thaler (1991), “Anomalies: The Endowment Effect,
Loss Aversion, and Status Quo Bias,” Journal of Economic Perspectives, 5, 1, 193–206.
Page 22
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
24THE
of 30
PROTECTIVE ORDER
V1 and V2 of Professor Rossi’s survey instrument, in which the “No” option corresponded with
“Stickers” and the price increase hypothetical was forward-looking with (b) the percentage of
“Stickers” in the pilot study for V3 of Professor Rossi’s survey instrument and the final survey
instrument, in which the “Yes” option corresponded with “Stickers” and the price increase
hypothetical was backward-looking.
49. As Exhibit 15 demonstrates, the percentage of “Stickers” increased dramatically from 51
percent in V1 and V2 of the survey to 73 percent in V3 and the final survey. 36 The difference
between the two groups is statistically significant at the five percent level.
Exhibit 15
Percentage of “Stickers” Identified in Question Q16 Under the
Forward-Looking and Backward-Looking Hypothetical Scenarios
Percentage of
Respondents
Backward-Looking
80% Hypothetical
73%
70%
Forward-Looking
60% Hypothetical
51%
50%
40%
30%
20%
10%
0%
Q16. “Would the 5% price increase cause you to make fewer Q16. “Thinking about the same 30-day period, would you have
purchases in the future from the Apple App Store?” made the same purchases of IAPs/subscriptions from the
Apple App Store with the higher prices?”
A. “No, the price increase would not cause me to make
fewer purchases from the Apple App Store.” A. “Yes, I would have made the same purchases and spent
<DROP IN AMOUNT ADJUSTED FOR 5% INCREASE>.”
Out of 74 combined responses for pilot study for V1 and V2 Out of 2,662 combined responses for pilot study for V3 and the final survey
Source: DX-4671
36The figure shows that the versions with the backward-looking framing (V3 and final survey
instrument) and where the “Yes” option corresponds to “Stickers” yielded significantly higher
percentages of respondents who chose to make the same purchases (73 percent) than those in the
versions with the forward-looking framing and where the “No” option corresponds to “Stickers”
(51 percent, in V1 and V2).

Page 23
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
25THE
of 30
PROTECTIVE ORDER
50. Finally, Professor Rossi’s hypothetical price increase was restricted to purchases made in
the past 30 days, resulting in a trivial nominal price increase for the vast majority of respondents;
more than 80 percent of respondents who completed question Q16 in Professor Rossi’s survey
were presented with a price increase of less than $2. 37 Whereas in the Initial Draft Survey
Professor Rossi showed respondents an estimated annual amount of aggregate spending on in-
app purchases and subscriptions, he limited the spending to the significantly narrower 30-day
window in the later revisions and the final survey instrument. This change (which significantly
lowers the hypothetical increase in the purchase price) likely influenced respondents’ predictions
of how they would react to Professor Rossi’s hypothetical price increase and induced many
respondents to make “the same purchases” (i.e., inflated the number of “Stickers”).
E. Question Q17
51. For respondents who indicated they would have changed their purchases, Professor Rossi
asked in question Q17 what the respondents would have done to spend less in the App Store,
after the hypothetical price increase. In question Q17, Professor Rossi overemphasized the need
to use a different device as well as the costs of switching devices, and presented answer choices
that were vague, artificially narrow, and did not properly reflect real-world choices. Responses
to this question are therefore likely to understate the willingness of real-world consumers to
reduce spending in Professor Rossi’s hypothetical price increase scenario by making payments
through means other than the App Store (e.g., using the web browser on an iPhone and making
the purchase on the app developer’s website or using a non-iOS Device, such as a laptop, to
make the purchase).
37 DX-4672.
Page 24
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
26THE
of 30
PROTECTIVE ORDER
Exhibit 16
Source: PX-2547.36–37
52. Specifically, Professor Rossi instructed his respondents to “keep in mind…” potential
switching costs associated with shifting purchases to other devices and provided a link to
additional information about switching costs (see Exhibit 16). While Professor Rossi included
prominent, lengthy reminders about switching costs, he failed to provide relevant context for
what “shift[ing] purchases” entails and whether such a shift always requires a shift to other
devices. Consequently, he implicitly and misleadingly positioned the decision as one of using an
iPhone or iPad versus switching to a different device entirely. 38 Professor Rossi therefore
presented the alternatives to making payments in the App Store as the less attractive and more
costly choices for his survey respondents (e.g., his question suggests that a respondent would
have to buy and use a new Android smartphone or a tablet to avoid making the payment through
the App Store). As a result, question Q17 biases respondents toward keeping payments in the
38While Professor Rossi includes a “None of the above” option, he provides no evidence that his
respondents would properly understand that using the web browser on an iPhone or iPad and
making the purchase on the app developer’s website would be covered under this option.
Page 25
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
27THE
of 30
PROTECTIVE ORDER
App Store and not considering other means for payments in order to reduce spending in the
hypothetical price increase scenario.
53. First, while Professor Rossi specifically reminded respondents of the “costs … to shift
purchases to other devices,” he did not explain how such a shift could result in spending less. In
fact, only the “Sticker” answer choice (“Would have kept my IAP/subscription spending in the
Apple App Store but spent less…”) mentions spending less. Neither of the answer choices that
involve shifting spending to other devices mentions spending less. Professor Rossi made no
attempt to explain how these answer choices could be relevant to spending less.
54. Second, because the answer choices that Professor Rossi presented in question Q17 were
vague, artificially narrow, and did not properly reflect real-world choices, results for question
Q17 likely understate the willingness of real-world consumers to reduce their spending by using
means other than making the payment through the App Store. For example, Professor Rossi:
a. Omitted alternative methods for reducing spending, such as making purchases
through a website using an iOS Device. 39
b. Failed to explain that respondents could continue to consume purchased products
(e.g., digital newspaper content) on their iOS Devices if they chose to make
payments for these products (e.g., a digital newspaper subscription) using other
devices. For example, Professor Rossi described device compatibility costs and
learning how to use new device features and apps as part of his switching costs.
Many of these costs would not be relevant for making payments. Discussing
switching costs in this manner may have led respondents to believe that purchased
content would need to be consumed from a device other than the respondent’s
iPhone or iPad.
c. Framed question Q17 such that respondents had to make an immediate switch
decision to fully or partially avoid the hypothetical price increase, (e.g.,
immediately purchasing a new Android phone as a remedy for the hypothetical
price increase). As a result, such an immediate switch decision would strike
many respondents as confusing and unrealistic.
F. Question Q18
55. For respondents who indicated in question Q16 that they would have changed their
purchases and spent less, Professor Rossi asked, in question Q18, by how much they would
reduce their spending (see Exhibit 17). He again failed to provide respondents with a realistic
setting and proper context. Nor did he allow respondents to indicate that they could not answer
this question because they were unsure or did not know (i.e., he forced all respondents to provide
39See “Netflix Stops Offering In-App Subscriptions for New and Returning Customers on iOS,”
The Verge, December 28, 2018, https://www.theverge.com/2018/12/28/18159373/netflix-in-app-
subscriptions-iphone-ipad-ios-apple, access date: 3/14/2021; DX-5308, Section 3.1.3.
Page 26
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
28THE
of 30
PROTECTIVE ORDER
a monetary value). As a result, responses to question Q18 and Professor Rossi’s estimate of
price elasticity of demand are unreliable.
Exhibit 17
Source: PX-2547.40
56. In the real world, the amount consumers pay would depend on their decision regarding
what to purchase (e.g., keep or discontinue an existing subscription) and the means through
which to make the payment for these purchases (e.g., pay via the App Store, pay using app
developer’s website using an iOS device or a non-iOS device). Therefore, real world consumers
would not be able to reduce their spending in the App Store by any random amount, as they were
essentially asked to do in Professor Rossi’s survey.
57. Further, Professor Rossi’s failure to provide a “don’t know / unsure” option for
respondents violates standard practices in survey research. 40 Though respondents indicated they
would spend less in previous questions, it is highly likely that some respondents did not know
how much less they would have spent in the App Store (in particular because this type of
40 See Diamond, at p. 390 (“[T]he survey can use a quasi-filter question to reduce guessing by
providing ‘don’t know’ or ‘no opinion’ options as part of the question … [b]y signaling to the
respondent that it is appropriate not to have an opinion, the question reduces the demand for an
answer and, as a result, the inclination to hazard a guess just to comply”). See also Krosnick, at
p. 557.
Page 27
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
29THE
of 30
PROTECTIVE ORDER
decision is not made in the real world). By forcing all of his respondents to provide an answer,
Professor Rossi elicited unreliable responses.
G. Professor Rossi’s Purported Evidence of Representativeness Is Misleading
58. Professor Rossi provided several comparisons of his survey sample to external data to
purportedly demonstrate that “the sample of respondents who passed the Reliability Test was
representative of the underlying target population of iOS device users who make at-issue
purchases.” 41 However, the tests he conducted provide a false impression of the
representativeness of his sample.
59. Specifically, Professor Rossi compared device distribution and gender distribution of
respondents in his data to the device distribution and gender distribution of iPhone users in a
survey presented in a RBC Capital Markets securities analyst report. The choice of the RBC
Capital Markets survey as a benchmark is arbitrary for two reasons:
a. The RBC Capital Markets survey does not reflect the relevant target population as
defined by Professor Rossi since it is not limited to those who made at-issue
purchases and is limited to iPhone users only.
b. The RBC Capital Markets survey was conducted by a securities analyst via
SurveyMonkey, and the analyst report that presented this survey did not disclose
information about the survey methodology, such as respondent selection or the
specific questions asked. Without this information, Professor Rossi has no basis
to assume that this survey accurately reflects the relevant population. 42
60. In the second comparison, Professor Rossi compared the median in-app purchases of his
survey respondents to the median in-app purchases in Apple Transaction Data. However, this
specific comparison of the datasets does not ensure Professor Rossi’s data matches the Apple
Transaction Data more generally. Professor Rossi’s own analysis shows that the comparison is
not robust. For example, the 75th percentile of his final survey sample (“Completers”) spent
$25, while the 75th percentile in the Apple Transaction Data spent $34.92. 43
41 Rossi, ¶¶ 47–49.
42 PX-2544.
43 PX-1085.

Page 28
HIGHLY
490-5EYES
FiledONLY SUBJECT
04/27/21 PageTO
30THE
of 30
PROTECTIVE ORDER
V. Oath
and correct.
April 23, 2021

Dominique Hanssens Date
8108 words

Page 29
1. My analysis in this matter focuses on the security, privacy, and reliability of iOS, and
particularly the question of whether Apple’s App Review process and App Store distribution of
apps provide and enhance the security, privacy, and reliability of iOS. It is my overall
conclusion that Apple’s App Review and App Store distribution provide significant security
benefits that meaningfully contribute to iOS being a safer and more trustworthy platform than
others. Allowing alternative app stores for iOS would prevent or weaken Apple’s App Review
and App Store distribution protections and would, in turn, weaken iOS and open it to new
threats. This overall conclusion is based on the following opinions:
2. Opinion 1. Security for computing devices includes detecting and preventing malware.
But it encompasses additional considerations such as ensuring user privacy, preventing scams,
ensuring device reliability, and protecting developers and consumers from software piracy.
(¶¶ 16-28.)
3. Opinion 2. Security assessments must consider the threat model, which is an analysis of
factors including who wants to attack the device / user, how frequently such attacks are likely to
occur, and what the consequences are (for the attacker and the victim) of a successful attack.
With over 1 billion active devices that are almost constantly on and connected to the Internet, a
userbase that frequently downloads apps and engages in financial transaction on the device, and
camera, microphone, and GPS hardware that follows owners nearly everywhere they go, and
nearly two million apps available for download, iOS—and Apple iOS devices such as the
iPhone—are faced with an extraordinary threat model. The portability of these devices has
facilitated their use for critical tasks such as navigation and emergency calls, but also means that
they are at higher risk of being misplaced or stolen, which places the contents of that device at
even greater risk. (¶¶ 22-28.)
4. Opinion 3. Computing security best practices call for layered defenses. This means that
there are multiple defenses in place for each of the multiple types of threats that the device is
likely to face. iOS layers defenses by combining on-device protections, such as access controls
to limit the exposure of data and functionality, with additional review or verification
mechanisms. Each layer of protection strengthens the overall security posture of the system, by
lowering the risk that untrustworthy or malicious apps may reach an iOS device and reducing the
impact of such apps should they make it to the device. (¶¶ 22-43.)
5. Opinion 4. Apple’s App Review (and accompanying app distribution model) is one of
the defense layers that Apple has employed to protect iOS devices and users. Using a
combination of human reviewers and sophisticated computer analytical tools, the App Review
provides significant security and non-security benefits by screening apps for a variety of
potential problems including scams, privacy intrusions, piracy, objectionable content, as well as
problems with reliability and crashing. By positioning itself as the sole source of app
distribution, Apple prevents users from unintentionally or unknowingly downloading apps that
would have failed the App Review process, and also prevents apps that fail the App Review
process from simply relocating to another store with lower (or no) standards. (¶¶ 25-43, 53-76.)

05640-YGR-TSH Aviel D. Rubin, Ph.D.
2
6. Opinion 5. The App Review process, combined with other protective layers employed
by Apple, has resulted in iOS being safer than other platforms, facing fewer attacks and malware
infections than other platforms. Studies have found that Android devices are fifty times more
likely to be infected with malware when compared to iOS, made up 26.64% of infected devices
in 2020 (as opposed to iPhones, which accounted for only 1.72% of infected devices), and saw
three times as many reported Common Vulnerabilities and Exposures (“CVEs”) as Apple, with
Android amassing 2,395 CVEs and iOS 832 CVEs between 2016-2019. Apple has achieved this
result in spite of the extraordinary threat model that iOS device faces. (¶¶ 44-61.)
7. Opinion 6. The introduction of alternative app stores on iOS devices would jeopardize
the security, safety, and trustworthiness of the iOS platform. Many other distribution sources
simply will not prioritize security, safety, and trustworthiness. We know this because outside of
the iOS platform, there exist stores that primarily traffic in adult content, malware, and/or pirated
software. Even distribution sources that mean well would have trouble meeting the standards of
Apple’s App Review. Some of them will lack the resources to build the various tools and
employ the reviewers that Apple currently has on staff. Others will lack the incentives. For
example, developers and third-party stores whose financial model depends largely on ad
revenues will have less incentive to protect user privacy because much of ad revenue is based on
the ability of advertisers to target and know intimate details about end users. And finally, all
other sources will lack Apple’s knowledge of iOS and iPhone hardware and their security
vulnerabilities, as well as the extensive body of knowledge that Apple has accumulated from
more than a decade of app review and analysis of threats posed by apps. Internal knowledge of
iOS and iPhone architecture cannot be simply revealed to third parties because of the potential
associated security threats. (¶¶ 77-117.)
8. Opinion 7. Fragmentation of app distribution and review would undermine security.

Security is only as strong as its weakest link; where multiple app stores are available, a bad actor
could seek out the option that is least likely to protect against malicious or otherwise problematic
apps. And Apple’s App Review benefits from its ability and knowledge gained from holistic
review across its platform. (¶¶ 87-99.)
9. Opinion 8. The current Android marketplace in China demonstrate the problems with
fragmentation. In China, where Google’s Play Store is banned, numerous third-party app stores
operate in its stead. Google has little control over the various Android app stores, which are
known to host a higher prevalence of fake, cloned, and malicious apps than the Google Play store
for which Google conducts app review. In fact, China hosts the top three stores from which
users are most likely to download malware—Xiaomi, Baidu, and Pconline—and more generally,
nearly 35 percent of Android apps were found to secretly steal user data unrelated to the app’s
functionality. The multitude of app stores in China has thus led to generally looser security
standards, numerous stores that violate security and privacy regulations, and, consequently, a
proliferation of malicious apps in China’s Android market. (¶¶ 100-04.)
10. Opinion 9. Changing iOS likely would open new threats and negative security impacts.
In particular, it would entail changing security measures that Epic’s experts have identified as
significant and important. Epic’s experts propose a hypothetical world with significant negative
security impacts that they have not fully assessed. (¶¶ 105-24.)

3
11. Opinion 10. Diversifying In-App Purchase (“IAP”) channels would inhibit Apple’s anti-
fraud abilities. The App Store’s requirement for use of IAP enhances user security, including by
use of Apple’s cryptographic hardware. (¶¶ 125-29.)
12. I have 30 years of experience in the field of computer science, and specifically in Internet
and computer security.
13. I received my Ph.D. in Computer Science and Engineering from the University of
Michigan, Ann Arbor, in 1994, with a specialty in computer security and cryptographic
protocols. I have been Professor of Computer Science at Johns Hopkins University since 2003.
I am also the Technical Director of the Johns Hopkins University Information Security
Institute—the University’s focal point for research and education in information security,
assurance, and privacy. Johns Hopkins University, through the Information Security Institute’s
leadership, has been designated as a Center of Academic Excellence in Information Assurance
by the National Security Agency and leading experts in the field.
14. I also have significant industry experience. I spent six years at AT&T Labs in the Secure
Systems Research Department, where I focused on Internet and computer security. Prior to
AT&T Labs, I spent two years at Bellcore in its Cryptography and Network Security Research
group, also focusing on Internet and computer security issues. More recently, I served as the
founder and President of Independent Security Evaluators, a computer security consulting firm.
Among our responsibilities was acting as an independent testing lab for Consumer Union, which
produces the Consumer Reports magazine. For Consumer Union, I managed an annual project
where we tested popular anti-virus projects. I am also currently the founder and chief scientist of
Harbor Labs, a software and networking consulting firm specializing in medical device security
and privacy of healthcare data.
15. I serve, or have served, on several technical and editorial advisory boards. I have served
on the Editorial and Advisory Board for the International Journal of Information and Computer
Security, the Journal of Privacy Technology, and Springer’s Information Security and
Cryptographic Book Series, as well as acting as an Associate Editor of the Institute of Electrical
and Electronics Engineers’ (“IEEE”) Security and Privacy Magazine, the Association for
Computing Machinery’s (“ACM”) Transactions on Internet Technology, and the
Communications of the ACM journal. I also have served in the past as a member of the Defense
Advanced Research Projects Agency’s Information Science and Technology Study Group, a
member of the Government Infosec Science and Technology Study Group of Malicious Code,
Associate Editor of the Electronic Commerce Research Journal, Co-editor of the Electronic
Newsletter of the IEEE Technical Committee on Security and Privacy, and a member of the
board of directors of the USENIX Association (the leading academic computing systems
society).

4
III. Opinions
A. Security Encompasses Broad Categories and Must Consider the Threat

Model
1. Security Encompasses Safety, Privacy, Trustworthiness, Piracy,

Objectionable Content, and Reliability
16. While the definition will vary depending on the context, security generally refers to the
process of protecting data as well as device and system functionality.
17. For purposes of my opinions in this case, I take a broader view of security than Epic’s
experts, Dr. Lee and Dr. Mickens. In particular, Drs. Lee and Mickens appear to be almost
exclusively focused on the narrow goals of malware exclusion and exploit resistance. Malware
is typically defined as software designed to disrupt, damage, or gain unauthorized access to a
system. Exploit resistance generally refers to a system’s ability to enforce the security measures
in place (e.g., preventing users who do not have a password from accessing a password-protected
system). With this narrow focus, Drs. Lee and Mickens conclude that certain of Apple’s App
Review goals do not relate to security because they instead relate to broader subjects such as
safety, performance, business, design, and legal compliance.
18. I agree that malware exclusion and exploit resistance are important security goals. But I
take the broader view that security encompasses issues including privacy, trustworthiness, and
reliability. For example, if an app is targeted to young children, and asks the user to enter her
age and home address, that is a potential security concern. If an app says that it is a program that
allows a user to play Tic-Tac-Toe against the device, but wants to access the device’s
microphone and camera for reasons entirely unrelated to Tic-Tac-Toe, that is a potential security
concern. If an app developer scams users by falsely describing the application as one that can
detect a stroke, or by artificially inflating an app’s reviews to entice customers to purchase the
app, those also are potential security concerns. Privacy relates to protecting data from
unauthorized access or disclosure, and is intertwined with security: security and privacy are two
sides of the same coin, where security controls dictate the level of privacy enforced, and privacy
technologies can guarantee a higher degree of security.
19. Unreliable apps also expand the attack surface, or different points of entry by which
security can be attacked. If an application is supposed to remind a user to take medication at a
certain time, but constantly crashes and fails to do that, that is a potential security concern. As
this and the earlier examples demonstrate, apps are more and more frequently used in the
management of daily life and home, and unsafe or unreliable apps could have physical security
implications. Reliability and security are intertwined, because reliability ensures that security
controls work as expected. Decreases in security and reliability thus likely coincide. For
example, a software crash could indicate that a threat has been embedded in an app’s code or
otherwise create an opening for an attacker to take advantage of the cause of that crash to, for
example, transmit a custom input that redirects the app to a location containing a virus or other
malware.

5
20. Reviewing for legal compliance likewise aids in protecting security. Piracy, for
example, is a security risk, not just a content moderation issue. Pirated apps threaten users
because they are known to often contain malicious functionality. And pirated apps threaten
developers because the download of pirated apps reduces the revenue stream for the legitimate
version of those apps. Other illegal apps pose similar threats. Apps that, for example, enable
illegal gambling could exploit user information or wealth.
21. For similar reasons, reviewing for objectionable content can contribute to security. Drs.
Lee and Mickens had identified “objectionable content” review as “non-security concerns.” But
objectionable content such as pornography is frequently associated with malware. Kaspersky, a
global cybersecurity company that provides antivirus and other cybersecurity products, found
that over 25% of attacks on mobile devices come from porn-related malware. This arises in part
because apps and sites distributing pornography have fewer traditional revenue mechanisms
available to them, and thus have been found to be more likely to partner with malicious actors in
their efforts to monetize their sites. All of these examples illustrate the broader point that
security is not limited to just anti-virus scanning and exploit resistance mechanisms.
2. iOS Faces a Heightened Threat Model
22. Any evaluation of the security of iOS thus must consider its context, objectives, potential
attackers, and the manner in which users use their systems, or the iOS “threat model.” Threat
models help formalize the security risks to a system, by enumerating vulnerabilities, weaknesses,
and defects, as well as impact of their exploitation. For example, it is important to ask: What
might a potential attacker be motivated to attack, and why? What data could be targeted by the
attacker? Would an attack require sophisticated hardware? What mechanisms could the attacker
use in seeking to recover private data from an iPhone? How physically accessible is an iPhone,
and the data that it contains? How many users could be impacted? What would the potential
cost of an attack be?
23. With over 1 billion active devices, an App Store that hosts almost 2 million apps that
have been downloaded over 180 billion times, a user base that frequently downloads apps and
engages in financial transactions on the device, and camera, microphone, and GPS hardware that
follows users nearly everywhere they go, Apple iOS devices are faced with an extraordinary
threat model. iOS devices are small, portable, typically on 24/7 and kept close at hand by users
for critical tasks such as navigation and emergency calls. They are very likely to hold a user’s
financial information, personally identifiable information (“PII”), protected health information
(“PHI”), and location information. And they might be misplaced, lost, or stolen at a higher
frequency than other types of computer systems. Their convenience has led users to store more
private information on them, and users rely on their functionality. All of these reasons, however,
make iOS devices rewarding targets to attackers and thus heighten the risks they face.
24. A factor in the threat model is the sophistication of potential attackers. Malware
developers and other malicious actors continue to become more sophisticated and expend
significant resources to build malicious and other ill-intentioned apps for mobile devices.
Purplesec, a cybersecurity company, observed in DX-4956, a report published in 2020, that the
total malware infection growth rate has been increasing by hundreds of millions every year

6
(812.67 million infections reported in 2018, a significant increase from 702.06 million reported
in 2017). Among the better-resourced are Advanced Persistent Threat (“APT”) groups, which
are usually directed and supported by established nation-states to conduct cyberattacks on
government, industry, and infrastructure targets for political or economic agendas. APT groups
use malicious apps to target high-profile victims and cause long-term damage on critical assets
through, for example, spyware that collects recordings, locations, messages, and call logs from
infected users. Social engineering attacks also are on the rise—they were observed as
comprising 98% of cyber-attacks in 2020. Social engineering attacks are exploit problems
designed to manipulate users into taking actions that will disable or circumvent existing on-
device or other affirmative protections. Cybersecurity cannot be static, because malicious actors
are always seeking out new vulnerabilities in systems and ways to circumvent and evade anti-
malware and other security mechanisms.
25. In view of the wide variety of potential attacks, and the mechanisms they can use,
computing security best practices calls for the employment of layered defenses. This is true for
iOS and more generally. The concept of layered defenses is based on the understanding that the
more layers of protection that a system has, the harder the system is to exploit. Each additional
layer of protection enhances the overall security posture of the overall system. In computer
systems generally, and in iOS in particular, layered security combines access controls to limit the
exposure of data and functionality as well as additional security review or verification
mechanisms. By doing so, there is lowered risk of unwanted apps ever reaching an iOS device
and users being confronted with social engineering attacks, as well as reduced impact of
unwanted apps should they be able to make to the device.
26. Take the example of a Tic-Tac-Toe app that I mentioned earlier, where the Tic-Tac-Toe
app would access an iPhone’s microphone and camera. That app, on its face, may not be
sufficiently limited by access controls, because there may be a legitimate reason why users
playing Tic-Tac-Toe may want to talk to each other. However, additional security review and
verification mechanisms, such as Apple’s App Review process, are better positioned to evaluate
not just that there is microphone and camera access, but whether that microphone and camera
access has malicious ulterior motives. And the microphone and camera are not the only
hardware on an iOS device that could expose a user. Even an iPhone’s Bluetooth functionality,
for example, can be used to determine its user’s location—thereby potentially putting the user’s
safety and privacy at risk—if an app is permitted to poll for nearby Bluetooth devices. A thief,
for example, could develop an app that would use this Bluetooth capability to provide an alert
anytime a user was (a) far enough away from their home that the house was vulnerable to
burglary or (b) in a particular location and personally vulnerable to attack or pick-pocketing. For
this reason, Apple scrutinizes requests to use Bluetooth functionality without a legitimate reason.
27. Apple has implemented sandboxing and other on-device protections, as well as taken
additional measures to enforce user privacy by, for example, requiring users to opt in before
sharing data and implementing differential privacy techniques that de-individuates user data.
Social engineering attacks, however, are designed to manipulate users into taking actions that
avoid these protections. For example, consider a game app that, while being run, will instruct a
user to enter their social security number to proceed to the next level. Here, all the app did was
display some text content. That would be unlikely to trigger any of the security properties that

7
Dr. Mickens identifies as providing sufficient security for iOS but likely would, as I will explain,
be detected during App Review.
28. Another example of a social engineering attack is presented by the “Update your Adobe
Flash plugin” pop-up scam, where multiple malicious websites relied upon general familiarity
with the “Adobe Flash Player” program to disguise malware in a pop-up advertised as an Adobe
Flash Player update. Users who thought they were agreeing to update their Flash Player were
instead deliberately installing adware (malware that would hide on their device, show unwanted
advertisements, and/or track user behavior) or worse. Scams and phishing campaigns do not
directly subvert on-device security measures, but instead exploit the (misplaced) trust between
the user and the vehicle through which a user would download the app.
B. Apple’s App Review Provides Significant and Better Security and

Non-Security Benefits than Competitor Platforms
29. Apple recognized the importance of layered security in designing its security architecture
to meet dual goals. First, the App Review and app distribution process seeks to prevent unsafe
apps from ever reaching user devices in the first place if threats and vulnerabilities are detected.
Second, Apple’s on-device security seeks to protect against and limit any damage that can be
inflicted by an app that is installed on a device.
30. It is generally unwise to first trust users to download malicious apps, and then try to
subsequently detect malicious apps and deny giving malicious apps the permissions they might
request. Given the unpredictability of user behaviors that may lead to vulnerabilities and
exploitations, users might still keep or run a malware even though an anti-malware identifies it,
or send the piece of malware to other users without running it. It is better to prevent user devices
from ever being infected with malware—as Apple endeavors to do with iOS—rather than use on-
device solutions to attempt to identify and remove malware once it already resides on a device.
31. Apple’s App Review security layer is a necessary complement to on-device protections
such as sandboxing to prevent apps from accessing data and functionality that they should not be
permitted to access: App Review checks an app to make sure it is not requesting unnecessary
entitlements (or special access rights to other hardware or software on an iOS device), which is
often an indication of malicious behavior. Sandboxing then enforces these entitlements, only
allowing apps access to data and functionality that the user permits the app to have. Sandboxing
also prevents apps from accessing and modifying data that is written by other apps.
32. Apple’s use of the App Store as the single app distribution mechanism adds a protective
layer for iOS devices. Apple currently conducts review on every app and app update distributed
through the App Store using a combination of computer automated and manual human review.
By centralizing the distribution of apps and prohibiting the distribution of apps outside of the
App Store, Apple forces any would-be attacker to navigate this additional layer of defense.
Without exclusive app distribution, apps could bypass the App Review layer of defense.
33. Apple’s app distribution layer provides additional protections even if App Review does
not catch a breach in the first instance. Where apps are all distributed through the App Store, if
an untrustworthy or malicious app is identified, Apple can pull this app (as well as other similar
8
ones) from the App Store to stop further distribution, or, as Steve Jobs put it, “turn off the
spigot” so no more people can download that app (or similar ones).
34. Permitting alternative third-party app stores would increase instabilities and security risks
for iOS. As I’ll explain, alternative platforms that permit apps to be downloaded outside a
reviewed distribution channel have demonstrably greater security vulnerabilities and concerns.
Also instructive are the limited scenarios in which Apple permits apps to be downloaded outside
of the App Store; Apple has observed efforts to take advantage of those scenarios to circumvent
security requirements.
35. Epic and its experts characterize Apple’s App Review process as generally ineffective,
easily replicable by third parties, and providing minimal security benefits beyond what on-device
security mechanisms already provide. As noted above, they reach this conclusion based on a
narrow definition of what constitutes a security risk. And in any event, none of these criticisms
are true. Apple’s App Review constitutes a critical component of Apple’s layered security that
provides significant and effective security protections that enhance and improve the security of
iOS.
1. Apple’s App Review Process Provides Significant and Comprehensive

Security Protections That Complement On-Device Protections in
Providing a Safe and Trustworthy iOS Platform
36. Apple’s App Review is necessary to complement Apple’s iOS on-device security in
protecting against malicious apps. App Review, and particularly its review by humans of every
app and app update approved for distribution through the App Store, provides comprehensive
advantages in identifying and mitigating malicious and untrustworthy activities within an app. In
other words, conducting App Review means that iOS apps are more likely to function as
disclosed to users and are less likely to contain malware, undisclosed and unintended features,
and other unsafe features. For the examples I provided earlier, and for other types of
untrustworthy or malicious apps, on-device security alone would not provide sufficient
protection.
37. App Review is better positioned than on-device security mechanisms to evaluate user-
generated content and determine whether entitlement access is being requested for a legitimate or
ulterior motive. Apple’s App Review, for example, is better able to detect social engineering
attacks like a game app that instructs users to enter private information like their social security
numbers. App Review also is better able to flag the Tic-Tac-Toe app that will act as spyware via
its access to an iPhone’s microphone and camera. A software running on the device will not find
it suspicious that a Tic-Tac-Toe app should want access to the microphone and camera. App
Review is better positioned than on-device security mechanisms to detect and prevent adware.
App Review also is better positioned to determine whether an app contains hidden and
undisclosed behavior, or will do everything it promises to do. It is difficult for a machine to
know whether an app purportedly about renaissance art actually contains such content.
38. The human review component of Apple’s App Review is critical; its combination with
the static and dynamic analysis performed by Apple’s proprietary computer systems
differentiates and strengthens iOS security. Human reviewers are better able than computers to
9
accurately assess, for example, whether user-generated content is offensive, violates restrictions
on content in apps for children, constitutes false or misleading content, or seeks information in
violation of privacy guidelines. Human review also is critical in light of the “halting problem”
faced by computer tools, where it is difficult for computer programs to determine when earlier
programs will “halt,” or terminate. Apps are comprised of multiple computer programs, where
each program will be “called” to run in turn. But the next-in-line program does not know when
it will begin to operate until it is called; that program is not, unlike a human, able to take a global
look at the overall app architecture or predict future behavior. For this reason, computer tools
may have a more difficult time determining whether an app has hidden a problematic feature in a
program, especially when the malicious components had not been previously identified. For
example, in the game app I mentioned earlier, users were instructed to enter their social security
number in a malicious game app. If the malicious component is novel or obfuscated and hidden
in a certain way, App Review computer tools or on-device protections might not detect it. By
contrast, human reviewers might detect the issue because they would see the text content as part
of the app’s overall media content.
39. Humans are also better equipped to evaluate the motivations for an app that instructs its
users to actively restart their device or disable device security features, which, in turn, could
expose users to uncontrolled or untrusted apps and code or permit a device to run undisclosed
operations in the background. An example of this type of threat is cryptojacking malware, or
malware that harnesses its compromised device’s resources to run background operations that
mine cryptocurrency. The question of whether cryptomining is authorized can be context-
dependent, which makes it difficult for traditional anti-malware to determine whether an app
mining cryptocurrency is doing so with the user’s consent and knowledge, or as a result of
cryptojacking. Absent Apple’s App Review process, it would be up to the users to decide
whether to grant potentially malicious apps access to the camera, network, microphone, etc. or
decide whether to disable certain on-device security features that the app claims will affect its
operation. App Review is critical to maintaining the users’ sense of trust and preventing users
from having to make difficult cybersecurity-based decisions.
40. App review’s human review component also means that it is better positioned than on-
device mechanisms to detect new types of issues and threats. Computer tools rely upon
heuristics—or algorithms that are written by humans to define patterns or other indicia of
potentially unsafe activity—but are also fundamentally limited by that reliance. Heuristics are
well-suited for identifying already-known threats and performing repetitive analysis, but are less
able to detect something that was not previously known and therefore not addressed in the
existing algorithms.
41. Apple’s App Review process benefits from a significant number of resources that
enhance its accuracy and efficiency. First, App Review process benefits from its proprietary
machine learning and other computer tools that it has developed for the app review process.
Apple has committed to and invested heavily in the development of these tools, which are used
internally by Apple and are not available to third-party reviewers.
42. Second, Apple’s App Review process benefits from the information that is has garnered
through over a decade of app review. Such accumulated information—all of which is held as
proprietary—includes a large internal corpus of previously identified threats associated with
10
rejected malicious apps, including suspicious keywords, malicious IP addresses and URLs, and
information used to determine whether an app may be pirated. This information, which is
developed in connection with the human review that Apple conducts of every app and app
update approved for distribution through the App Store and Apple’s underlying knowledge of the
iOS ecosystem, has multiple purposes. Apple utilizes this data to train its machine learning
algorithms to better detect malicious and otherwise problematic apps. Apple also utilizes this
information internally to improve iOS security.
43. Third, Apple’s knowledge of its own iOS ecosystem and operation and their security
vulnerabilities, as well as its private APIs, plays a critical role in Apple’s App Review. Apple’s
internal knowledge of the detailed operation of iOS, its potential vulnerabilities and
development, private APIs, and first-party entitlements are highly associated with security
concerns. APIs, for example, are interfaces through which an app can access system functions.
Apple has created a list of private APIs and determined that they should be used by Apple only
for its internal purposes. Such information, if subverted, can grant bad actors heightened
privileges to perform powerful security attacks on iOS devices. For these reasons, disclosure of
confidential and proprietary information about iOS and iPhone hardware can create significant
security risks. Restricting third-party developers’ access to Apple’s private APIs helps ensure
that apps cannot interfere with the core functions and stability of iOS.
2. iOS Is Safer than Other Operating System Platforms
44. I disagree with the conclusion of Drs. Lee and Mickens that other platforms, such as
Windows and Android, offer “rough parity” in their security protections to those of iOS. A
number of objective third-party sources indicate that iOS is safer. These third-party sources
reinforce my own conclusion that the particular attributes of Apple’s App Review render iOS a
safer and more trustworthy platform.
a) Third-Party Analyses Demonstrate that iOS Is Safer than

Other Platforms
45. Non-iOS mobile devices historically have been the victims of malware far more often
than iOS devices. For example, the National Vulnerability Database (“NVD”), maintained by
the National Institute of Standards and Technology (“NIST”), collects Common Vulnerabilities
and Exposures (“CVEs”) reported for various software platforms. A CVE is a broad term that
refers to computer security flaws and includes, for example, authorization bypass efforts,
reliability problems such as buffer overflows, the exposure of sensitive information to
unauthorized actors, improper privilege management, and other flaws in software, hardware, or
computer components that can be exploited and negatively impact confidentiality, integrity, or
the availability of that component. The NVD’s CVEs are considered by the industry to provide
indicia of a platform’s relative security or trustworthiness because they provide a standardized
identifier for given security flaws and enables the assessment of information across multiple
platforms. The NVD acts as the U.S. government repository of standards-based vulnerability
management data and is recognized by the industry as a leading source of software vulnerability
data. As shown in DX-4962, DX-4969, and DX-4966, the chart below reflects the number of
CVEs on iOS compared to the numbers on Android and Debian Linux operating systems in

11
Android on metrics of perceived platform quality and user satisfaction because of Apple’s
diligence in ensuring app quality.
50. The Windows/PC platform also faces more vulnerabilities than iOS. In Nokia’s study,
Windows/PC devices accounted for an even higher percentage of infected devices than Android,
at 38.92% in 2020. And Windows has a history of being victim to very damaging malware
attacks. In 2017, for example, the NotPetya ransomware caused an estimated $10 billion in
damages and disrupted the global supply chain by decommissioning a significant percentage of
shipping ports employed by a certain shipping entity.
51. Throughout the App Store’s existence, there have been incidences of new malware
finding its way onto the App Store. Examples like these have been used by Epic and its experts
to suggest that Apple’s App Review fails to provide meaningful protections to users. Simply
put, these cases do not show the lack of necessity for the App Review process. The third-party
analyses that I just discussed, as well as Apple’s internal statistics recording how many apps are
rejected or otherwise taken down because of privacy violations, the introduction of hidden
features, and obfuscation techniques (over 150,000 for privacy violations alone in 2020), among
others, show how distorted Epic’s experts’ claims are. RiskIQ recognized that Apple’s App
Store is one of the fastest growing app stores, with 465,676 new apps observed in 2019, but
described it as “Fort Knox…. [because] it rarely hosts dangerous apps.” And, more generally,
the App Store hosts almost 2 million apps that have been downloaded approximately 180 billion
times, collectively. Apple reviews approximately 100,000 App Store submissions per week,
which annualizes to more than five million apps and app updates per year. Even if 100 bad apps
made it through the App Review process and became available on the App Store, that would still
be less than 0.01 percent of the apps on the App Store and an even smaller percentage of the apps
and app updates reviewed by App Review. And, importantly, Apple’s App Review process is
ongoing—it will continue to monitor and take down apps that are already available in the App
Store, if it determines those apps are untrustworthy or malicious, and it modifies its App Review
process when it learns of such apps to prevent similar apps from getting through App Review in
the future. This evidences the relentlessness of threats and danger of degrading security.
52. For similar reasons, marketplace payout levels for “zero day vulnerabilities” are not
instructive in evaluating whether Apple’s App Review provides security benefits (or better or
worse security benefits than those offered on Android devices). According to Dr. Mickens, the
higher payouts offered for Android by Zerodium, one zero-day acquisition firm, demonstrate that
iOS phones are easier to compromise than Android phones. But as Dr. Mickens himself admits,
market data must be interpreted with a degree of skepticism. There can be many reasons for the
varying levels of iOS and Android payouts including, as Dr. Mickens notes, the possibility that
demand for iOS exploits is now depressed by the fact that the bounties for iOS vulnerabilities
used to be higher than those for Android. Also possible is that Zerodium may offer a higher
payout for Android because of Android’s adaptability to various types of infrastructure and
hardware (while iOS is intended for Apple devices only). I note as well that there are multiple
zero-day acquisition firms, including government organizations and hacker groups operating in
less visible, less trackable marketplaces, that offer different (and higher) payouts for iOS than
Android. The amount of a payout simply cannot be assumed to directly correlate with level of
security offered on a platform.

13
b) Apple’s App Review Renders iOS a Safer and More

Trustworthy Platform
53. The results of these third-party studies are consistent with and reinforce my conclusion
that the iOS’s layered defenses, and particularly its inclusion of the App Review layer, make iOS
a safer and less vulnerable platform than competitor platforms. Windows, Android, and non-
Android variants of Linux (shortened as “Linux”) have much more lenient app distribution
models, and depend on on-device security and third-party anti-malware solutions to then find
solutions for the malware once it is already on device.
54. The Google Play Store, and the Android mobile platform more generally, provide a
particularly instructive comparison because they permit multiple ways to circumvent any app
review like Apple’s App Review process. Unlike iOS, Android permits the installation of apps
from multiple sources, including third-party stores, sideloading, and preloading by OEMs.
Android also maintains fewer authorization mechanisms; it does not, for example, require apps
to be signed with certificates obtained from Google or another principal authority. Android’s
official documentation recognizes third-party stores and sideloading as legitimate “[a]lternative
distribution options.” As such, the developer identity associated with sideloaded apps are not
checked so that there is no deterrence for malicious Android app distribution via sideloading.
Sideloading introduces security risks to users: it allows the installation of unreviewed apps that
might install malware or otherwise might grant themselves entitlements to a broad array of
hardware and software in order to, for example, access privileged functionality without alerting
the user. These unreviewed apps also could be pirated or otherwise entail intellectual property
violations. Additional security risks are posed because it is almost impossible to effectively limit
further distribution of an already-identified malicious or vulnerable app, which can be
downloaded through such a wide variety of means. It also is difficult to keep sideloaded apps
up-to-date and secure.
55. These design choices stand in stark contrast to Apple’s App Review and centralized app
distribution security layers and, as I just explained, lead to significantly higher infections and
other vulnerabilities in Android than in iOS. Apps that are sideloaded onto an Android device
may undergo no app review at all and, for this reason, could contain any manner of malware or
spyware. This happened in 2019, for example, when spyware was distributed through sideloaded
Android apps that advertised themselves as Evernote, Google Play, and other massively popular
apps, but instead were “fake apps” that performed eavesdropping, screen recording, and
password collection from user devices.
56. The presence of third-party app stores containing unreviewed apps thus has the potential
to significantly diminish the overall security of a platform. And this applies not just to pre-
distribution app review, but also to apps after they become available for distribution in a store.
Apple continues its App Review process even after an app becomes available for distribution in
the App Store and will take action against apps that exhibit malicious or user-unfriendly
behaviors after they have become available in the App Store. Apple is able to do all of these
things because of its App Review and control of all aspects of its platform. Android, however,
does not and cannot. iOS’s ban on the installation of unsigned, untrusted apps enhances iOS
security relative to Android.

14
policy also does not restrict the use for call and SMS-centric apps for user data collection or
spam, unlike Apple’s Guidelines. This guideline can have significant consequences for a
potential user data breach or malicious blocking of legitimate calls and messages to a user—but
would not be enforced if it does not exist.
60. Windows similarly demonstrates the value of app review. Windows has three main
issues not present on iOS that hurt its overall security posture of the platform. First, a sizeable
portion of Windows computers do not have hardware-backed cryptographic modules and thus
Windows OS cannot use these mechanisms for secure key storage and disk encryption across
large swaths of its user base. Second, Windows allows for installation of device drivers to allow
the operating system to interface with a wide variety of hardware. Malicious or vulnerable
device drivers can be used to subvert sandboxing and also to perform other malicious behavior
with elevated privileges. Third, Windows supports the installation of software applications from
untrusted sources. This has resulted in many malware attacks. In a common scenario, a user
receives a malicious email attachment or file download. They execute it and ignore warnings
from User Account Control (“UAC”). The user's computer is then infected with malware.
61. Indeed, in the Windows Vista operating system, it became common that users would
arbitrarily run malicious programs with administrator privileges using UAC. Microsoft has
recognized and warned users about these security risks. But iOS, in contrast to Microsoft, can do
more than warn users that using administrative credentials to run an unknown program could
render a computer vulnerable to attack. App Review specifically checks for and will reject apps
that maliciously requests elevated privileges.
3. The Enterprise and Ad Hoc Distribution Programs Do Not

Demonstrate that App Review Is Not Needed
62. Apple provides two distribution options through which apps may not be reviewed by
Apple’s App Review—the Developer Enterprise program and the Ad Hoc program. Dr.
Mickens points to these scenarios as acknowledgement by Apple that App Review is not
necessary for security because iOS on-device mechanisms are sufficient to keep users safe. I
disagree for several reasons.
63. First, Dr. Mickens’s argument ignores the most important aspect of the Enterprise
Program: it is intended for use only for company-specific apps. As the name implies, the
Enterprise program allows a business enterprise to distribute apps to the company’s employees.
The fact that the apps are created by an employer, and distributed to employees (typically for
free), creates a specific security context that does not exist when third parties are providing apps
to strangers (sometimes for money) with whom they have had no prior or existing relationship.
Far from proving Dr. Mickens’s point that the downloading of an app can be “decoupled” from
the verification of the signature on an app, the Enterprise Program merely demonstrates that
when there is a relationship such that an employee trusts his or her employer, Apple permits apps
from that employer to bypass App Review.
64. This similarly applies in Ad Hoc distribution, which allows a developer to test an app on
up to 100 devices. This program allows a developer to test an app on different devices (e.g., iPad
Air, iPhone X, and iPhone 12) and on different versions of Apple’s iOS operating system to
16
ensure compatibility. Before apps can run via Ad hoc distribution, their developer must select a
valid provisioning profile and compile the app through Xcode. The provisioning profile includes
a development certificate used to sign the app. Similarly, to run an in-development app on a
developer’s devices, the app needs to run on the developer’s device with a provisioning profile
downloaded through the developer’s Apple Developer account, which also includes a
development certificate used for code signing.
65. In short, the distribution scales in both circumstances are extremely limited. Enterprise
apps are intended for distribution only to users affiliated with their enterprise, and Ad hoc
distribution is limited to up to 100 devices that must be specifically identified and registered.
66. Second, despite the limited distribution and the different security context in which these
programs operate, there have been several well-known problematic apps distributed through non-
reviewed channels, including malware by eSurv and Hacking Team. There are entities that have
been revealed as attempting to deliberately utilize the Enterprise program in a manner that would
violate Apple’s App Store Review Guidelines.
67. Epic is one such entity. It apparently explored the possibility of using the Enterprise
program in a manner that certain Epic employees flagged was contrary to Apple’s terms and
conditions. Specifically, Epic explored using Apple’s Enterprise certificates in order to bypass
the App Store review process so that Epic could more conveniently distribute its apps
(essentially via sideloading). This appeared to be financially motivated, as Epic was “looking at
ways to reduce the 30% cut that Apple take[s].” DX-4066.002. Epic not only knew that “using
an Enterprise account for external distribution like this [was] firmly against Apple’s T&C [terms
and conditions],” DX-4066.002, it was also aware that “[c]ertain malware in the past has been
able to do this [to bypass app review].” DX-4616.002.
68. Notably, the breaches identified above occurred when the distribution was not between
an employer to employer, nor from an app developer to a small group of testers. Instead, they
occurred when developers attempted to distribute and obtain apps outside of the App Review
process on a broader basis. In other words, they occurred when developers attempted to use the
Enterprise Program to bypass App Review for broad distribution.
69. In addition, even if developers are well-intentioned, if certain quality checks are not
performed because the app has not undergone App Review, broken or unintended functionality
can result.
4. macOS Does Not Prove that App Review Is Not Needed
70. Similar to his argument with respect to the Enterprise Program, Dr. Mickens also argues
that the ways in which Apple permits app distribution on macOS computers means that App
Review is not necessary for meeting Apple’s safety standards. I disagree with this argument as
well.
71. As I mentioned earlier, the threat model is critical for assessing security for a computing
system. iOS, quite simply, faces an extraordinary threat model that is different from—and poses
a greater spectrum of risks than—macOS. For one thing, iOS devices contain highly sensitive

17
personal information—often more sensitive than that stored on a computer. iOS devices, unlike
macOS devices, have sensor hardware—such as GPS units and accelerometers—that track their
users’ location and movement patterns and generate data about their users’ current physical
environment. iOS devices holding this sensitive data, along with financial data, are smaller than
macOS devices and—with their microphones and cameras—are more likely than macOS devices
to be with their users at all times. And for iOS devices, an unstable app that causes an iPhone to
crash could have devastating consequences; picture, for example, an app that causes your iPhone
to crash and be unable to make phone calls when you get a flat tire in the middle of the night.
Heightened security systems—and particularly one that includes App Review’s review for
entitlements, hidden features, requests to access sensitive hardware that are unconnected to the
app’s purpose, and other threats that could cause an iPhone to crash, among other things—are
thus reasonable and necessitated for iOS in a way that they are not for macOS.
72. In the opposite direction, macOS users may expect and require greater access privileges
than iOS users. Historically, computer users have held a higher level of privilege in order to
perform system/network administration tasks, virtualization, and software development
(including for software that will interact with peripherals such as USB devices). To connect a
printer to a computer in order to print a term paper or a novel written on that computer, for
example, a user typically requires administrative privileges in order to download and install the
drivers for a printer. Regardless of whether macOS and iOS share the same kernel, they do not
share the same threat model—and that threat model is critical to evaluating whether App Review
is necessary for security.
73. I also note that even with respect to macOS, Dr. Mickens acknowledges that one of the
app distribution models entails app review. For others, Dr. Mickens points to anti-malware
scanning as contributing to the safety of macOS. macOS devices, however, are better positioned
to conduct on-device anti-malware scanning than iOS devices. Anti-malware technologies like
macOS’s XProtect usually contain a real-time scanner, which continually monitors system
activities for the presence of malware, and an on-demand scanner. A hook to the operating
system alerts the real-time scanner when a file is executed, allowing the scanner to check a file
for malware signatures or behaviors. Anti-malware scanners also contain an on-demand scanner,
which could be run by the device user at a specified time or interval to check an arbitrary storage
location for malware. The arbitrary location may include certain files, folders, or the contents of
an entire hard drive.
74. Anti-malware software is able to operate on macOS because macOS generally provides
its users with the elevated privileges needed to perform malware scanning. In iOS, those
privileges are limited for the reasons explained above and for reasons that Dr. Mickens argues
are necessary for providing security on iOS devices—the sandbox compliance that Dr. Mickens
identifies as one of the three security properties are entirely enforced by an OS and important to
security. In other words, the iOS sandboxing protections that Dr. Mickens identifies as
important would prevent anti-malware from operating on apps distributed by third-party app
stores. It is for these reasons that Apple currently relies on a preemptive automated scanning
process during the App Review process in order to detect whether iOS apps are harmful before
they become available for distribution through the App Store. Apple also continues to scan apps
that have already been uploaded to the App Store and will remove them if malware is detected.

18
Apple has a number of tools in place to detect malware on existing apps, that it runs at periodic
intervals to capture content at different times.
75. I also should note that, even if anti-malware scanning could occur in iOS, it could be less
effective against the types of threats that are prevalent on mobile devices. In an industry study of
threats to mobile devices, adware was identified as the most prevalent threat to mobile devices,
at 48.02%, and riskware the second most prevalent, at 20.14%. Riskware is defined as legitimate
programs that pose potential risks due to security vulnerability, software incompatibility, or legal
violations, where malicious actors can take advantage of these programs to access and steal
sensitive data or admin-level processes. By contrast, and for context, adware and riskware only
make up 7.7% and 3.8% of portable executable threats on the Windows operating system. Even
if current anti-malware would hypothetically be applied to iOS, the efficacy could not be
determined.
76. The problems that would arise if iOS adopted the “macOS app distribution models” that
Dr. Mickens identifies can be previewed by examination of cases of jailbroken iOS devices.
Jailbroken iOS devices, like Dr. Mickens’s “macOS app distribution models,” allow for the
download of apps outside the App Store. Jailbreaking refers to a process that modifies Apple’s
iOS operating system to enable the installation of unauthorized software, including applications
from other interfaces, that are not approved by an app review process (like sideloading). It has
been well-documented, however, that jailbroken iOS devices suffer from more malware than
non-jailbroken iOS devices. Malware can be distributed via unreviewed apps and, moreover, can
use elevated privilege levels possessed in light of their lack of review to perform malicious
activity. Dr. Mickens’s “macOS app distribution model” proposal would, however, move iOS
towards a “universal” jailbroken iOS phone and the greater exposure to threats that this would
entail.
C. The Introduction of Alternative App Stores Would Decrease the Security,

Safety, Reliability and Trustworthiness of the iOS Platform
77. Dr. Lee suggests that third-party app stores could achieve the same security goal as the
App Store, and Dr. Mickens suggests that, in the event of third-party app stores, iOS users would
not see diminished security. I disagree with that conclusion.
78. The introduction of third-party app stores for iOS would decrease iOS security, safety,
and trustworthiness, as evidenced by the cases of Google and statistics indicating that third-party
app stores host 99.9% of discovered mobile malware. DX-4956.004. Irrespective of whether
they would be able to or intend to achieve the same security goals, the reality is that they could
not. Moreover, there is no guarantee that all, or even most, third-party app stores would commit
to upholding user security and privacy and intend to achieve such security goals, particularly if
those standards come at the expense of efficiency and revenue.
79. In fact, I understand that Mr. Sweeney would not anticipate third-party stores to adhere to
security standards; I understand that he believes that app stores can make decisions about the
quality and other attributes of the apps they will distribute. In other words, Mr. Sweeney does not
consider it to be a problem that third-party app stores give up security controls for time and cost

19
efficiency, as long as the third-party app stores think the surrendered security controls are not
necessary, without thinking about the security implication behind this situation.
1. Third-Party App Stores May Not Have the Incentives to Provide as

Secure an App Store Experience as Apple
80. Epic’s experts suggest that even if app distribution were opened up, the security of the
iOS platform would remain uncompromised because at least some third-party app stores
theoretically could engage in the app review process as Apple is currently doing.
81. To begin with, neither of Epic’s experts opine that all third-party stores would even
attempt to engage in the app review process that Apple currently undertakes. They only
speculate that some of those stores might attempt to emulate Apple’s processes.
82. Outside of the iOS platform, we know that there currently are distribution sites that
specifically traffic in the types of apps—such as pirated apps—that Apple prohibits. If permitted
to operate and distribute iOS apps, these stores would have no incentive, and are unlikely to
attempt, to duplicate Apple’s app review efforts. Even third parties that don’t explicitly traffic in
illegal and malicious content are unlikely to match Apple’s App Review efforts for several
reasons.
83. First, many third-party stores, including those that target niche markets, lack the
resources or commitment to review apps in the way that Apple does. Apple has outpaced its
competitors in protecting user privacy. As noted above, Apple’s process is a comprehensive one
that includes not only human reviewers, but teams of dedicated engineers who create tools,
including machine-learning tools, specifically to help Apple combat efforts to subvert the app
review process.
84. Second, the incentives of third-party stores may drive them to deliberately adopt a
standard lower than Apple’s. For example, certain large companies are heavily dependent on ad
revenue, which in turn, is heavily dependent on the ability of an app to track user behavior.
Other companies may choose to maintain different standards. For example, Apple’s App Store
Review Guidelines reject apps that alter or disable standard device inputs like device volume
buttons, but the Google Play Store’s Developer Program Policy does not have a similar
requirement. Returning to the Tic-Tac-Toe app example, if it also included an instruction to
users to reconfigure their devices to raise the volume on the microphone to enhance listening
sensitivity, Google might allow that app for distribution via the Google Play Store where Apple
might not. As mentioned earlier, Google similarly does not have guidelines pertinent to privacy
protection such as the Apple App Store Review Guidelines that limit background activity to
specific functions and restrict the calling of and collection of SMS data. Another example is the
GOG app store, which operates on PCs and has a business model of restoring old, unworkable,
or unoptimized games. Because GOG’s purpose is to make unworkable games work again, not
to provide secure apps, it likely prioritizes security less than the App Store.
85. Third, even if a third-party store has the best of intents, practical consequences such as
meeting a deadline to release a product may result in them “bypassing” certain review guidelines
on a one-off basis. For example, the videogame Cyberpunk 2077 recently was released with a
20
considerable number of bugs and glitches. Sony had to pull Cyberpunk 2077 from the
PlayStation Store and offer users full refunds because the game was unplayable on certain
PlayStation consoles. Cyberpunk 2077 is an example of a game that was released at full price
before it had been sufficiently debugged, and it demonstrates how developers may prioritize
profit over reliability. I raise this example not to impugn this specific product, but to suggest that
when companies are faced with deadlines and marketing campaigns, those pressures may be at
odds with even well-intentioned developers and app review guidelines.
86. Even Epic’s own CEO and Vice President of the Epic Games Store recognize that third-
party app stores could hold differing incentives. Steven Allison, Epic’s Vice President of the
Epic Games Store (“EGS”), will testify that EGS, unlike Steam, would not support certain kinds
of content like anti-games, which are offensive and sexual in orientation and will be curated
based on standards against porn and hate. As Mr. Sweeney will testify, Epic curates its own Epic
Games Store and recognizes that app stores can present customers with different quality,
selection, and policies.
2. iOS Security Would Be Impaired if Even One Third-Party App Store

Does Not, or Chooses Not to, Maintain Strong Security Measures
87. In the area of computer security, it is generally understood that “security is only as strong
as the weakest link.” What this means is that in a fragmented distribution landscape, bad apps
need to find only one app store with less than adequate security measures in order to jeopardize
overall safety of iOS and infiltrate the iOS ecosystem. If there were nine stores, an attacker
could simply submit his app to all nine stores, and as long as one store were to accept the app, it
would then become available to the public. The scenario is not limited to intentional attacks. If
an app unintentionally contains a bug that causes it to crash, it also would be available to the
public as long as one of the stores accepted the app. Third-party app stores thus would increase
the attack surface of iOS and weaken its overall security.
The situation becomes more dire when one considers that not all stores will even pretend to care
about security. Stores referred to as “rogue app stores” are known to deliberately host and
distribute pirated content and apps containing malware and user data theft. In China, Android
app stores are known to violate security and privacy regulations, with nearly 35 percent of
Android apps secretly stealing user data unrelated to their functionality. These third-party stores
either choose not to or are not capable of enforcing security and privacy guidelines like Apple’s
and the presence of even one such store on iOS could significantly diminish iOS’s overall
security.
88. The presence of third-party app stores thus has the potential to significantly diminish the
overall security of Apple’s App Store (and the iOS platform). When Apple discovers malicious
apps that have successfully circumvented the App Store review process, it adjusts the review
process to prevent such apps from successfully being approved in the future. If Apple discovers
apps that have not circumvented the App Store review process per se but that are exhibiting
malicious or user-unfriendly behaviors after installation, Apple similarly adjusts its processes to
prevent this from reoccurring. If Apple discovers new malware on the iOS platform, it can
explicitly adjust its custom-written malware scanners to scan apps already on the App Store and
to detect such malware in the future. Apple is able to do all of these things because it controls all
21
aspects of its platform. Drs. Mickens and Lee are not only wrong to opine that Apple’s App
Store review process adds only marginal security benefits at best, but are also wrong to opine
that third-party app stores can provide equivalent security to Apple’s app store review process.
89. Furthermore, the fragmentation proposed by Epic’s experts would limit Apple’s ability to
deter or otherwise take action against such attackers. Where Apple’s App Store acts as the only
way to distribute apps, Apple could freeze or terminate that attacker’s account in order to prevent
them from any further legitimate distribution of apps and app updates that violate Apple’s
Guidelines, as well as check for similar issues in other apps. In the world proposed by Epic, iOS
would, like Android, no longer have a single “spigot” of app distribution and therefore be less
able to prevent the distribution of known malicious apps on iOS.
3. The Presence of Third-Party App Stores Would Weaken the Security

Provided by Apple’s App Review as well as by Well-Meaning Third
Parties
90. The centralization of iOS and App Store distribution not only prevents multiple stores
from acting as the “weakest link,” but it also strengthens Apple’s ability to implement measures
to protect iOS security. By contrast, weakened protections can lead to the deterioration of user
trust in iOS, which might lead to more severe risk scenarios in which users refuse to apply
Apple-recommended security measures such as downloading software updates.
91. Given Apple’s centralized iOS app distribution channel, Apple’s App Review has
transparency over all apps distributed through the App Store, which covers the absolute majority
of apps utilized by iOS devices. Apple’s App Review process thus has and relies on a catalog of
historic review decisions Apple has made with respect to tens of millions of apps, which is used
to continuously update Apple’s tools and serves as a resource to train and educate Apple’s
human reviewers. Apple has made numerous choices intended to present a safe, reliable, and
trustworthy app experience, including all of these as well as restrictions on the addition and
removal of apps that permit users to remove preinstalled apps. Apple also takes the
responsibility to eliminate “weakest links” for iOS app security and app distribution to better
protect its users, instead of forcing individual users to identify secure, trustworthy app
marketplaces on their own, just to protect themselves. Indeed, Epic’s experts do not dispute that
curation of apps provides additional layers through which security and privacy can be protected.
Approving low-quality apps that pose security and trust risks would, for example, degrade iOS
users’ experience and likely cause a loss of goodwill with respect to users and, ultimately, the
attractiveness of Apple’s app experience to developers.
92. Apple’s catalog of review information would see reduced effectiveness if third-party app
stores were allowed to approve and distribute their apps on iOS. I mentioned earlier the “Update
your Adobe Flash Player” plugin scam, where multiple websites would display a pop-up
intended to get a user to click on and install adware or other types of malware. In the app
context, where all apps are reviewed by Apple for distribution through the App Store, Apple
would be able to catalog, detect, and block each instance of apps containing the website links
containing the malware plugin. Where, however, third-party app stores also could distribute
their apps on iOS, the likelihood increases that one or more of those other app stores might not
block an app pointing to an “Update your Adobe Flash Player” malware plugin, perhaps because
22
that app store had not been able to catalog all of the website links associated with the malware
plugin. And this would be a problem not just for malware websites, but also new types of
malware, new trends in malicious apps, new methods of social engineering attacks, and other red
flag information might not make it into the databases for Apple and all other third-party app
stores distributing apps on iOS.
93. This could pose problems not just for Apple’s App Review, but for all third-party stores
endeavoring to enforce security guidelines on iOS. This would effectively render all app review
conducted by all of the app stores less useful over time. With multiple third-party app stores, the
Apple App Store, and each third-party app store, would see only a subset of all existing data—
such as a subset of malware signatures—resulting from the malicious apps that were caught in
their respective app review process. By allowing multiple app stores, the app stores all would
become less effective as key information becomes decentralized, leaving app stores with
fragmented knowledge of how the iOS platform is operating. Whether Apple was able to
enforce security standards or not, and even where security breaches occurred because of a third-
party app store’s ineffective review process, users might attribute those security breaches to the
iOS platform, and, by consequence, Apple. This ultimately would risk users’ safety and erode
the trust of the iOS platform.
94. The “Cuphead” game provides a real-life example where fragmentation permitted an app
that otherwise would have been caught during App Review to make it through (temporarily).
Cuphead is a game that was available on the PC third-party app store and gaming platform
Steam. Developers then created a copycat of Cuphead and submitted it to the App Store with
screenshots from the legitimate game and under a name similar to that of the Cuphead developer,
StudioMDHR. Apple ultimately pulled the app from the App Store after StudioMDHR reported
the app as an imposter to Apple.
4. Users Have Limited Ability to “Choose” between Safe and Unsafe

Third-Party Stores
95. Epic’s experts seem to opine that even if some app distribution stores will have lower (or
no) security standards, that would not degrade iOS security and trustworthiness because
individual users can make informed decisions about which stores to shop at, and which apps to
download. This assumption is problematic for several reasons.
96. First, the general user population may lack the security-oriented technical
understanding—and the incentive to gain such understanding—needed to make accurate
decisions regarding security and privacy. For these reasons, social engineering attacks, spyware
operating in the background, and cryptojacking—particularly those that circumvent app review
channels—have been particularly successful. With social engineering as an example, DX-4956,
a report published by cybersecurity company Purplesec in 2020 identified that “98% of cyber-
attacks rely on social engineering” and that the total malware infection growth rate has been
increasing by hundreds of millions every year (812.67 million infections reported in 2018, a
significant increase from 702.06 million reported in 2017). Similarly, 9Game.com is still the
largest mobile game market platform in India despite being considered by RiskIQ as the “most
dangerous” app store with 61,669 Android apps associated with potential security threats—the
highest concentration of potentially insecure Android apps. DX-4401.005.
23
97. Even if some users have a certain degree of technical sophistication, it is sometimes still
very hard for them to identify that they are under attack, let along perform attribution. After all,
one goal of attackers is to keep the victim from realizing that an attack has occurred; this allows
the attacker to maximize the number of victims as well as the “payout” before the attack is
discovered and perhaps shut down.
98. Epic’s position also relies on the assumption that a user has a realistic choice among
multiple app stores. As the PC and Chinese Android marketplaces show, and as Mr. Federighi
will testify, this is not a realistic assumption outside of the iOS App Store. Users may not always
have control over what software they install, and if the software is not available in the App Store,
users may have to go to a less secure app store in order to obtain that software. This will, as I
explained above, open up iOS to the “weakest link” and make iOS users less secure.
99. Indeed, if a third-party app store provides exclusive content that is only available on its
platform, users might have no choice but to use that store regardless of whether it is secure or
reliable. Exclusivity, for example, has been one of Epic’s most prominent business strategies.
Epic has heavily leveraged content exclusivity by making certain content only available through
the Epic Games Store and Mr. Sweeney has confirmed that Epic seeks to negotiate content
exclusivity deals that prevent Steam from having the same content that is available in the Epic
Games Store.
5. The Case Study of the Android Marketplace in China Demonstrates

the Security Problems that Arise From Fragmentation
100. The various concerns I raise here are not merely theoretical. The Android marketplace in
China illustrates the real-life consequences of fragmentation of app distribution.
101. As Dr. Evans explains, the Android operating system for mobile phones is created and
maintained by Google. Throughout the world, Google operates the Google Play app store
through which many developers choose to distribute their Android apps. Apps distributed
through the Google Play Store are reviewed by computers as well as, sometimes, humans, in
accordance with Google’s Developer Program Policy. But developers also have the option of
distributing their apps outside of the Google Play store for Android devices.
102. In China, the Google Play Store is banned. And in place of a centralized Google Play
Store, a variety of other app distribution stores have popped up instead. The multitude of
Android app stores in China has not improved Android security in that country. To the contrary,
China faces a very significant and frequent risk of malware arising from the numerous Android
app stores that are available to Chinese users—and the multitude of those app stores has
facilitated that frequency.
103. As seen in DX-4934, RiskIQ’s 2020 Mobile App Threat Landscape Report, the top three
stores (Xiaomi, Baidu, and Pconline) where users were most likely to download malware are all
from China and heavily used by Chinese users. This has translated to Chinese users facing
increasing security risks when obtaining apps from these app stores. As far back as 2013, studies
have shown that the fragmentation in China’s Android market has resulted in “nearly 35 percent
of the Android apps . . . secretly stealing user data unrelated to the app’s functionality.” DX-
24
4555. Just within the first month of 2021, 157 apps on China’s Android app stores were
discovered by China’s Ministry of Industry and Information Technology to be violating security
and privacy regulations; these included widely-used apps from large corporations, such as apps
developed by Tencent.
104. The proliferation of malicious apps in China’s Android market is largely attributed to
looser security standards that accompany the multitude of app stores distributing apps. In a 2018
study analyzing over 6 million Android apps obtained from 16 Chinese Android app stores and
Google Play, the Chinese Android app market was found to have a higher prevalence of fake,
cloned, and malicious apps in Chinese stores than in Google Play, possibly due to market
operators indulgently oversighting copyright and security checks over the apps. As found in a
2020 study, various Android apps in China require more default permissions than their iOS
counterparts and therefore entail fewer function-specific authorization requests than on iOS
devices. The default permissions in Android grant access to certain types of data by default
when users install and use the app. In this case, Chinese users not only face looser security
standards, they are also left at their discretion to make cybersecurity-related decisions such as
whether to change authorization permissions or privacy settings. As I discussed above, not
forcing users to make those types of decisions facilitates a much safer mobile app distribution
ecosystem.
6. Implementing Epic’s Experts’ Proposals Would Entail Changing iOS

in a Manner That Could Open New Threats and Have Negative
Security Impacts
105. Apple specifically designed its iOS software and hardware to prevent the download and
installation of apps outside the App Store. Changing iOS to permit third-party app stores as Drs.
Lee and Mickens appear to suggest, however, could create new threats and negative security
impacts.
a) On-Device Malware Scanning Would Require Weakening of

the Sandbox and Would Not Necessarily Provide Sufficient
Protection
106. As I noted earlier, Dr. Mickens argues that the introduction of third-party stores on iOS
would not hinder security because Apple could apply anti-malware protection to scan apps on
iOS devices. However, the use of anti-malware scanning would require changing, and therefore
weakening, Apple’s current sandboxing.
107. The anti-malware apps that Dr. Mickens proposes would need operating system hooks to
perform real-time monitoring, and file access inside other apps’ sandboxes to perform on-
demand scanning. Apps performing these activities would require high-level system privileges,
and they would break through iOS sandbox protections with every scan activity. These scenarios
lead to heightened security risks. These scenarios also may lead to detrimental impairment to
device usability. Constantly running anti-malware scans and updates negatively impacts user
experience, since these activities would significantly drain a device’s battery and reduce overall
performance.

25
b) Sharing Entitlements and Apple’s Private APIs Will Create

Vulnerabilities and Potential Instabilities
108. Drs. Lee and Mickens also suggest, or otherwise appear to assume, that third-party app
stores and/or app developers would be provided with a list of Apple’s private APIs and
entitlement management ability. Both of these suggestions also present security concerns,
particularly if used by third-party app developers outside established security review procedures
and standards.
109. Apple’s App Store Review Guidelines currently forbid the use of Apple’s private APIs,
which might be used to perform unauthorized activities or to extort private information; they
might also change any time and break app stability. As Dr. Lee acknowledged in his expert
report, private APIs can be used, for example, to allow attackers to circumvent iOS sandbox or
obtain private data without permission. Apple’s private API detection during App Review is
backed by a complete list of private APIs that it internally collects and manages. This is an
example where Apple’s knowledge base distinguishes itself; Apple has developed automated
processes to detect whether the APIs called by an app are private APIs or some of the thousands
of APIs on the iOS platform. Apple’s computer tools here operate accurately and efficiently
(and can outperform humans in this type of review). Apple’s control of private APIs is a part of
Apple’s commitment to user privacy and security; restricting third-party use of Apple’s private
APIs helps ensure that apps cannot interfere with the core functions and stability of iOS.
110. Apple also manages the entitlements granted to third-party apps in connection with
protecting against potential security, privacy, and reliability threats. Drs. Lee and Mickens have
not, however, addressed the management of entitlements and signing in their proposed multi-app
store scenario, including who would be responsible for evaluating requested entitlements or
checking signing of apps for third-party app stores.
111. Currently, when developers want to test their app, Apple provides them with a
provisioning profile, which is a type of system profile used to launch one or more apps on
devices and use certain services. Entitlements are granted to apps when a user is given this
provisioning profile. According to Apple’s current process, app developers must request special
entitlements from Apple, who will review an app’s use case to ensure privacy policies are
followed appropriately and in accordance with the legitimate purpose of that app.
112. However, permitting third-party stores could mean that developers deploying apps for
third-party app stores would not need Apple-provided provisioning profiles and therefore grant
themselves their own entitlements (as well as first-party entitlements, which are entitlements that
can only be used by Apple in the current environment). By consequence, apps distributed
through third-party app stores could easily be granted elevated privileges that allow apps to
bypass certain on-device security.
113. Dr. Mickens states that a majority of security measures enforced during app review could
be implemented using on-device security, but if apps can easily circumvent these protections
because they are granted (or can grant themselves) certain entitlements, then these on-device
protections would be largely irrelevant. This furthers my opinion that app review is necessary,
and that allowing third-party app stores onto iOS could present serious security concerns.
26
114. For example, when Epic distributed Fortnite to Samsung devices using its own third-
party downloader, the game was installed silently via a private Samsung Galaxy Apps API. A
silent download implies that users are not given any prompts to confirm the permissions that the
game requests. This signifies the importance of controlling entitlements. If other apps followed
in Epic’s footsteps, they could easily access privileged functionality without alerting the user. In
the absence of Apple as an arbiter evaluating and granting entitlements in connection with efforts
to guarantee a level of security and privacy on par with Apple’s standards, the lack of restriction
on entitlements for apps distributed through third-party app stores could undermine the security,
privacy, and reliability of the entirety of the iOS ecosystem.
115. Moreover, the consequences of granting entitlements and signing without regard for the
potential risks created by those processes can be very substantial as it would result in a less
secure app landscape where users are free to acquire apps from arbitrary third-party origins.
Previously, I explained that such a process of downloading apps from arbitrary sources and
insecure third-party app stores contributes to the fact that Android is less secure than iOS.
Downloading from external sources on Android allows for the direct distribution of unreviewed
apps that are potentially malicious, and there is no way for users to verify whether these apps
meet a certain security standard. At the same time, there is no single “spigot” through which
Android OS companies or OEMs can control or limit the dissemination of malicious apps once
they are detected, which is opposite of what Apple is able to do on iOS. Apple can remove the
app from the App Store and revoke the developer’s certificate, preventing them from uploading
new apps or updates signed by the revoked certificate to the App Store.
116. If Apple is not permitted to itself evaluate entitlements and signing of apps to be released
on the iOS platform, Apple would be forced to rely upon third-party certificates for app
verification. This could encompass, for example, needing to allow third parties to install root
certificates on iOS for app verification. Furthermore, it exposes Apple and its users to potential
propagation of malicious apps on iOS using a stamp of a third-party that is compromised, has
lost control of their private key, maintains an insecure app review process, or even has a rogue
employee. This could critically and irreversibly hurt the overall security of the iOS platform.
117. Again, an example with respect to Epic’s activity on Android devices—which lack
entitlement and verification control—is instructive. Epic launched Fortnite Installer on Android
devices via third-party stores and sideloading in August 2018. A Man-in-The Disk (MiTD)
vulnerability—which is an attack that allows an intruder to intercept and potentially alter data
that moves between Android external storage and an installed mobile app—was discovered in
the installer soon after the release. The MiTD vulnerability allowed an attacker to hijack the app
installation process and install arbitrary apps in the background of the user device. Those
arbitrary apps installed through this vulnerability could have granted themselves full access
permissions without the user’s knowledge. In this instance, Epic had little to no control of
Fortnite Installer being continuously downloaded, since the installer could be uploaded to any
third-party app store by any user that had already downloaded the APK, and Epic could not
control any malicious exploits that had been introduced into the devices that downloaded
Fortnite using the installer.

27
c) Epic’s Proposals Have Additional Significant Negative Security

Impacts that They Have Not Fully Assessed
118. Even assuming that third parties would be able to replicate Apple’s App Review process,
there is no guarantee that they would seek to implement comparable security measures. It is
notable that Epic and its Epic’s experts failed to set out a framework for the relationship between
Apple and any third-party store. Thus, it is unclear what they envision third-party app stores
would—in practice—be permitted and entitled to do. The current state of the industry, however,
and the examples of the Chinese Android app marketplace and Google’s Google Play Store
review process, suggests that third parties would be unable to, or not inclined to, implement the
same level of security, privacy, and reliability review as that of Apple. The introduction of third-
party app stores thus would be likely to create additional vulnerabilities and security threats for
iOS.
119. Epic’s experts suggested that Apple should provide security guarantees and help third-
party app stores, if they were to exist, achieve the same security goals as the App Store. Epic’s
experts assume, however, that Apple will continue its security and trustworthiness efforts at the
same pace and to share the fruits of those efforts with all third-party stores (apparently without,
however, those third-party stores sharing in the cost of vetting apps, monitoring the iOS
ecosystem, connecting with third parties, and providing support to developers and users). If,
however, Apple could not set security guidelines for or otherwise prohibit the distribution of
unsafe, illegal, and/or malicious apps, security, privacy and trustworthiness could be
significantly degraded. This could occur, for example, if Apple could not prohibit a third-party
app store from distributing a pirated and fake version of well-known apps that seeks to obtain
private and sensitive user information, such as users’ credit card information. Similarly, if third-
party developers and app stores were to discover and exploit a security vulnerability, I would be
concerned if Apple were not able to update its iOS platform to address that security vulnerability,
regardless of whether an update results in those apps no longer being operable on iOS.
120. Dr. Mickens focuses on macOS security mechanisms such as the “warning dialog” that
macOS will display when a user tries to install an unsigned, unnotarized app, and the fact that the
user can override that warning and install the app. But it is unclear whether Apple actually could
issue such warnings in a hypothetical world where Apple is not permitted to take actions that
have the “effect of impeding or deterring competition among app distributors (including
competition between third-party app distributors and the App Store).” The absence of such
warnings, however, would have significant negative effects; although I believe that such
warnings are not sufficient to protect users, there may be few other ways for unsuspecting users
to differentiate the level of security that Apple offers from what other app stores offer. The open
question of whether Apple could provide security measures or have involvement in third-party
app stores thus presents another security risk not considered by Dr. Mickens.
121. Epics’ experts proposals could have additional negative security impacts. As I explained
earlier, third parties lack Apple’s internal feedback loop available for Apple’s App reviewers.
App reviewers are able to make suggestions to the App Store Review Guideline, and to escalate
app review findings and issues to other Apple teams handling policy, processes, and products,
both through internal Apple channels. As Mr. Kosmynka will testify, internal communications

28
between app reviewers and iOS engineers have led to iOS updates with security and privacy
advancements based on new issues identified during app review.
122. However, third-party reviewers would also be incapable of acquiring and utilizing
Apple’s open channels of communication available to Apple reviewers. Even if some feedback
loops can be established between third-party reviewers and Apple, it would be hardly as efficient
as Apple’s internal communication channels, which would lead to amplified response time for
incident response. In addition, Epic and its experts are silent on whether each app store should
establish a one-on-one channel with Apple, or Apple could just establish a communication
platform that every store could participate. They are further silent on who would be in charge of
building and maintaining such communication channels.
123. There is also no guarantee that third-party app stores would provide feedback to Apple on
their security findings. There is also no guarantee that, if they are willing to share, they will
share all necessary information in a timely manner. It has to be noted that there is competition
among app distributors (including competition between third-party app distributors and the App
Store). Such competition might inhibit certain parties’ intention to share information.
124. Potentially differing levels of security across app stores thus raise questions as to who
will scrutinize third-party app stores, Apple’s ability to ensure the safety of its iOS platform, and
the extent to which security, privacy, and reliability on iOS would be negatively impacted by
permitting third-party app stores on iOS. On Android, which permits sideloading, there can be
no effective scrutiny, and Android thus has faced third-party app stores acting as malicious app
vectors as a result. Epic’s proposed scenario does not address or try to quantify the impact on
security, how third-party app stores would be held accountable if they get caught “cheating” and
not following platform rules; or if they become a significant vector of malware for the iOS
platform? Similarly, Epic’s proposed scenario does not address a situation where Epic, for
example, could take actions through app stores on iOS that hurt Apple’s users, but Epic would be
indifferent because it could collect revenue by offering the same products elsewhere. If Apple is
prohibited from creating and enforcing security policies, Apple could be constricted in what
steps it can take to provide security measures or otherwise ensure that app stores meet security
guidelines. Apple would be unable to leverage the experience and knowledge with its own
technology and standards in seeking to ensure that third parties meet those standards. And what
recourse would Apple have if third-party app stores tarnish the platform’s reputation or place its
users at risk or directly in harm’s way? Apple could have to incur even more costs in an effort to
monitor its ecosystem and provide support to users who face those risks (and potentially attribute
some of that responsibility to Apple as the iOS platform provider). Epic and its experts fail to
realize that it will be a negative-sum game for iOS security if certain third-party app stores give
up security controls and choose not to commit to Apple’s security and privacy objectives. Where
third-party app stores consider deviating from the high standards of security and trust that Apple
has set, the protective mechanisms that Apple has put in place would have diminished effect and
Apple, its developers, and its users would face additional costs and risks. This will compromise
the integrity of the iOS platform as well as ultimately put app users in harm’s way.

29
D. Diversifying In-App Purchase Channels Could Inhibit Apple’s

Anti-Fraud Abilities
125. IAP is a single, secure, and efficient solution that consumers have learned to trust. The
IAP functionality provided by Apple aids in protecting security and privacy in the iOS platform,
including through sophisticated fraud detection. These benefits are shared by consumers,
developers, and Apple itself.
126. Allowing third-party payment systems, similar to allowing third-party app distribution
mechanisms, could lead to less secure payment mechanisms and differing security standards that
would facilitate bad acts. Apple has developed significant and extensive security protections,
including the use of cryptographically signed attestations, tamper-resistant Hardware Security
Modules (“HSMs”), and other mechanisms for safeguarding user data. IAP protects the privacy
and security of iOS users by withholding their private information from developers as well as
Apple employees. Users’ different payment methods and payment details are stored in the
tamper resistant HSM on a server, so even Apple employees do not have access to them. Thus,
when a customer wants to make an in-app purchase, once the user is authenticated, the
transaction can happen seamlessly and securely. Apple also provides cryptographically signed
attestations that tie an application’s state to a particular device. This allows a developer to have
the assurance that customers are not cheating their application with multiple devices that are
shared by different users. This proprietary functionality enables IAP to benefit from the ability to
utilize such features as the on-device store kit, which can be used to verify whether a receipt for
a purchase is authentic.
127. Third parties, however, may be unable to, or choose not to, use Apple’s proprietary on-
device cryptographic hardware, hardware-based attestations, or IAP APIs, which Apple
maintains confidentially in order to protect PII of Apple’s customers. I also have not seen any
commitments, by Epic or Epic’s experts, that third parties should be required to utilize reputable
payment handlers as third-party payment mechanisms. Thus, where multiple parties are
processing payments of digital goods in apps, a fraudster could choose one payment method that
is easier to exploit, or bounce between payment methods and payment processors to avoid
detection. Breach of a third-party payment system would potentially expose private data,
including financial information and PII, to attackers.
128. Allowing third-party payment systems also would curtail Apple’s ability to monitor and
detect fraud and abuse. Use of Apple’s IAP functionality centralizes the purchase of digital
goods and services in apps with IAP, safeguards user data, and maintains visibility into the entire
payment process. For example, Apple checks that a developer server has confirmed receipt and
that the customer has in fact purchased the content before it is delivered to the customer. This
maintains the integrity and traceability of the transaction and confirms developer’s receipt of
transaction, and the delivery of digital goods to appropriate customers. It also enables Apple’s
fraud algorithms to process as much information as possible across all the transactions. Apple
uses this deep learning system to determine whether an account has been compromised. These
types of learning techniques are more accurate when more data points are available. IAP fraud
protection is enhanced by the very fact that it operates in a centralized system for the entire iOS
ecosystem. The balkanization of in-app payment processing systems thus would limit the
amount of data that Apple can aggregate and analyze overall—and consequently, the fraud
30
OAKLAND DIVISION

JAMES E. MALACKOWSKI
vs.
APPLE INC.,
Time: 8:00 a.m.
Defendant, Counterclaimant. Courtroom: 1, 4th Floor
Judge: Hon. Yvonne Gonzalez Rogers
Ex. Expert

05640-YGR-TSH James E. Malackowski
1. My work is focused on providing an understanding of the value of Apple’s intellectual

property (“IP”) assets, to which it has exclusive rights, and the use of Apple’s extensive IP by
others. I also analyze Apple’s investment in innovation, which has resulted in a significant IP
portfolio, of which a substantial portion is directed to the technology at issue in this case. The
significance of this technology to the iOS ecosystem is amplified because Apple’s innovation is
integrated across iOS hardware, software and app developer tools and each element, as well as
the iOS ecosystem as a whole, benefits from Apple’s ongoing dedication to innovation. Epic’s
requested remedies and the proposals of its experts not only seek to allow Epic to free-ride on
Apple’s IP through the equivalent of a compulsory license, but expect that Apple will continue to
innovate, for Apple’s own iOS ecosystem as well as for the benefit of Epic and its proposed app
store, and will provide Epic with free access to Apple’s ongoing innovation.
2. Opinion 1 – Apple’s IP rights are generally exclusive rights. ¶¶ 15-21, pp. 3-4.
3. Opinion 2 – Apple has valuable IP in its iOS ecosystem. Apple has made a substantial
and sustained investment in research and development resulting in a significant IP portfolio, a
substantial portion of which is directed to the technology at issue in this case. ¶¶ 22-25, pp. 4-6.
4. Opinion 3 – The significance of Apple’s IP relating to iOS, the App Store, and software
and tools licensed to app developers is amplified because Apple’s innovation is integrated across
iOS hardware, software and app developer tools. ¶¶ 26-38, pp. 6-8.
5. Opinion 4 – Apple’s IP rights are integral to the development, testing and distribution of
apps for the iOS ecosystem. Apps and app developers, as well as the iOS ecosystem as a whole,
benefit from Apple’s ongoing dedication to innovation. ¶¶ 39-43, pp. 8-9.
6. Opinion 5 – Epic acknowledges its use of Apple IP in the development, promotion and
distribution of the iOS version of Fortnite and the importance of that use to the development of
its Fortnite iOS app. ¶¶ 44-46, pp. 9-11.
7. Opinion 6 – The analyses of Epic and its experts in this matter are flawed because they
fail to consider Apple’s IP rights and the importance of Apple’s patents, copyrights, trademarks
and trade secrets, Apple’s history of innovation and the value of those innovations to consumers
and app developers. ¶¶ 47-60, pp. 11-14.
8. Opinion 7 – Epic is seeking essentially a compulsory license to all of the IP necessary to
distribute apps to iOS users, without any provision for compensation to Apple for the use of its
IP, and without any restrictions that Apple may deem necessary to protect the security, privacy
and reliability of the iOS system for itself, consumers and other developers. This proposal is
fundamentally inconsistent with the exclusive rights conferred by patent, copyright, trademark
and other IP laws. ¶¶ 61-71, pp. 14-16.


1
9. Opinion 7.1 – Epic’s requested injunction and its experts’ proposals seek to require
Apple to invest in innovation to benefit Epic and its proposed app store, without compensation to
Apple. ¶¶ 64-68, pp. 15-16
10. Opinion 7.2 – Epic’s remedies and its experts’ proposals seek to prevent Apple from
enforcing its IP rights or even setting the terms according to which such IP rights would be
licensed. ¶¶ 69-71, p. 16.
11. I am the Chief Executive Officer of Ocean Tomo, LLC, and am responsible for its
services, which relate to intellectual property, including financial expert testimony, valuation,
strategy consulting, patent analytics, investment management and transaction brokerage. Prior to
founding Ocean Tomo in 2003, I conducted intellectual property valuation, strategy and
investment management analyses while holding positions with other firms. I have specialized in
intellectual property issues for more than 35 years.
12. My experience extends to matters of general business valuation and commercial disputes,
both domestic and foreign. On more than fifty occasions, I have served as an expert in U.S.
Federal Court, U.S. Bankruptcy Court, State Court and the Ontario Superior Court of Justice on
questions relating to intellectual property economics, including the subject of valuation and the
equities of a potential injunction.
13. I am a frequent instructor for graduate studies on intellectual property management and
markets as well as speaking on emerging technology markets and related financial measures. I
have substantial experience as a Board Director for leading technology corporations and research
organizations as well as companies with critical brand management issues. As an inventor, I
hold more than twenty U.S. patents. I am Certified in Financial Forensics, a Certified Licensing
Professional, and a Registered Certified Public Accountant in the State of Illinois.
14. My expertise in intellectual property has been recognized by leading intellectual property
publications and organizations:
IAM Magazine:
A founding and continuous member of the IP Hall of Fame Academy, comprised
of individuals with acknowledged expertise in international intellectual property
issues;
One of the World’s Leading IP strategists since 2007, comprised of individuals
who lead in the development and implementation of strategies that maximize the
value of IP portfolios; and
One of the World’s Leading Patent Professionals, based on recognition in the
industry for exceptional skill sets and profound insights into patent matters


2
National Law Journal: listed in its inaugural list of 50 Intellectual Property

Trailblazers & Pioneers
Licensing Executives Society:
Past President of LES International and of LES USA & Canada, Inc.; and
Member of the Standards Development Organization Board, regarding the design
of and teaching of business process aspects of intellectual capital management
World Economic Forum, one of fewer than twenty members of the Network of
Global Agenda Councils to focus on questions of IP policy
III. Opinions
A. Apple’s Intellectual Property Rights Are Generally Exclusive

15. The U.S. Constitution guarantees the right of exclusivity to IP assets as an incentive to
innovate, resulting in the creation of new goods and services. Exclusive IP rights not only
incentivize innovation, they can also protect an IP owner’s rights in: maintaining or improving
the quality of their product or service; ensuring consumer safety or improving product security
and privacy; improving the ease with which consumers can use a service; and preventing free-
riding.
16. IP owners, having invested in innovation and obtained patents, copyrights, trademarks
and/or trade secrets to protect their inventions, have the right to seek a return on their investment.
IP protections offer an incentive to invest in costly research and development by providing the
exclusive opportunity to obtain the reward of economic benefits resulting from the investment.
One strategy is to use the IP owner’s right of exclusion to prevent use of their IP by others. An
alternative available method for an IP owner to obtain a return on its investment in innovation is
to license some or all of its IP. An IP owner does not have an obligation to grant a license to its
IP, nor is an IP owner denied the ability to enforce its IP because it refused to license its IP.
17. IP owners can select the strategy they find most beneficial if they decide to license their
IP rights, including whether the license is exclusive or non-exclusive, whether they will seek a
payment, the form of payment they choose to seek (one-time payment, running royalty, tiered
royalties, structured payment terms or a combination), the length of the license provided,
restrictions on use of the IP, and terms regarding the ownership of related IP developed as a
result of the license.
18. Apple has the right to the exclusive use of its IP, to determine if and how to provide to
others access to its IP, and to seek a return on its investment in innovating and developing IP.
Apple has enforced its IP rights in various courts and before regulatory agencies, including
multiple patent-infringement suits against, for example, Samsung; copyright infringement
litigations against, for example, Psystar; testimony before the Copyright Office to protect its
copyrights; trademark litigation against, for example, Amazon; and trade secret litigation against
employees who leaked confidential Apple information.



3
19. In some instances, Apple has exercised its right to IP exclusivity. For example, unlike
Google and Microsoft, Apple does not license its operating systems, including iOS, for use on
third-party devices. Apple has chosen to reserve and exercise its right to exclusively use its
intellectual property for the design and production of its own devices.
20. When it launched the iPhone in 2007, Apple initially chose not to permit the use of iOS
by third-parties to create native apps for download and installation on iOS devices. That
changed with the introduction of the iPhone Developer Program in 2008 and Apple’s release of
its software development kit (“SDK”) for third-party developers’ use in creating native
applications for the iPhone. To facilitate app development by entities ranging from individuals
to large companies, Apple has provided access to Apple’s IP, including its IP that covers
approximately 150,000 iOS application programming interfaces (“APIs”) and other app
developer tools, to the members of the Apple Developer Program and registered iOS developers
pursuant to its Apple Developer Agreement and Developer Program License Agreement
(“DPLA”), as well as other license agreements.
21. Access to Apple’s IP and the App Store resulted in an explosion in the number of apps
available on the App Store, from more than 500 apps available at the launch of the App Store on
July 11, 2008 to 1.4 million apps available in 2015 and 1.8 million apps available in November
2020, with more than 180 billion cumulative downloads by app users.
B. Apple Has Valuable IP in Its iOS Ecosystem

22. Apple’s investment in research and development has grown over time. Since its
FY 2005, when Apple began working in earnest on the iPhone, Apple’s annual research and
development expenditures have grown from approximately $0.5 billion in FY 2005 to almost
$18.8 billion in FY 2020, for a combined total of more than $101 billion in this period, resulting,
in part, in iOS ecosystem innovations protected by patents, trademarks, copyrights, and trade
secrets.

23. Apple’s innovations and IP rights (patents, copyrights, trademarks and trade secrets) in
the iOS ecosystem obtained during its innovation related to the iPhone and iOS include IP rights
related to technology at the center of this case, such as IP rights relating to Apple’s App Store,
IAP, iOS development tools, App Review, and iOS security. These IP rights include:



4
Almost 3,200 U.S. patents and applications relating to app distribution and
development, including development tools, APIs in frameworks and related services,
in-app purchases, and security and privacy and other technology designed into iOS
and iPhone hardware and software that are used by app developers in developing iOS
apps, including, among others:
108 U.S. patents and 10 U.S. patent applications referring to Apple’s CoreAudio,
CoreGraphics, and CoreVideo APIs, [DX-4889.200-.203] which function to
interact with the iOS device’s media hardware and media pipelines for processing
digital video;
4 U.S. patents and 1 U.S. patent application referring to CoreMotion, [DX-
4889.200-.203] which allows the use of movement and environment-related event
data;
22 U.S. patents and 6 U.S. patent applications referring to UIKit, [DX-4889.200-
.203] which defines core components of an iOS application offering features such
as event handling infrastructure or multi-touch, animation support, and app
extension support and resource management;
5 U.S. patents and 12 U.S. patent applications relating to In-App Purchase [DX-
4889.200-.203]; and
1 U.S. patent and 1 U.S. patent application referring to CloudKit, [DX-4889.200-
.203] which enables authentication and structured asset storage services;
165 U.S. patents and 91 U.S. patent applications referencing the App Store [DX-
4889.200-.203];
Approximately 1,237 U.S. patents and 559 U.S. patent applications referencing iOS
[DX-4889.207] as well as more than 200 registered copyrights including the term
iOS, [DX-4889.197] as well as copyrights on Apple documentation, source code, and
software;
Registered trademarks for the term App Store®, developer tools such as CloudKit®
and Swift® and slogans used to promote the App Store, such as “There’s an app for
that”®; [DX-4889.198] and
IP rights in security technology central to the ongoing protection and function of the
Apple App Store and iOS devices operating third-party native apps, including U.S.
patents and applications on security and authentication-related technologies and tools
as well as trade secrets.
24. Apple’s continuing investment in its iOS ecosystem, from its initial development of
iPhone-related technology through today, has enhanced value by adding features to new
generations of the iPhone and new versions of iOS, as well as enhancing other elements of the
iOS ecosystem. Each component of the iOS ecosystem, as developed and improved over time,
has and contributes value, including the App Store, IAP, iOS development tools, and iOS
ecosystem security, including App Review.



5
25. Apple’s commitment to and investment in innovation related to the iOS ecosystem has
resulted in success of the iOS platform, dramatic growth of the App Store, and continuing trust
of customers in the security of the iOS platform from Apple’s consistent focus on innovations in
security, as well as a substantial portfolio of IP. Corroborating my opinion regarding the
significance of Apple’s IP, an independent analyst has estimated the value of Apple’s intangible
assets, including intellectual property assets, to be over 60 percent of Apple’s business enterprise
value.
C. Apple’s Integrated Innovation Across iOS Ecosystem Amplifies the Significance of

Apple’s IP
26. Apple’s integration of its innovation across its iOS hardware, software and app developer
tools amplifies the significance of Apple’s IP relating to iOS, the App Store, and software and
tools licensed to app developers. Apple’s improvements in its iOS ecosystem are integrated,
with improvements in one aspect combining with improvements in other aspects to enhance the
overall iOS ecosystem. The integrated nature of this development benefits Apple’s iOS
hardware, software and app developer tools, which have evolved in combination and are
integrated to reinforce and improve each other.
27. The integrated nature of the development of innovations in the iOS ecosystem has been
present since the introduction of the iPhone, when Mr. Jobs explained that the iPhone was a
combination of three revolutionary devices and offered benefits from the combination of
innovations in its novel touch screen, human interface, and other hardware, as well as the
“desktop” features, apps, and other software provided by its operating system (then known as
“OS X,” before being renamed “iOS” in 2010, which is a foundational layer of software,
allowing apps to run and access features of the device, such as the touch screen). After the
launch of the App Store, this integrated innovation approach was used in developing and
improving APIs that work with and build on hardware and software features. Mr. Cook
confirmed that the iOS ecosystem continues to benefit from this integration of innovation across
proprietary hardware, software and service elements.
28. This integrated nature of development extends to Apple’s iPhone security architecture,
which builds security measures into iPhone hardware that support and integrate with the security
protections provided by software and app review. Apple provides layers of protection to help
ensure that apps are free of known malware and additional protections are enforced regarding
access by apps to user data. Apple’s security controls are implemented to provide a stable,
secure platform for apps.
1. API and Software Innovations
29. Apple’s APIs and other software and tools offered to app developers integrate with and
rely on iOS and iPhone hardware features such as iPhone motion sensors, media hardware and
media pipelines for processing digital video, as well as touch screen. By developing its iOS
hardware, operating systems and software tools in coordinated efforts across multiple Apple
teams, including those focusing on chips, software, APIs, and the camera, Apple is able to
streamline getting innovations to iOS app developers, allowing them to access new features more



6
quickly, facilitating development of apps with enhanced performance and quality, which leads to
more users downloading those apps.
30. A majority of the engineering performed by Apple supports and benefits app developers,
whether it is for developing hardware features, software features in the operating system,
Apple’s APIs or capabilities of the App Store for distributing and supporting apps. Apple’s
investment in engineering and other resources in its model of design integration produces app
developer tools and services that not only gives app developers access to new iPhone and iOS
features, but also enables app functionality that previously did not exist and provides easier to
use and more powerful capabilities by which app developers can incorporate these new iPhone
and iOS features into their apps, all while maintaining user privacy and security.
31. Since Apple’s launch of its first SDK in 2008, containing 10,000 APIs, Apple has
provided a suite of tools that facilitated use of iPhone’s multimedia hardware and software to
support creation of full console games, not feature-constrained cell phone games. Multi-touch is
one example of the benefits of Apple’s iPhone and API technology for app development,
including for game apps. At Apple’s 2008 event launching its SDK, app developers highlighted
multiple APIs, including Cocoa Touch, which used iPhone’s advanced multi-touch capabilities,
allowing everything from single finger touch to multi-finger touch to gestures. The touch layer
of the iPhone has benefited from continued Apple investment resulting in, for example, enhanced
smooth scrolling and a very fast rate of frames per second.
32. Apple has continued to add and update features in new generations of the iPhone and new
versions of iOS, such as the iPhone’s built-in sensors, camera technology, audio technology,
networking technology and data storage technology tools and services made available to app
developers and included by app developers in apps. For example, Apple’s innovation in graphics
technology includes its enhanced Metal framework, which directly communicates with the
graphics processors to render advanced 3D graphics and perform data parallel computations,
maximizing the graphics and computing potential of iOS. Apple has protected Metal with 15
U.S. patents and applications, as well as copyrights and trademarks. Epic used Metal for
Fortnite and promoted Metal at the 2015 Worldwide Developer Conference (“WWDC 2015”)
and the 2018 Worldwide Developer Conference (“WWDC 2018”), including characterizing
Metal as revolutionizing graphic design and allowing app developers to create richer 3D worlds.
33. The benefits of the integrated nature of Apple’s iOS ecosystem innovations for app
developers are illustrated by Nex Team’s HomeCourt app, introduced in 2018, which
demonstrated a new application for the camera using the CoreML app. Apple holds a U.S.
patent application relating to CoreML, [DX-4889.200] which integrates complex machine
learning technology developed by Apple with data from iPhone hardware for use by an app
developer. HomeCourt uses CoreML to track basketball shots by simply pointing an iPhone at a
court. Without any sensors on the basketball court, ball, or player, HomeCourt can measure a
player’s release time and angle of a shot, how fast the player is moving before shooting, and the
player’s vertical leap. HomeCourt took advantage of the new A12 bionic chip in the latest
iPhone models, XS, XS Max, XR, with significantly faster speed to offer, for the first time, “real-
time AI-powered analysis.”



7
34. Apple’s integrated approach to the development of its APIs, software, and security
innovations benefits Apple as well as iOS app developers, enhancing Apple’s ability to compete
for the attention of app developers, and providing incentive for app developers to release their
apps first on the iOS platform. Apple’s approach reduces app developers’ work from the
programming side, while providing developers access to new powerful features of each iOS
device.
2. Security Innovations
35. Mobile devices, particularly those where native third-party apps can be installed, face
significant and distinctive challenges that have forced Apple to innovate with respect to its
security approaches and protections. App developers benefit from Apple security and privacy
innovations that are designed into iOS and iPhone hardware and software, structured asset
storage, and other software technology, as does the functionality and operation of Apple’s App
Store. The layers of protection that Apple provides help ensure that apps are free of malware,
and that the App Store provides a stable and secure platform for apps.
36. The Apple iOS platform combines hardware, software and services that work together to
provide security, with the ultimate goal of keeping personal information safe. Apple’s security
approach includes building security measures into iPhone hardware that support and integrate
with the security protections provided by software and app review. Apple works to ensure that
App Review security tools are designed and updated for the latest software and hardware, to
ensure that they are ready to perform their security functions upon the release of the latest
software and hardware.
37. Apple has obtained U.S. patents on security and authentication-related technologies and
tools.
47 U.S. patents and an additional 37 U.S. patent applications referencing
authentication technology;
35 U.S. patents and an additional 41 U.S. patent applications referencing
authorization technology; and
8 U.S. patents and an additional 7 U.S. patent applications referencing framebuffer
and security. The use of framebuffer objects can permit an app to move complex
image decompression and drawing logic into a separate process for the purpose of
increasing security.
38. Apple also protects as trade secrets certain of its investment in security technology. In
the security arena it is important to maintain a level of secrecy about the details of Apple’s
proprietary security mechanisms to avoid providing a roadmap of Apple’s defenses to potential
attackers, making trade secrets an important element of Apple’s security-related IP.
D. Apple’s IP Rights are Integral to Development, Testing and Distribution of Apps

39. Apple’s IP rights are integral to development, testing and distribution of apps for the iOS
ecosystem. Apple has invested in innovating and developing tools, software and other



8
technology for the development and testing of apps and distribution of apps through Apple’s App
Store on Apple’s iOS devices. The App Store, and Apple’s innovations relating to the App
Store, have benefited consumers and developers by creating a secure and trusted marketplace for
apps.
1. App Developers Access Apple’s IP Rights Pursuant to Licenses
40. Apple makes available to app developers an extensive library of developer
documentation and API code, including tutorials, sample code, articles and APIs, that are
protected by Apple IP, including patents, copyrights, trademarks and trade secrets. App
developers use Apple’s IP to develop apps for the iOS ecosystem, to distribute apps for use in the
iOS ecosystem by consumers, and to monetize apps through the iOS ecosystem, if the developer
chooses to charge for its apps and/or in-app purchases of digital content. In other words, a
developer cannot participate in the iOS app marketplace without using Apple’s IP.
41. Apple has chosen to create the “Apple Developer Program” in order to enable app
developers to access and use protected Apple technology and services for their own development
and testing of apps, as well as the ultimate distribution of such apps through the App Store.
Through the Apple Developer Program, app developers can obtain limited licenses, pursuant to
the Developer Agreement and the DPLA, to access a broad array of tools, software and other IP
created, updated and enhanced by Apple, including the SDKs, operating systems, beta OS
releases (providing access to developers prior to public release of updates), advanced app
capabilities, and other software, as well as the App Store Connect suite of web-based tools to
distribute pre-release versions of apps via TestFlight and to publish approved apps on the App
Store. These licenses work in conjunction with various software-specific licenses, such as
license agreements that govern each download of Xcode, SDK, or sample code, as well as
Apple’s published, publicly-available Guidelines.
42. The license terms under which Apple provides app developers access to its IP are stated
in the agreements it enters with app developers, such as the terms of the Apple Developer
Agreement, the DPLA and individual software licenses, and are consistent for all app developers.
Apple’s licensing practice is long-standing and serves Apple’s business objective of maintaining
a dynamic App Store for the benefit of users and app developers while protecting Apple’s IP
rights. Developers, including Epic, are permitted to use Apple’s proprietary software, tools
and/or services only pursuant to licenses from Apple. The licensing arrangements help prevent
developers from free-riding on Apple’s innovation.
43. Similarly, Apple has maintained a consistent policy of requiring iPhone users to accept
and agree to the terms of an iOS and iPadOS Software License Agreement under which “Apple
and its licensors retain ownership of the Apple Software itself and reserve all rights not expressly
granted.”
E. Epic Acknowledges Use of Apple IP

44. Epic acknowledges that it has used Apple IP in its development, promotion and
distribution of the iOS version of Fortnite. Apple has more than 2,500 U.S. patents covering



9
aspects of technology used by app developers such as Epic, as well as copyrights, trademarks,
and trade secrets. Epic has used, for example, thousands of Apple APIs, multiple versions of
Apple SDKs and XCode builds, and the XCode integrated development environment that
provides developers with software features necessary to design, develop and debug software for
use on iOS and macOS. In addition, Epic has benefited from engineering support, services and
capabilities provided by Apple to support the development of advanced app features and address
issues with Fortnite.
45. Epic has admitted to using at least the following Apple APIs, which are referenced in
more than 280 Apple U.S. patents and applications. [DX-4889.200]
U.S. Patents U.S. Applications Total

AudioToolbox 2 - 2
AVFoundation 2 1 3
CloudKit 1 1 2
CoreAudio 38 4 42
CoreGraphics 72 7 79
CoreMedia 26 4 30
CoreMotion 4 1 5
CoreVideo 19 1 20
Foundation 26 5 31
GameController 6 2 8
GameKit 1 - 1
iAD 4 1 5
In-App Purchase 5 12 17
Metal 11 4 15
MultipeerConnectivity - 1 1
QuartzCore 5 - 5
StoreKit 1 1 2
UIKit 22 6 28
UserNotifications 12 3 15
WebKit 52 10 62
Total 309 64 373
Total Removing
Duplicate Patents 235 52 287
46. Apple places copyright protection on its software, documentation and source code
covering iOS content, APIs, beta iOS releases, SDKs and other materials used by an app
developer like Epic. The names of certain frameworks Epic acknowledges using are
trademarked by Apple, including CloudKit®, iAd® and Metal®. Epic has benefited from the



10
use of Apple’s trademarks and brand when, for example, Epic showcased its launch of Fortnite
during the WWDC 2015, presented about Fortnite on iOS during the WWDC 2018, and
collaborated in other marketing efforts with Apple. Apple has also promoted Fortnite through its
marketing channels including App Store banners featuring Fortnite, as well as posts and paid
advertisements on social media.
F. Epics’ Experts’ Analyses Are Flawed Because They Ignore Apple’s IP Rights
47. The analyses of Epic and its experts, Drs. Evans, Athey and Mickens, are flawed because
they fail to consider Apple’s IP rights and the importance of Apple’s patents, copyrights,
trademarks and trade secrets. None of these experts addressed the importance of Apple’s IP and
the role of Apple’s IP rights, nor did they acknowledge the need to consider or evaluate Apple’s
IP, in providing their opinions regarding the potential distribution of apps outside the App Store
and security aspects of this proposed app distribution, including Apple’s App Review process.
48. Epic’s experts’ analyses and proposals appear to assume and rely on broad access to
Apple’s IP, such as access to Apple’s IP-protected iPhone features, iOS, APIs and proprietary
tools used for app reviews, potentially including those tools protected as trade secrets or others
not shared with third-parties. This assumed access is incorporated in the proposals of Epic’s
experts without consideration of any need to obtain authorization from Apple, including the
terms and conditions of license agreements, or the need to provide the compensation that would
be due to Apple for use of Apple’s exclusive IP rights.
49. It is my opinion that Epic’s experts’ analyses and proposals, and as a result their ultimate
conclusions, are flawed because of this failure. In fact, both Dr. Evans and Dr. Athey
acknowledged in their depositions that Apple’s IP would be used in the but-for world each
propose, confirming the flaws in their analyses and the need to consider Apple’s IP. Evans Dep.
468:25-469:9; Athey Dep. 212:18-24.
1. Dr. Evans Ignores Apple’s IP and History of Innovation
50. Dr. Evans opined that Apple has invested little in innovation and provided limited
services to developers. Although Dr. Evans acknowledged that “Apple was able to use iOS and
the features of the iPhone to provide a well-designed and easily used app store,” and that the
iPhone was characterized as “revolutionary” offering multi-touch and “desktop-class email, web
browsing, searching and maps,” he asserts that “other app stores would have had access to those
same iOS and iPhone features for their stores.” Dr. Evans, however, did not identify any inquiry
that he performed into the specific nature of Apple’s innovation, the contributions made by
Apple’s innovation and features, the investment that Apple has made in its innovations, or the
protection of those innovations by Apple IP. Instead, to reach his conclusion regarding Apple’s
lack of investment in innovation and limited services to app developers, Dr. Evans only listed
features that he claims relate to Apple’s lack of investment in innovation and provided anecdotal,
high-level descriptions, without specific consideration of Apple’s innovation or contributions.
51. Despite his failure to evaluate the Apple IP rights protecting its innovations with respect
to the App Store, Dr. Evans testified during deposition that he recognizes that a developer must



11
use Apple’s IP to develop apps for the iOS ecosystem. He testified that he would expect it to be
the case that Apple makes available to app developers an extensive library of developer
documentation and code that is protected by Apple IP rights. Evans Dep. 468:25-469:9. These
admissions reinforce my opinion that Dr. Evans’s failure to evaluate Apple’s innovation and IP
rights, the role they play in Apple’s iOS and iPhone features used by apps, the App Store and
Apple’s app developer tools and services, and the terms and conditions under which others may
be granted access to Apple’s IP rights, casts doubt on his analysis of the hypothetical world that
he contends would exist “absent Apple’s exclusionary practices.”
52. Dr. Evans reached his conclusion regarding Apple’s lack of innovation without analyzing
or acknowledging either Apple’s IP relating to the iPhone, iOS, App Store or app developer tools
and services, or Apple’s investment in research and development. Over the life of the iPhone,
Apple has continued to innovate and introduce improvements and enhancements in each
component of its ecosystem – hardware, software and services – that benefit app developers and
the App Store. Dr. Evans’ but-for world appears to assume continued access to Apple’s IP in its
iOS and iPhone features by third-party app stores, without any consideration of the terms,
conditions, or compensation for that access, even though in his deposition Dr. Evans admitted
that developers must use Apple’s IP to develop apps for the iOS ecosystem. Evans Dep. 468:25-
469:9.
53. Dr. Evans’ but-for world not only disregards Apple’s existing IP and its extensive
research and development, that world is not consistent with Epic’s own requirement in the Epic
Online Services Developer Agreement that “establish[es] terms regarding [a developer’s] use of
the [Epic] SDK and related Services” and sets out various license restrictions and acceptable
uses. Epic acknowledges that the terms and restrictions in a license agreement can provide for
compensation, consistent with Epic’s interest in obtaining a return on its proprietary software
rather than “provid[ing] Unreal Engine as a public service.” Epic has sought and obtained 15
active granted U.S. patents, 154 U.S. copyrights and 315 trademarks, giving it the right to
exclude use by others and to obtain a reward for the right to use its invention. The same
principles are equally applicable to Apple, but Dr. Evans disregards them in his but-for world.
2. Dr. Athey’s Middleware Analysis Ignores Apple’s IP Despite Her Deposition
Admission that Her Proposed Economic Middleware Would Use Apple’s IP
54. Dr. Athey defines “economic middleware” as technologies that reduce user application-
related switching costs; reduce user application-related mixing-and-matching costs; reduce
developers’ costs of providing services that enable user app migration and synchronization to
multiple platforms, further reducing user app-related switching and mixing-and-matching costs,
and reduce developers’ multi-homing costs. Dr. Athey opines that she would expect greater
innovation in economic middleware absent Apple’s challenged conduct. In offering her opinion,
Dr. Athey never analyzes the degree to which Apple IP would need to be used to develop such
middleware for iOS apps. Inherent in Dr. Athey’s analysis is her assumption – confirmed during
her deposition – that her proposed economic middleware would require connection to the iOS
operating system through APIs. Because those APIs are protected by Apple IP, her proposal
would require the use of Apple’s IP. Athey Dep. 212:18-24. Dr. Athey also acknowledged in



12
her deposition that developers of iOS apps must use Apple IP, and that it is not possible to
develop an iOS app without the ability to call iOS APIs or to put an app on the iOS platform
without using Apple’s tools and APIs on which Apple holds many patents. Athey Dep. 209:1-7,
25-210:2. In addition, Dr. Athey testified that the iOS apps connect to iOS through the APIs
exposed by the operating system and that Apple would have to remove its IP-protecting
restrictions to allow multiplatform app stores to operate on the iOS platform. Athey Dep.
212:18-213:24.
55. Dr. Athey’s deposition admissions confirm the flaws in her opinions. Apps using Dr.
Athey’s hypothetical middleware would need to use – and operate with – the IP-protected iOS
that Apple has developed, which reflects and protects Apple’s earlier and ongoing developments
in iOS, the App Store, and developer tools. Dr. Athey’s proposed economic middleware, and the
hypothetical but-for world in which it would operate, fail to take into account Apple’s IP rights,
which are exclusive rights that Apple enjoys under federal and state IP laws. Nor did Dr. Athey
evaluate whether Apple would have to make changes in order to accommodate her proposed
economic middleware and, to the extent changes would be required, what those changes would
be. Dr. Athey also does not address whether or how Apple would contract with multiplatform
app stores, or the form of a license, if any, between Apple and the middleware entity.
3. Dr. Mickens Assumes Use of Apple’s IP Without Permission
56. Without explicitly acknowledging that Apple has IP rights protecting its iPhone and app
security, Dr. Mickens acknowledges that Apple’s security process uses iOS technology as well
as App Review and that Apple must invest in security technology on a sustained basis to address
evolving security threats.
57. Dr. Mickens provides an assessment of app security that emphasizes the role and
importance of Apple’s proprietary iOS software. Dr. Mickens concludes that many, but not all,
aspects of app security can be achieved based on an operating system alone, without the
assistance of App Review.
58. According to Dr. Rubin, and as I understand Apple’s witnesses will testify, Apple’s
security architecture, and the way in which Apple provides iOS and app security, is far more
sophisticated than reliance only on iOS on-device security. He also observes the importance of
updates to Apple’s existing App Review and iOS security and other software, and the evolving
nature of threats posed to iOS devices. Apple maintains many security tools for its internal use
that are protected to ensure that they do not provide a roadmap for circumvention by potentially
malicious app developers.
59. Apple’s IP rights protect the technology throughout Apple’s layers of protection in its
hardware and software, including updates. Apple maintains these protections by updating its
proprietary App Review tools as a regular and ongoing process consistent with the release of the
latest hardware and software from Apple and to identify the latest threats.
60. Dr. Mickens ignores that Apple has chosen not to license iOS, and instead to maintain its
exclusive right to iOS as proprietary software. Dr. Mickens’ analysis appears to rely on


13
assumptions that third-parties would be provided access to Apple’s proprietary iOS features as
well as IP-protected APIs, security patents, and trade secrets for the foreseeable future. It is only
with use of these Apple IP-protected security assets, but without addressing terms and conditions
of or compensation for such use, that Dr. Mickens is able to propose that third-parties could
attempt to provide comparable levels of security to that provided by Apple.
G. Epic Essentially Seeks a Compulsory License to Apple’s IP Necessary to Distribute

and Operate iOS Apps
61. I understand that Epic challenges various of the restrictions in Apple’s DPLA and has
indicated that it intends to seek injunctions that would prevent Apple from “restricting,
prohibiting, impeding, or deterring the distribution of iOS apps through a distribution channel
other than the App Store,” including by enforcing its contractual provisions, guidelines, or
policies. I also understand that Epic intends to seek an injunction that would prevent Apple from
“[d]enying iOS app stores [and iOS apps downloaded through a distribution channel other than
the App Store] access to iOS functionality that the App Store [and apps downloaded through the
App Store] has access to.”
62. Epic’s requested injunctions, and its experts’ proposals, are fundamentally inconsistent
with the exclusive rights conferred by patent, copyright, trademark and trade secret laws. Epic is
essentially seeking a compulsory license to all of Apple’s IP necessary to distribute apps to iOS
users, without apparent consideration of Apple’s IP that would be required, the investment made
by Apple to create such IP, or the compensation due to Apple for the use of its IP. They would
force Apple to make available aspects of iOS technology that it has not previously licensed. In
addition, Epic’s apparent request for a compulsory license does not appear to contemplate
permitting Apple to implement restrictions on the use of its IP that Apple may deem necessary to
protect the security, privacy and reliability of its iOS system for itself, consumers and other
developers. Epic’s experts failed to address the terms and conditions of any use of Apple’s IP in
their proposals that app developers be allowed to distribute iOS apps through third-party app
stores on iOS devices.
63. Epic’s requested injunctions and expert proposals would essentially eliminate the
limitations under which Apple has agreed to license its proprietary technology. Rather than a
limited, non-exclusive, personal, revocable, non-sublicensable and non-transferable license to IP
for the right to develop, test and distribute apps using Apple IP through Apple’s protected App
Store, Epic seeks an unlimited license to develop, test and distribute apps using Apple IP.
1. Epic’s Demand for Apple to Innovate and Invest for Epic’s Benefit
64. If implemented as requested, Epic’s proposed injunction and its experts’ proposals would
require Apple to innovate and invest for the benefit of Epic and its proposed app store,
apparently without compensation to Apple, reducing Apple’s incentive to innovate and invest in
IP.
65. Apple has made substantial investments in innovation related to the iPhone, iOS, APIs
and related security. The analyses and proposals of Epic’s experts appear to assume and rely on



14
broad access to Apple’s IP protecting these innovations, such as access to Apple’s IP-protected
iPhone hardware and software features, security protections, APIs, and proprietary tools,
including those protected as trade secrets, used for app review.
66. Epic’s experts, Drs. Evans, Athey and Mickens, each appear to confirm, but disregard,
that Epic’s proposed injunctions and their preferred outcomes would require ongoing,
presumably unlicensed, access to Apple’s IP, including its iOS technology, by app developers
and third-parties providing app stores on iOS devices, including access to aspects of its iOS
technology not previously licensed or made available to third-parties. Although Drs. Evans and
Athey assume access to Apple’s IP, their analyses and opinions are not accompanied by any
discussion of obtaining authorization from Apple, the terms and conditions of any license
agreements, evaluating the IP rights that would be used by Epic in distributing apps to iOS users,
or providing (or attempting to determine) the appropriate amount of compensation that would be
due to Apple for use of Apple’s exclusive IP rights. Similarly, the suggestion by Dr. Mickens
that third-parties be permitted access to Apple’s proprietary security technology and processes
ignores Apple’s IP rights in iOS and its proprietary security technology and processes as well as
related materials, such as underlying source code and other documentation.
67. I understand that the scope of the compulsory license that Epic requests would not be
limited to only Apple’s currently licensed IP. Epic’s proposals demand continuing and
equivalent access to any Apple IP that would be available for use by apps distributed through
Apple’s App Store, without evaluating or identifying such IP. This IP would include both
Apple’s current technology and innovation in iOS, as well as additional iOS technology that
Apple does not presently license and potential future IP-protected technology resulting from
Apple’s ongoing innovation and development that would be used by future apps. Epic’s
requested injunctions and expert proposals would not only seek to compel Apple to grant a
license to its existing IP, but to maintain and update that IP for the benefit of Epic, and any other
app developer, whether or not that developer had ever entered into an Apple license, apparently
without compensation and under terms and conditions not set by Apple. Epic does not evaluate
the extent to which accommodating Epic’s requests would require Apple to modify iOS to enable
third-party app store access; to develop or modify iOS and iPhone security procedures to protect
users and the iOS ecosystem from malware or other vulnerabilities that could be introduced
through the middleware or third-party app stores; or to facilitate updates made by the developers
of the apps distributed through the third-party app stores.
68. Based on an assumption that Apple will continue to invest in developing IP as it has in
the past, Epic’s experts similarly assume not only continued access to and use of Apple’s
currently existing technology, but also additional Apple technology as Apple continues to
innovate, including any that may be required to be developed by Apple, such as Apple IP needed
to download, install and execute an app on an iOS device from third-party app stores. There is
no consideration of the appropriate compensation for the use of Apple’s IP in Epic’s experts’
proposed alternative worlds, nor the appropriate terms and conditions for use of Apple’s
exclusive IP rights. If imposed as requested, the compulsory license implied by Epic’s
injunction requests and its experts’ proposals would potentially increase Apple’s costs while
reducing the potential return or reward available to Apple from its innovations.



15
2. Impact of Epic’s Remedies and Its Experts’ Proposals on Apple’s IP Rights

69. iOS is Apple’s proprietary operating system, protected by Apple’s IP, which has been
developed and enhanced through Apple’s investments since at least Apple’s FY 2005. Apple’s
IP-protected innovations in its iOS ecosystem integrate its development of and improvements to
iOS hardware, software, and security innovations in the iPhone and App Store with its
development and innovations in app developer APIs and other software tools.
70. Epic’s proposed injunctions and the proposals of its experts would impose additional
costs and burdens on Apple as well as limit its right to compensation for its investment in its IP,
thereby reducing the potential return or reward available to Apple from innovation and impairing
the value of Apple’s IP. Instead, Apple would be forced to grant a license to Apple’s software
and technology on terms set by Epic – not Apple – that are favorable to Epic, for Epic’s own
purposes.
71. Epic’s injunctions and its experts’ proposals seek to preclude Apple from claiming the
well-established rewards from innovation and obtaining intellectual property protection:
compensation and/or the right to exclude. Epic’s requested relief would prevent Apple from
enforcing the IP rights in which it has invested and obtained exclusive ownership rights,
including preventing Apple from enforcing, or even determining, the requirements and
restrictions set out in its various license agreements and guidelines, thereby violating Apple’s
right to decide whether, and if so on what terms, it will license its technology protected by
patents, copyrights, trademarks and/or trade secrets, which are governed by laws giving Apple
exclusive rights.

IV. Oath
and correct.
____
Dated: April 23, 2021
Word count: 7,355


16

Apple&#39;s Expert Written Testimony - Epic Games v. Apple

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Apple&#39;s Expert Written Testimony - Epic Games v. Apple

Uploaded by

Copyright:

Available Formats

Case 4:20-cv-05640-YGR Document 490 Filed 04/27/21 Page 1 of 2

1 THEODORE J. BOUTROUS JR., SBN 132099 MARK A. PERRY, SBN 212532

19 EPIC GAMES, INC., Case No. 4:20-cv-05640-YGR-TSH

20 Plaintiff, Counter- NOTICE OF FILING OF DEFENDANT

Gibson, Dunn &

9 Dated: April 27, 2021 Respectfully submitted,

12 By: /s/ Rachel S. Brass

14 Attorney for Defendant Apple Inc.

Gibson, Dunn & 1

UNITED STATES DISTRICT COURT

NORTHERN DISTRICT OF CALIFORNIA

EPIC GAMES, INC., No. 4:20-CV-05640-YGR-TSH

April 23, 2021

Epic v. Apple, No. 4:20-CV- 1 Written Direct Testimony of

3. Opinion 2. This overarching conclusion is strongly supported by quantitative and qualitative

Epic v. Apple, No. 4:20-CV- 2 Written Direct Testimony of

Epic v. Apple, No. 4:20-CV- 3 Written Direct Testimony of

Epic v. Apple, No. 4:20-CV- 4 Written Direct Testimony of

II. Background and Qualifications

Epic v. Apple, No. 4:20-CV- 5 Written Direct Testimony of

Epic v. Apple, No. 4:20-CV- 6 Written Direct Testimony of

Examples of digital game transaction platforms

Platform Compatible devices

Epic v. Apple, No. 4:20-CV- 7 Written Direct Testimony of

Worldwide Fortnite user accounts by platform (July 2017 – December 2020)

Source: DX-4761 (summarizing DX-5339, DX-4314, DX-3686)

Epic v. Apple, No. 4:20-CV- 12 Written Direct Testimony of

Epic v. Apple, No. 4:20-CV- 13 Written Direct Testimony of

Portability of FY2019 top game apps in the App Store

Top 25 by downloads 32% 32%

Source: DX-4777 (summarizing DX-5338, DX-4324, online public materials)

For apps offering purchased digital content

Top 25 by downloads 100% 28% 8%

Source: DX-4798 (summarizing DX-5338, DX-4324, online public materials)

Epic v. Apple, No. 4:20-CV- 15 Written Direct Testimony of

E. Empirical evidence shows that consumers treat game transactions on other

Epic v. Apple, No. 4:20-CV- 16 Written Direct Testimony of

Other electronic devices used by iOS App Store users

Other Electronic Devices 458 92%

Source: DX-4765 (summarizing DX-5339)

Epic v. Apple, No. 4:20-CV- 20 Written Direct Testimony of

Source: DX-4765 (summarizing DX-5339)

Epic v. Apple, No. 4:20-CV- 21 Written Direct Testimony of

Source: DX-4766 (summarizing DX-5339)

Epic v. Apple, No. 4:20-CV- 22 Written Direct Testimony of

Total users Percent of users that are

Hours played Revenue

Epic v. Apple, No. 4:20-CV- 23 Written Direct Testimony of

Epic v. Apple, No. 4:20-CV- 25 Written Direct Testimony of

i. Analysis of users who downloaded a console or PC game transaction platform

Source: DX-4792 (summarizing DX-5338)

ii. The launch of Fortnite on the Nintendo Switch

Epic v. Apple, No. 4:20-CV- 27 Written Direct Testimony of

Source: DX-4822 (summarizing DX-5339, Nintendo, “Fortnite”)

Epic v. Apple, No. 4:20-CV- 28 Written Direct Testimony of

Source: DX-4823 (summarizing DX-5339, Nintendo, “Fortnite”)

Epic v. Apple, No. 4:20-CV- 29 Written Direct Testimony of

Source: DX-4771 (summarizing DX-5339)

Epic v. Apple, No. 4:20-CV- 30 Written Direct Testimony of

Increase in revenue on other “platforms” as a

3. iOS single-homers 26.5% 37.5% 30.6% 22.4%

4. iOS multi-homers 18.7% 35.3% 63.2% 49.4%

Source: DX-4824 (summarizing DX-5339)

Epic v. Apple, No. 4:20-CV- 31 Written Direct Testimony of

Apple's Expert Written Testimony - Epic Games v. Apple

Apple's Expert Written Testimony - Epic Games v. Apple