Federal Court of Australia

Anthony v Apple Inc [2025] FCA 902

ORDERS

THE COURT ORDERS THAT:

1.    The further hearing of the proceeding be stood over to a date to be fixed.

2.    Save for the oral summary of these reasons given by Beach J at the time of delivery of this judgment and any republication in any form of that summary, pursuant to s 37AF(1) of the Federal Court of Australia Act 1976 (Cth) and on the ground set out in s 37AG(1)(a), subject to further order, disclosure and publication of these written reasons and the content thereof shall only be made to the parties’ external legal advisors in this proceeding and in proceedings NSD 190 of 2021, NSD 1236 of 2020 and VID 342 of 2022, and to no other person.

3.    Liberty to apply.

4.    Costs reserved.

Note:    Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.

ORDERS

THE COURT ORDERS THAT:

1.    The further hearing of the proceeding be stood over to a date to be fixed.

2.    Save for the oral summary of these reasons given by Beach J at the time of delivery of this judgment and any republication in any form of that summary, pursuant to s 37AF(1) of the Federal Court of Australia Act 1976 (Cth) and on the ground set out in s 37AG(1)(a), subject to further order, disclosure and publication of these written reasons and the content thereof shall only be made to the parties’ external legal advisors in this proceeding and in proceedings NSD 190 of 2021, NSD 1236 of 2020 and VID 341 of 2022, and to no other person.

3.    Liberty to apply.

4.    Costs reserved.

Note:    Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.

REASONS FOR JUDGMENT

BEACH J:

1    There are four proceedings before me, the joint trial of which took place over a period of four months on questions of liability.

2    In the first proceeding, the Epic parties have made claims against the Apple parties principally concerning alleged contraventions of s 46 and other provisions of the Competition and Consumer Act 2010 (Cth) (CCA).

3    The Epic parties are involved in developing entertainment software for personal computers, smart mobile devices and gaming consoles, with their most popular game being Fortnite. They allege that the Apple parties’ contravening conduct has forced app developers including Epic to only use Apple’s App Store to distribute their software applications (apps) to the broad base of iOS device users, and to only use Apple’s payment system for processing the purchases of their in-app digital content by iOS device users. Apple removed Fortnite from the App Store on 13 August 2020.

4    I have delivered a separate set of reasons dealing with the first proceeding (Epic Games, Inc v Apple Inc [2025] FCA 900).

5    In the second proceeding, the Epic parties have made claims against the Google parties principally also concerning analogous alleged contraventions of s 46 and other provisions of the CCA. They allege that the Google parties’ contravening conduct has hindered the ability of app developers including Epic to distribute apps to Android devices in a realistic and practical way other than through Google’s Play Store. It is said that Google has achieved this by imposing various contractual and technical restrictions, which stifle or block Android device users’ ability to download app stores and apps directly from developers’ websites and prevent or hinder competition in the distribution of apps to Android devices. It is also said that Google has inappropriately imposed on app developers Google’s payment system for processing the purchases of in-app digital content by Android device users. Google removed Fortnite from the Play Store on 13 August 2020.

6    I have delivered a separate set of reasons dealing with the second proceeding (Epic Games, Inc v Google LLC [2025] FCA 901).

7    The third and fourth proceedings, which these present reasons deal with, are representative proceedings commenced by the representative applicants under Part IVA of the Federal Court of Australia Act 1976 (Cth) where similar allegations have been made by them on behalf of group members against the Apple parties and the Google parties concerning the same contraventions of s 46 of the CCA as alleged in the first proceeding and the second proceeding. One of the class actions is against the Apple parties concerning the overcharging of commissions on the App Store. The other class action is against the Google parties concerning the overcharging of commissions on the Play Store.

8    In each case it is said that the contraventions caused app developers to pay materially higher commissions on paid app downloads and in-app purchases of digital content than they would otherwise have had to pay had the contravening conduct not occurred. For convenience I will refer to the representative applicants in both proceedings on their own behalf and in their representative capacity as the class applicants.

9    Now in each of these proceedings the class applicants allege that the contravening conduct of Apple and Google contrary to s 46 occurred during the period from 6 November 2017 to 20 June 2022 (the relevant period). So, unlike Epic in the other two proceedings, they make no claim concerning ongoing conduct of Apple and Google concerning s 46.

10    Now at the trial of these representative proceedings, which as I say I heard concurrently with the two main proceedings being Epic v Apple and Epic v Google, the issues to be determined were confined to liability and an additional issue that was unique to the representative proceedings, being whether there had been an overcharge of commission by the respondents. That latter question has required me to consider relevant retrospective hypothetical scenarios including what commissions Apple and Google would have charged through the App Store and Play Store respectively absent the contravening conduct.

11    As I have indicated in my reasons in the other cases, these reasons should be read together with my reasons in Epic v Apple and Epic v Google, which I will assume a familiarity with and which provide the necessary foundation for these reasons in the two representative actions.

12    The class applicants’ case on liability is on all fours with Epic’s case in its separate proceeding against Apple and its separate proceeding against Google save that, as I say, the class applicants allege contraventions by the respondents of Part IV of the CCA limited to the relevant period.

13    And as I have said, the additional point relevant for present purposes is that it is said that Apple and Google each separately caused app developers to pay materially higher commissions on paid app downloads and in-app purchases of digital content than they would otherwise have done had the contravening conduct not occurred. I will refer to the commissions that would have been paid absent the contravening conduct as the counterfactual commissions.

14    It is said that in respect of each of Apple and Google, there would have been materially lower counterfactual commissions by reason of there being competition or substantially more effective competition in the relevant markets during the relevant period. I will elaborate on two different dimensions of counterfactuals in a moment.

15    Now the overcharge issue only arises if there is a finding of liability based on the same case that Epic brings against Apple and Google, which liability findings I have made.

16    Now in the two Epic proceedings I accepted Epic’s position on the various market definitions.

17    Further, the effect of my liability findings in the two Epic proceedings is that each of Apple and Google had substantial power in the relevant markets. Further, speaking very generally for the moment, the purpose and effect of the relevant Apple and Google conduct was to lessen substantially competition in the relevant markets.

18    So, my findings set out in my reasons in the two Epic proceedings against Apple and Google concerning market definition, market power and whether Apple’s and/or Google’s conduct had an anti-competitive purpose should be taken to be part of the foundation for these reasons.

19    Further, in terms of anti-competitive effect applying the future with and without test, I would make a similar point save that I have also added some additional analysis concerning the counterfactual scenarios which should be read together with what I have said in the other reasons. I will say more about counterfactuals in a moment.

20    Further, I have paid little attention to the views of Mr Holt, who I have referred to in my other reasons, on market definition or market power, save that Mr Holt has expressed views concerning the question of the economic and accounting profitability of the App Store and the Play Store which do have some relevance to questions of overcharge and counterfactual questions which I will explain in a moment, and which as a consequence I have considered notwithstanding that such matters are also relevant to whether inferences can be drawn as to the existence of substantial market power.

21    Now it has been necessary for me to consider two categories of hypotheticals which are conceptually different.

22    The first hypothetical is a retrospective hypothetical, which principally only arises in the class actions on the question of whether there has been an overcharge from the start of the relevant period to the present or from a later point in time to the present. The retrospective hypothetical question is as to what fee Apple and Google would have or could have charged absent the impugned conduct? Such a question has to be answered on the balance of probabilities. But of course, uncertainty and imponderables are involved in the calculus or, more appropriately characterised, the estimation. But one has to do the best one can. Moreover, there may be different figures for different scenarios at different times or over different periods. There may not be one figure, whether just varying with time or otherwise.

23    The second hypothetical is a future hypothetical applying the with and without test, that is, an assessment of the likely or realistic future with the impugned conduct going forward compared with the likely or realistic future without such impugned conduct going forward.

24    The second hypothetical goes more to the question of whether any contravention has occurred involving inter-alia s 46. The first hypothetical goes more to the question of overcharge and the issue of monetary relief.

25    But the second hypothetical can in one sense be posed at an earlier time in that the future with and without test is framed at the time the contravention is said to have occurred, and looking forward from that earlier time.

26    Now as the class applicants point out, in a counterfactual that assumes the absence of such conduct, it is consistent with economic theory that competition or effective competition, including through actual or credible threat of new entry, will result in a lowering of prices.

27    I agree with the class applicants that whilst greater competition on quality cannot be ignored, its scope is limited and the likelihood is that such competition would, in the counterfactual, be overwhelmed by price competition.

28    In the counterfactual without the elements of the impugned conduct, acting as commercially rational firms without market power, Apple and Google would have taken legitimate steps to respond to competition. As the class applicants point out, the rational legitimate step absent the elements of the contravening conduct would have been to lower prices. That Apple and Google would have taken such a step, and promptly, is supported by two further matters, according to the class applicants.

29    First, they say that to the extent that either held any concern as to the security of their operating systems, the legitimate way to alleviate that concern to the extent it reasonably existed would have been to deter entry by lowering prices.

30    Second, they say that each had a ready means of recouping revenue lost by lowering prices, being advertising.

31    Further, the class applicants say that in a commercially realistic counterfactual, it is unlikely that Apple and Google would have acted in any other way. By delaying any price reductions, or relying only on claims of perceived quality, the respondents would have risked rival app stores establishing themselves in the relevant markets. Those rivals would have eroded Apple’s and Google’s market share through price competition.

32    So, the class applicants say that the credible threat of new entry in the absence of the contravening conduct makes it more likely that there would have been a prompt reduction in commissions charged by Apple and Google for paid app downloads and in-app purchases.

33    I must say that there is a lot to be said for the class applicants’ points. I will address this later.

34    Now the expert evidence adduced by the class applicants sought to establish before me that the effective counterfactual commission in both the Apple and Google class actions would have been in the range of between 10 to 20% during the relevant period with a most likely single figure, as an average, of 15%.

35    Now I have concluded on the evidence as a whole that the counterfactual commission rates would likely have been materially less than were charged by Apple and Google and potentially others during the relevant period, but by how much I cannot say. I will need to hear further from the parties as to this once they have had a chance to consider my reasons.

36    I should also say that I am not convinced that there should be a single figure for the counterfactual commissions. Any relevant counterfactual commission may vary depending upon the time period one is considering or possibly the type of app or possibly developer revenue type considerations.

37    Finally by way of introduction, although at this first stage I am dealing with counterfactual commissions and overcharge questions, I am not dealing as such with any loss and damage. Determination of the loss and damage caused to the applicants and group members by Apple’s and Google’s impugned conduct is outside the scope of the initial trial in these proceedings.

38    For convenience, I have divided these reasons into the following sections:

(a)    The structure of the two class actions – [40] to [72].

(b)    Matters for determination – [73] to [105].

(c)    Some legal principles – [106] to [133].

(d)    The counterfactual commission – introduction – [134] to [151].

(e)    App Store profitability – [152] to [232].

(f)    The counterfactual scenario(s) – Apple context – [233] to [365].

(g)    Comparator analysis – Apple context – [366] to [516].

(h)    The counterfactual commission(s) – Apple context – [517] to [549].

(i)    Play Store profitability – [550] to [639].

(j)    The counterfactual scenario(s) – Google context – [640] to [719].

(k)    Comparator analysis – Google context – [720] to [834].

(l)    The counterfactual commission(s) – Google context – [835] to [857].

(m)    Conclusions – [858] to [860].

39    Now before proceeding further, let me say something about the applicants and the group members in each of the class actions.

The structure of the two class actions

40    Let me say something first about the Apple class action.

The Apple class action

41    Mr David Anthony is the first applicant in the Apple class action. During the relevant period, Mr Anthony owned an Apple iPhone and made purchases of apps on the App Store for use on his iPhone. Mr Anthony also purchased in-app digital content within iOS apps during the relevant period.

42    Mr Anthony says that he entered into three in-app transactions on an iOS device, which have a total value of $21.47. The apps referenced in Mr Anthony’s pleading in respect of which those transactions took place all appear to be gaming apps. Mr Anthony does not identify whether the three in-app transactions on which he relies involved transacting with a developer who was charged a 30% commission rate on those transactions by Apple.

43    Dark Ice Interactive Pty Ltd is the second applicant in both the Apple and Google class actions. During the relevant period, Dark Ice was a developer of both iOS and Android apps. In that period, it distributed its iOS apps for download by app users through the App Store and for download by app users of its Android apps through the Play Store. Within its apps, and during the relevant period, Dark Ice also distributed in-app digital content to app users. The apps that Dark Ice developed were distributed through the App Store and the Play Store.

44    The Apple class action was commenced as a representative proceeding on behalf of the following persons.

45    First, it was commenced on behalf of all persons who at any time during the relevant period purchased an iOS app on an iOS device from the Australian App Store and/or purchased in-app digital content within such an iOS app, and thereby suffered loss or damage by reason of Apple’s conduct.

46    Second, it was commenced on behalf of all persons who at any time during the relevant period supplied an iOS app for iOS devices via the Australian App Store and/or supplied in-app digital content within such an iOS app, and thereby suffered loss or damage by reason of Apple’s conduct.

47    Let me say something more about Dark Ice in the Apple context.

48    Dark Ice entered into a developer agreement and a developer program license agreement (DPLA) with Apple on 6 May 2013. It is the developer of various iOS apps being: an iOS app, Pocket Cal kJ, which adopts the “freemium” model that I have discussed in my reasons in the two Epic cases; an iOS app, Pocket Cal kJ Pro, which is a paid app, that is, the user pays to download the app; an iOS app, Pocket Weight Track, which is free to download and does not offer the user any in-app purchase transactions; and an iOS app, PHL Portal, which is free to download and does not offer the user any in-app purchase transactions.

49    Now since 22 December 2020, Dark Ice has been enrolled in the App Store’s small business program, paying commissions to Apple for paid downloads of its apps or paid in-app digital content in an amount of 15%. It pays no commission to Apple in respect of Pocket Weight Track or PHL Portal, as those apps are free to download and do not offer any in-app purchases.

50    From 6 May 2013 until late 2020 or early 2021, Dark Ice paid commissions to Apple of 30% on paid iOS app downloads and downloads of in-app digital content within its iOS apps. During late 2020 or early 2021, Dark Ice joined the App Store small business program. From this time, it paid commissions of 15% to Apple on paid iOS app downloads and purchases of in-app digital content within its iOS apps.

51    Now it should also be noted that the apps developed by Dark Ice are not intended to offer an iOS user the kind of experience and functionality on the user’s iOS device that would make the apps competitive with all other types of apps, such as game apps.

52    Further, Dark Ice as a participant in the small business program pays 15% commission to Apple only on the download of the Pocket Cal kJ Pro and such of those in-app purchases which take place on the Pocket Cal kJ and Pocket Cal kJ Pro apps. Apple says that on the class applicants’ own case taken at its highest, that is the same rate which it is said would apply if the allegedly contravening conduct did not occur.

53    Now in all other respects, Dark Ice pays nothing beyond a nominal USD 99 developer fee to Apple for its use of Apple’s myriad of tools and technology to develop and maintain its iOS apps and its distribution of those apps through the App Store and all other services.

54    The class applicants’ case is that, based upon Mr Holt’s calculations, Apple’s effective commission rate based on transaction data for Australia was 29.1% in 2017, 28.8% in 2018, 28.3% in 2019, 28.1% in 2020, 26.8% in 2021, and 26.6% in 2022.

55    Now the class applicants have relied upon an effective rate of commission that they say Apple obtained during the relevant period. But Apple says that the effective rate is misleading because it fails to appropriately weigh the commission rate paid to Apple by reference to the total revenue generated by the developer through transactions undertaken on Apple’s iOS platform. Instead, the so-called effective rate simply divides the total commission for all developers by total paid transactions for all developers.

56    Apple also says that the class applicants’ effective rate is misleading because it does not take into account that a large percentage of developers who distribute paid apps or offer in-app purchases are eligible for the small business program, which charges commission at a rate of 15%. Many such developers including Australian developers are currently enrolled in the program. Further, Apple says that the effective rate ignores that no commission is charged on multiple types of transactions, for example, downloads of free iOS apps.

57    Now as I have said, in the Apple class action there are two distinct categories of group members. There are iOS device group members, being all persons who at any time during the relevant period purchased an iOS app or in-app digital content on an iOS device through the Australian storefront of the App Store and who suffered loss or damage by reason of the alleged contravening conduct. Further, there are iOS developer group members, being all persons who at any time during the relevant period supplied an iOS app on iOS devices via the App Store or in-app digital content within such iOS app and who suffered loss or damage by reason of alleged contravening conduct.

The Google class action

58    Mr Brett McDonald is the first applicant in the Google class action. During the relevant period, Mr McDonald owned an Android smart mobile device and made purchases of apps on Google’s Play Store for use on his Android smart mobile device. Mr McDonald also made in-app purchases within Android apps during the relevant period. Further, as I have said, Dark Ice is the second applicant.

59    The Google class action was commenced as a representative proceeding on behalf of the following persons.

60    First, it was commenced on behalf of all persons who at any time during the relevant period purchased an Android app on an Android smart mobile device from the Australian Play Store and/or purchased in-app digital content within such an Android app, and thereby suffered loss or damage by reason of Google’s conduct.

61    Second, it was commenced on behalf of all persons who at any time during the relevant period supplied an Android app for Android devices via the Australian Play Store and/or supplied in-app digital content within such an Android app, and thereby suffered loss or damage by reason of Google’s conduct.

62    Let me say something about Dark Ice’s position vis-à-vis Google.

63    Dark Ice is also a developer of Android apps and in-app digital content within Android apps.

64    During the relevant period, its Android apps and in-app digital content within Android apps were distributed to Android devices via the Play Store.

65    Since the commencement of the relevant period, Dark Ice has supplied the Pocket Cal/kj Android app via the Play Store. While the Pocket Cal/kj Android app is free to download, it offers an in-app purchase of Pocket Cal/kj Plus at a price of $4.49. The Pocket Cal/kj Android app was initially released on 13 March 2012 and the current version was released on 4 October 2022.

66    Since the commencement of the relevant period, Dark Ice has supplied the Pocket Cal/kj Pro Android app via the Play Store as a paid Android app. Its current price is $1.99. The Pocket Cal/kj Pro Android app was initially released on 16 April 2012 and the current version was released on 15 January 2016.

67    Until 1 January 2022, Dark Ice paid 30% commission on all in-app purchases within its Android apps downloaded through the Play Store and, after that date, the commission paid by Dark Ice on in-app purchases reduced to 15%.

68    From the inception of the predecessor to Google Play Billing in 2011, Google charged developers a 30% fee on all sales of apps and on in-app purchases. This was extended to subscriptions in 2012. During the relevant period, Google offered exceptions to its headline commission by reducing its commission rate to 15%. These exceptions included the following matters.

69    From 1 January 2018, Google reduced its service fee for subscriptions that had been active for more than one year from 30% to 15%. From 1 July 2020, Google reduced its service fee from 30% to 15% for the first USD 1 million earned each year by each developer. In October 2021, Google announced that it would reduce its service fee for subscriptions from 30% to 15% in the first year, effective from 1 January 2022. Google also offered reduced service fees of 15%, or lower in some cases, for developers eligible for the play media experience program, the living room accelerator program, the audio distribution accelerator program, and the books and comics accelerator program. Google also entered into various ad hoc agreements offering fees below 15%.

70    Given that Google offered exceptions to its headline commission rate, it is necessary to adjust the headline rate to account for these discounts in order to calculate the effective rate that it actually charged Android app developers on relevant purchases during the relevant period. This can be determined by taking an average of the various commissions charged by Google, weighted by the value of the transactions subject to the various commission rates, thereby producing an effective commission.

71    Mr Holt calculated that, taking into account the mix of commission rates charged by Google on relevant purchases globally, its effective commission rate was 29.8% in 2017, 29.5% in 2018, 29.4% in 2019, 29.4% in 2020 and 28.5% in 2021.

72    Further, and as I have said, in the Google class action group members consist of all persons who, during the relevant period, purchased an Android app on an Android smart mobile device from the Play Store and suffered loss or damage by reason of the Google respondents’ contravening conduct. Group members also consist of all persons who supplied an Android app and/or in-app digital content on Android devices via the Play Store and suffered loss or damage by reason of the Google respondents’ contravening conduct.

Matters for determination

73    Now an unproductive debate took place before me as to the precise matters to be determined at trial based upon what had been said or done by the previous docket judge, Perram J, at an earlier joint case management hearing. I should say now that Apple’s and Google’s complaints on these questions lacked substance. Moreover, nothing said or ordered by Perram J has in any way foreclosed or constrained how I have proceeded and what I have considered appropriate to decide at this stage. Let me now briefly address the complaints of Apple and Google.

74    In the Apple class action the class applicants were confined at this stage of the trial to rely on the case run by Epic save in one limited respect, which was to seek to establish that the impugned conduct caused the commissions paid by Dark Ice and iOS app developer group members for purchases of iOS apps and/or in-app digital content within an iOS app by iOS device group members to be materially higher than the commissions that would have existed had that conduct not occurred. A similar and analogous point can be made in the Google class action concerning Android apps and devices and the relevant developers and users.

75    Now I have deferred the determination of the extent of any pass through, going to the question of how the burden of any established overcharge has been split between developers and users, which ultimately feeds into loss and damage questions, which I am not addressing during this phase.

76    Now Apple and Google have said that at the trial before me, the class applicants carried the burden of proving that Apple or Google as the case may be overcharged developers by reason of conduct alleged to be in contravention of Part IV of the CCA or s 21 of the Australian Consumer Law, in the relevant period, and carried the burden of proving the amount of any such overcharge.

77    Apple and Google say that the class applicants have failed to prove the alleged commission overcharge as pleaded and particularised. And they say that if I am not in a position to make quantified findings on the overcharge, then the proceedings must be dismissed.

78    Apple and Google say that on the pleadings, there is no doubt that the quantification of overcharge is in issue. Moreover, they say that it was ordered that such an issue was to be determined during the first hearing. So it follows that any quantification must occur now. And they point out that I cannot find that overcharge occurred, as alleged, without making a determination about the counterfactual commission rate, which I am not able to do. So they say that the proceedings must be dismissed.

79    They say that on 6 April 2023 it was ordered by Perram J that such issues should be determined at the first stage trial. And they say that the effect of those orders was to require the class applicants to quantify the overcharge, but not how it was to be distributed between developers and users.

80    Both senior counsel for Apple and senior counsel for Google stressed that the initial trial was conducted on the basis that overcharge was to be quantified. But if it is not or I am not in a position to do so, then the proceedings must be dismissed.

81    Apple and Google say that the class applicants’ evidence did not quantify, and the applicants have not proved, the existence or quantum of such overcharge. And it is said that I should not permit the question of quantification of the overcharge to be deferred to some subsequent hearing. It is said that such a course would occasion real prejudice to Apple and Google, which has called lay and expert evidence from witnesses on these issues, including witnesses cross-examined by senior counsel for the class applicants and in respect of whom adverse credibility submissions are now made.

82    Apple and Google say that any deferral of the question of quantifying overcharge is inconsistent with the basis on which the class applicants were allowed to join in the present proceedings and take advantage of the initial hearing.

83    Apple and Google say that if the class applicants fail to meet the burden of proving overcharge at the initial stage of hearing, then I must dismiss the class applicants’ cases.

84    Moreover, Apple and Google say that I ought not permit the whole question of the quantification of the overcharge to be deferred to some subsequent hearing. But I do not accept this.

85    In the first place, the class applicants have not asked me to defer the issue. Rather, the class applicants say that I am in a position to determine the issue, but have also indicated that other options open to me include not determining a single rate but a range, or finding simply that there was a material overcharge without further specification. Indeed, it is that last option that I have selected from the menu.

86    Now Mr Darke SC for Apple and Mr Moore SC for Google asserted before me that it was not open to me to fail to determine a separate question or a formulated common question.

87    But it is not uncommon for judges deciding in the course of writing up their reasons for judgment in complex and lengthy cases, that having heard the evidence it is not appropriate to determine a question as agreed and formulated by the parties or as ordered at an earlier stage of the proceedings. And that does not entail an inevitable dismissal of the proceedings or any particular negative or positive answer to a question posed.

88    In my view, the proper course is for me to deliver reasons which include findings relating to counterfactual commissions and any overcharge, with further consideration of the necessary issues and outstanding questions, legal, forensic or economic, to be debated in light of those reasons.

89    The class applicants should have the benefit of my reasons concerning the contraventions that I have found and the corresponding counterfactuals considered in that light before deciding what course to take.

90    Now before proceeding further, let me deal with a separate pleading point that both Apple and Google have made but which lacks substance.

91    Apple says that the class applicants’ case on the question of overcharge is based on the commission rate developer group members would have paid to Apple, not any alternative app store, absent the impugned conduct.

92    Google similarly says that the overcharge alleged is the difference between what the second applicant and developer group members paid to Google during the relevant period, and the commissions that those developers would have paid to Google in the counterfactual.

93    But the class applicants’ latest version of their statement of claim against Apple pleads that but for the conduct Apple would have faced vigorous and effective competition from competing app stores and direct downloading options. And it pleads that competitive benefits in the counterfactual would have included lower commissions paid by app developers “to app distributors” for purchases of iOS apps and in-app purchases and competing app distributors who would have competed with Apple on commissions for app distribution and in-app purchases. Plainly the expression “to app distributors” is not limited to Apple.

94    Further, [124] pleads:

124. At all material times during the Relevant Period, the conduct of Apple Inc and/or Apple Pty Limited, referred to at paragraphs 97–98 and 108–109 (s 46), paragraphs 112–113 (s 47), paragraphs 117–118 (s 45) and paragraph 123 (s 21) (individually or in combination, the Contravening Conduct) caused the commissions paid by:

(a)     The Second Applicant; and

(b)     iOS App Developer Group Members

to Apple for purchases of iOS apps and/or in-app digital content within an iOS app by iOS Device Group Members to be materially higher than the commissions that would have existed had the Contravening Conduct not occurred.

95    The first part of this pleading addresses the factual, being that Dark Ice and group members paid commissions to Apple. The second part of the pleading addresses the counterfactual, being the commissions that would have existed had the contravening conduct not occurred.

96    But the pleaded counterfactual is not limited to commissions paid to Apple. It is the commissions that would have been paid in the counterfactual, whether to Apple or rival app stores. Moreover, the particulars that follow are particulars of the allegation as to commissions that would have existed in the counterfactual and are in some respects agnostic as to the payee of such commissions.

97    Further, the question which Mr Holt was asked to address and did address was what commission rate would have been charged in the counterfactual, with no limitation in that question as to the identity of the payee of the commission. His report makes it clear that in the counterfactual the commissions that would have been paid would not necessarily have been paid to Apple at all.

98    As the class applicants point out, their pleading asserts the existence of competing app stores and distribution sources in the counterfactual and the payment by developers in the counterfactual of lower commissions to app distributors not limited to Apple. Nowhere in the pleading or the particulars is it alleged that the developers who paid commissions to Apple in the factual would only have dealt with and paid commissions to Apple in the counterfactual.

99    The class applicants make similar points concerning the latest version of the applicants’ statement of claim against Google.

100    I am inclined to the view that Apple understood that the class applicants’ case was not limited to the commission that would have been paid to Apple in the counterfactual. I am inclined to the same view concerning Google.

101    Further, in his expert reports in both cases, Mr Holt analysed the counterfactual commissions generally but not limited to commissions paid to just Apple and Google.

102    Further, in the joint report in the Epic v Apple proceeding, Professor Hitt, Mr Holt and Mr Houston agreed that relevant considerations in determining the counterfactual position in relation to commission rates included matters that went beyond what Apple would have charged in the counterfactual, including whether there would be new entrants into the relevant market in the counterfactual and how the competition between rival app stores including both the incumbent app store, being Apple's App Store, and new entrants would play out in the counterfactual including competition on price, quality and innovation.

103    Further, as the class applicants point out, in the joint report in the Epic v Google proceeding, Professor Asker and Mr Holt joined issue on the question of commissions that group members would have paid in the counterfactual to any app distributor, not just limited to Google.

104    Moreover, each of the Apple and Google concurrent evidence sessions concerning overcharge were not limited to the issue of what prices Apple and Google would charge in the counterfactual, but proceeded on the additional basis that it was relevant to consider the likelihood and fact of entry in the counterfactual of rival iOS and Android app stores and iOS and Android sideloading options, and the extent to which developers in the counterfactual would multi-home or switch to a rival app store in order to take advantage of lower prices or take advantage of side loading options in order to pay a lower commission.

105    There is no substance to Apple’s and Google’s pleading point.

Some legal principles

106    For present purposes, it is sufficient to make the following observations on questions of causation and counterfactuals, recognising of course that I am not required in this stage of the trial to make any findings as to the loss or damage suffered by the applicants and group members.

107    I have said something about the legal principles concerning how to deal with counterfactuals in my reasons in Epic v Apple and Epic v Google. Let me add some additional matters relevant to the present context.

108    Now a counterfactual is a tool of analysis which assists in the establishment of a causal link between misconduct and loss, and of the identification of the loss. It is not a proxy for causation, which is a legal concept.

109    The class applicants bear the legal and the evidential onus of establishing causation on the balance of probabilities.

110    But whilst the plaintiff bears the burden of proving causation and loss in a retrospective hypothetical, once causation is established, uncertainties as to the precise metes and bounds of what would have happened in the counterfactual may sound in an assessment of loss and damage, with the evidential onus lying with the party asserting the relevant fact.

111    So, in a case alleging damage in the form of a loss of opportunity, the plaintiff will establish causation by showing that the contravening conduct caused the loss of a commercial opportunity which had some non-negligible value. The value itself, that is, the loss, is ascertained by reference to hypotheses and possibilities which can be evaluated as a matter of informed estimation.

112    In Generic Health Pty Ltd v Bayer Pharma Aktiengesellschaft (2018) 267 FCR 428, which involved an award of damages for patent infringement, Allsop CJ, Yates and Beach JJ held that when dealing with a retrospective hypothetical for the purposes of assessing compensatory damages, it is necessary to form an estimate of the likelihood that the hypothetical, that is, the counterfactual, would have occurred, and adjust any award of damages to reflect the degree of probability (at [181]) that the event would have occurred. We said at [181] to [186]:

We are dealing with a hypothetical situation of the past, namely, whether if a sale of Isabelle had not been made, then an equivalent sale of Yasmin would have been made. Accordingly, we must form an estimate of the likelihood that the hypothetical situation would have occurred: see Malec v JC Hutton Pty Ltd (1990) 169 CLR 638 at 639 per Brennan and Dawson JJ. In that respect, we must assess the degree of probability that such an event would have occurred and adjust any award of damages to reflect that degree of probability: Malec at 643 per Deane, Gaudron and McHugh JJ.

In another sense, the question can be looked at through the lens of the value of a lost opportunity. The lost opportunity was to make a sale of Yasmin if the sale of Isabelle had not been made. In our view, it is not in doubt that it was established before the primary judge that, on the balance of probabilities, for all sales of Isabelle there was a loss of an opportunity to make a sale of Yasmin. The question then is: what is the value of the lost opportunity? On that question, the value is to be “ascertained by reference to the degree of probabilities or possibilities” involved in the hypothetical counterfactual: Sellars v Adelaide Petroleum NL (1994) 179 CLR 332 at 355 per the plurality.

Another way to express the same point is as Brennan J described it in Sellars at 368:

Although the issue of a loss caused by the defendant’s conduct must be established on the balance of probabilities, hypotheses and possibilities the fulfilment of which cannot be proved must be evaluated to determine the amount or value of the loss suffered. Proof on the balance of probabilities has no part to play in the evaluation of such hypotheses or possibilities: evaluation is a matter of informed estimation.

This theme of informed estimation resonates with some observations made by Hayne J in Placer (Granny Smith) Pty Ltd v Thiess Contractors Pty Ltd (2003) 77 ALJR 768; 196 ALR 257 at [37] and [38]. Any estimation must be done “with as much precision as the subject matter reasonably [permits]”. Mere difficulty in estimating damages does not relieve a court “from the responsibility of estimating them as best it can”.

Interestingly, Staughton LJ in Gerber Garment Technology [1997] RPC 443 at 459 cited Sellars with approval and discussed the matter in terms of evaluating the loss of a chance. Gerber Garment Technology both at first instance ([1995] RPC 383 per Jacob J at 407 and 408) and on appeal ([1997] RPC 443 per Staughton LJ at 459 and 460) evaluated and applied a loss of a chance approach based on the possibilities and probabilities. Moreover, Jacob J at [1995] RPC 383, 395 also invoked the themes of Lord Diplock in Mallett v McMonagle [1970] AC 166 at 176, which resonate with the themes discussed in Malec at 639 and 643 that we have referred to above.

Whichever way one expresses it, in assessing the possibilities or probabilities of a hypothetical counterfactual, one is engaged in the task of estimation, even if the estimation involves an assessment of the counterfactual as being close to a certainty. But being close to a certainty is not the same thing as a certainty. If one is estimating, one still needs to apply a discount, albeit a very modest one, to reflect the assessment that one is not at a certainty. If one is looking at the value of a lost opportunity which is not certain to occur, then the valuation must involve some discount, even if a very modest one.

113    But Apple and Google say that the class applicants can derive no support from cases about assessment of lost commercial opportunities. True it is that, in such cases, a court may proceed by reference to hypotheses and possibilities not established on the balance of probabilities. But the loss of a chance is a distinctive form of loss which must be specifically pleaded. Where a plaintiff pleads its loss or damage in a defined way, it is not permitted to substitute a loss of opportunity case because its more defined case fails.

114    Further, they have pointed out what I said in Barnes v Forty Two International Limited (2014) 316 ALR 408 at [119] to [123] (Siopis and Flick JJ agreeing) concerning a damages claim that had been confected late in the day:

A fundamental requirement for the fair trial of allegations of contravention of law requires “the party making those allegations [in this case BlueFreeway] to identify the case which it seeks to make and to do that clearly and distinctly”: Forrest v Australian Securities and Investments Commission (2012) 247 CLR 486; 291 ALR 399; 91 ACSR 128; [2012] HCA 39 at [25] per French CJ, Gummow, Hayne and Kiefel JJ. BlueFreeway failed to do this in relation to its alternative causation and damages case.

…

Further, “if a plaintiff has suffered damage of a kind which is not the necessary and immediate consequence of the wrongful act, he must warn the defendant in the pleadings that the compensation claimed will extend to this damage”: Perestrello E Companhia Limitada v United Paint Co Ltd [1969] 1 WLR 570 at 579; [1969] 3 All ER 479 at 485–6 per Lord Donovan delivering the judgment of the court. This was not done by BlueFreeway…

Not only were the appellants “entitled” to a pleading by BlueFreeway identifying the head of loss as “loss of opportunity”, but they were also entitled to have that head of damage properly particularised: see, for example, David Benson Nominees Pty Ltd v Dicksons Ltd [2005] SASC 97 at [39]–[40] per Besanko J. None of this was done by BlueFreeway. For his Honour to permit BlueFreeway late in the day to put its case based upon the alternative scenario produced an unsatisfactory and inherently unfair state of affairs so far as the appellants were concerned. Pleadings serve “to ensure the basic requirement of procedural fairness that a party should have the opportunity of meeting the case against him”: Banque Commerciale SA (en liq) v Akhil Holdings Ltd (1990) 169 CLR 279 at 286; 92 ALR 53 at 58 per Mason CJ and Gaudron J. BlueFreeway’s further amended statement of claim did not ensure that basic requirement in relation to its alternative scenario. To ensure this basic requirement of procedural fairness, BlueFreeway should have been confined to its pleaded case that did not include this alternative scenario. This was not a case “in which the parties [had] deliberately chosen some different basis for the determination of their respective rights and liabilities”: at CLR 287; ALR 58–9 per Mason CJ and Gaudron J. How the trial was conducted provides no support for such an inference.

In summary, once his Honour had found against BlueFreeway’s pleaded causation and damages case, he should have dismissed the proceeding. The alternative scenario had not been pleaded, opened or run. Moreover, to allow it to be advanced for the first time in closing addresses produced inherent unfairness to the appellants. The course of evidence was not conducted on the basis of the alternative scenario. Moreover, the appellants lost the opportunity to call evidence and to appropriately cross-examine on such a scenario…

115    Now in the present case, the class applicants do not allege any opportunity was lost. Rather, they postulate an actual overcharge. They allege that there was in fact an overcharge, and class members thereby suffered loss. For the moment I accept that there is no occasion for the application of the principles applicable in loss of opportunity cases and I have put these to one side for the moment.

116    Let me say something more on counterfactuals.

117    In Berry v CCL Secure Pty Ltd (2020) 271 CLR 151, the High Court considered, relevantly, the onus of proof in a hypothetical counterfactual where a wrongdoer alleges that but for its contravening conduct, it would have deployed lawful means to bring about the same detriment to the plaintiff.

118    It was held that whilst a claimant for damages under s 82 of the then Trade Practices Act 1974 (Cth), which is relevantly the same as s 82 of the CCA, is responsible for formulating how the loss or damage claimed is to be identified, and bears the ultimate burden of proof in the sense of establishing their case on the balance of probabilities, the burden of introducing evidence is liable to shift according to the issues between the parties.

119    Berry is an illustration of the broader principle that whilst a legal onus is constant, an evidential burden can and often does shift. If, how, and to what extent an evidential onus will shift depends on how the applicant puts their case, and how thereafter the parties join issue on questions of causation and loss that arise in respect of the applicant’s chosen formulation.

120    Now the class applicants bear the legal burden of establishing causation, which in this case is done with the assistance of a retrospective hypothetical counterfactual.

121    But to the extent that Apple or Google claim that in a retrospective hypothetical counterfactual, they would have charged an additional fee so as to mitigate the financial impact on them of not being permitted to charge an anti-competitive commission, they may at the least have an evidentiary onus on that aspect.

122    Now in the present context in terms of the claims made by the class applicants, the overcharge part of the case is not concerned with conduct going forward and matters I may pay regard to in formulating any injunctive relief. It is concerned only with what occurred. So understood, the damages to be considered must be compensation for the unlawful conduct that in fact occurred.

123    An analogy may be drawn with damages in patent infringement cases. It is no defence to a claim for damages by a patentee for lost profits on sales it would have made but for the sales by the defendant of patent infringing products that the defendant could have competed lawfully by using a non-infringing product and causing the same loss of sales.

124    If, contrary to the foregoing, it is permissible as a matter of principle to consider alternative conduct of Apple or Google in the counterfactual of the type suggested by the respondents, any such conduct would have to have been lawful.

125    Now at this stage of the trial, the class applicants are only required to prove that the contravening conduct caused the commissions to be inflated and materially higher than they would have been in the counterfactual. That task is met by assessing the counterfactual commissions.

126    Now I agree with Apple and Google that it is problematic to suggest that loss is to be ascertained by just comparing the factual with a counterfactual in which the only adjustment is subtraction of the contravening conduct. That approach does not answer the question of what position the claimant would be in had the contravening conduct not been engaged in. That is different from just subtracting the contravening conduct and assuming, contrary to real commercial likelihoods, that everything else would be held the same.

127    In my view it is necessary to bring to account in an analysis of causation all those facts which go hand-in-hand with the supposition that the contravening conduct was not engaged in.

128    Further, I should make one other point. The decision in Australian Consumer and Competition Commission v Cement Australia Pty Ltd (2017) 258 FCR 312, relied on by Apple, is not relevant to my task of assessing whether the class applicants have established the commission overcharge.

129    As Epic correctly points out, that case concerned the principles applicable to the imposition of pecuniary penalties under s 76 of the CCA. The ACCC argued that the trial judge erred in the assessment of the appropriate penalty under the Act by failing to precisely quantify the loss and damage suffered as a result of the wrongful conduct. In support of its argument, the ACCC contended that the Court’s consideration of “any loss or damage suffered by the act or omission” as a relevant factor in the exercise of its power under s 76 ought to have been informed by the principles relevant to the assessment of damages under s 82, and in particular, the necessity of estimating the value of loss, however difficult.

130    That argument found no purchase with the Full Court, primarily on the basis of the obvious conceptual differences between s 76 (deterrence) and s 82 (compensation). To this end, Middleton, Beach and Moshinsky JJ noted that if the ACCC’s position was to be accepted, it was for the regulator to have adduced such evidence necessary to enable the trial judge to make the appropriate quantification of damage or market harm, and it failed to do so. It was in this context, that is, a fallback position from its primary (and correct) decision that precise quantification of loss was not necessary in a penalty context, that the Full Court made some general observations about economic methods of calculating a but-for price in competition cases.

131    Now it is worth setting out what Middleton, Beach and Moshinsky JJ said (at [497] to [509]):

The principles developed in connection with the assessment of damages, namely, that where it is clear that substantial loss has been incurred, the fact that an assessment is difficult because of the nature of the damage is no reason for awarding no damages or merely nominal damages, are not directly comparable with the present context.

First, the assessment of damages is a different exercise to the imposition of a civil penalty. The Court is principally concerned in the present context with general and specific deterrence. The perspective of putting a party in the position that they would have been in if a contract had been performed, or with putting an injured party in the position that they would have been in but for the statutory contravention or the tortious conduct has little to do with the matter.

The imposition of a penalty is affected by many questions which are then balanced as a matter of intuitive synthesis. The amount of loss or damage causally connected to the contravening conduct is only one of the relevant factors. Moreover, the exercise of intuitive synthesis is an evaluative exercise. To suggest that there must be quantitative precision on any one factor feeding into the question of what a penalty should be is misplaced. Contrastingly, in the assessment of damages or the value of a lost opportunity, some monetary quantification resulting from the process of some degree of estimation, albeit imperfect, must be made (see for example Sellars v Adelaide Petroleum NL at 350-355 and 368; Placer (Granny Smith) Pty Ltd v Thiess Contractors Pty Ltd (2003) 77 ALJR 768; 196 ALR 257 at [38]; and Malec v JC Hutton Pty Ltd at 642 and 643). And indeed that quantification is the singular goal of such a compensatory exercise. But in setting a penalty, the goal is to fix a monetary amount for the penalty sufficient to address the principal objective of deterrence, rather than to quantify loss or damage as compensation.

…

Further, if the ACCC’s principle was to be accepted, contrary to our view, it was a matter for the ACCC to have adduced such evidence as was necessary to enable the primary judge, to the requisite standard (s 140(2) of the Evidence Act 1995 (Cth)), to make the appropriate quantification of damage or market harm. It failed to do so. It did not seek to apply and justify quantification based upon any sufficiently probative method. It was not appropriate for the ACCC to say in effect to the primary judge or contend before us that his Honour had to do his best with the ad hoc and less than systematic approach which the ACCC advocated. In elaboration, we would note the following.

First, the only expert evidence concerning market harm relied on at the penalty hearing was from Dr Philip Williams, who was called by the respondents. The ACCC chose not to adduce its own expert evidence on the subject. Dr Williams’ evidence was to the effect that the matters relied upon by the ACCC did not provide a sound economic basis for quantifying the effect of the conduct. In the absence of contrary and probative evidence, his Honour was entitled to accept and act upon Dr Williams’ opinions.

Second, it may have been possible for the ACCC to probatively measure the benefits derived or the market harm suffered even if the identity of the entities that might have entered the market was unknown. Dr Williams referred to economic literature that suggested there were various methods that could have been employed to quantify market harm: Penalty Reasons, [641]. But the ACCC chose not to rigorously engage with any such methodologies. Moreover, its criticisms of the primary judge’s conclusions as to the lack of probative evidence (see for example at [818](14)) are unwarranted, particularly in the context where it only chose to put before his Honour its own, if we may say so, piece-meal analysis.

We accept that estimating damage from anti-competitive behaviour of the type found by the primary judge in the present case is difficult. The question of making such an assessment usually involves in an economic context a “but for” analysis. But in a legal context where compensatory loss is being assessed, sometimes a “but for” approach may not be suitable or comprehensive. Section 76(1) in using the phrase “any loss or damage suffered” may embrace but is clearly not confined to the lens of compensatory loss. It is to be analysed in part in its economic context. And for economists, a “but for” analysis is usually appropriate. The question normally posed in such a framework is: what would the price of the relevant product have been absent the anti-competitive conduct? This is compared with the actual price to determine loss. Of course, such an approach may not consider other types of loss such as harm done arising from the reduced use of the relevant product because of the higher price that may also have to be considered. But the issue principally posed for economists is how to calculate the “but for” price. Various but problematic methods may suggest themselves (see Brander JA and Ross TW, “Estimating Damages from Price-Fixing” in Pitel S (ed), Litigating Conspiracy: An Analysis of Competition Class Actions (Federation Press, 2006) at 335), including:

(a)     making before and after comparisons of prices;

(b)     using marginal cost or average cost as a proxy for price;

(c)     using prices in similar markets as analogues; and

(d)     undertaking econometric simulations to determine a competitive price benchmark.

But before the primary judge the ACCC did not put forward any probative analysis using any of these methods, although it did suggest rough and ad hoc approximations utilising some aspects of methods (a) and (c).

Third, in relation to some aspects of its analysis the ACCC relied upon what the respondents not unfairly described as a “patchwork quilt of figures” from budget assumptions or forecasts prepared by different personnel of the respondents at different times. But these were premised on counterfactuals and other assumptions in part correctly rejected by the primary judge. Moreover, the ACCC inappropriately equated perceived benefit with market harm. The ACCC’s approach was correctly rejected by the primary judge.

Fourth, …. Moreover, given that a comparison was required between the actual position of the respondents with the counterfactual position where the relevant contracts did not contain the impugned provisions (rather than the total loss of the contract(s) to a third party), it was incorrect for the ACCC to assert that the whole of the EBIT should be taken to have been causally connected to the contravening conduct. Indeed and more generally, in our opinion the ACCC never properly engaged with his Honour’s statement at [489] of the Penalty Reasons (set out at [313] above).

…

Finally and as a general observation, we of course accept that the mandatory consideration of loss does not necessarily require evidence from victims or expert evidence on a “but for” basis (Reckitt Benckiser at [93]); moreover, in some cases an assessment can be made even absent precise evidence on causation or mathematical precision. The primary judge was required to consider and analyse the information before him, and draw any necessary but reasonable inferences open to him in making some assessment. But to say as much is not to say that his Honour was in error in failing to make a quantification of the type contended for by the ACCC. We would also note that the forensic context that his Honour was dealing with was quite different and considerably more complex than that discussed by the Full Court in Reckitt Benckiser at [98]. Depending upon the context and the exercise, more or less precision in the assessment may be justified. At the end of the day the task is principally an evaluative exercise for the trial judge...

132    None of these observations assist Apple or Google in the present context.

133    Moreover, as I said to counsel in argument concerning hypothesised cartel behaviour involving price fixing, which is not the present context of course,:

… you can have the hypothesised cartel period, and you look at the prices in that period, and then compare them with the prices before the cartel started and/or the prices after the cartel is over, and you run multiple linear regression calculations, and you use a dummy variable for the within period cartel prices, and you’re able to work out through linear regression analysis what the price has done during the cartel period and causally relate that to the collusion as distinct from just other market or nonmarket factors. And that’s a very standard way that sort of thing is done in the US. Now, of course, we’ve got something much more complicated.

The counterfactual commission — introduction

134    Let me now introduce in more detail the class applicants’ cases against Apple and Google concerning the counterfactual commission(s).

135    Mr Holt calculated a counterfactual commission range that would apply across the relevant markets.

136    As he said in the concurrent evidence session in the Apple class action, what he measured was the effective average that would apply across the market. That could include some app stores charging higher amounts and some charging lower and it also allows for differentiation on quality. Further, it could allow for some app categories to have higher rates and some to have lower. Moreover, he said that there was nothing in his analysis that precluded any degree of individualised outcomes of that nature.

137    Similarly, in the concurrent evidence session in the Google class action, Mr Holt said that his analysis did not seek to estimate what the Play Store specifically would have done, but was a wider assessment of the overall cost of distribution in the Android app distribution market. He said that his range was not intended to cover all individual transactions, but the effective average, and he also said that his approach did allow for product differentiation.

138    Mr Holt’s counterfactual commission range was also an effective average range across the relevant period, rather than the settled range that would have been arrived at or after a period of competition within the relevant period. Mr Holt said that what he had done was to identify an effective overall average that he expected would apply over the relevant period.

139    Mr Holt explained the following in an exchange with me:

HIS HONOUR:     But that would change in terms of dynamics and how new entrants came in. So, if low quality, what they were doing etc, and how the incumbent reacted to those new entrants. How the incumbent might want to back out the potential for that new entry or threat to grow. But that might all occur dynamically – do you follow what I mean? So in a counterfactual, I just wonder whether I should be looking at one rate at all – some unitary, you know, single rate or I should be looking at it more dynamically over time.

MR HOLT:     I think either are potential alternatives, I think, would be open to you. The approach that I’ve adopted is to look at this as an effective average over the relevant period.

140    Mr Holt’s evidence was that having regard to the relevant comparators, a competitive level of commission in a competitive iOS app distribution market and a competitive Android mobile app distribution market would be in the range of 10 to 20%, with a most likely level of 15%.

141    Mr Holt said that the counterfactual commission of 10 to 20%, with a most likely level of 15%, includes a component attributable to the use of in-app payment solutions in the range of 2.2 to 10% of the relevant purchase. He said that this is consistent with Apple’s current monetisation strategy where the commission charged on paid app downloads and in-app purchases is said to include the value related to payment processing and related activities. And he said that it is also consistent with Google’s monetisation strategy, where payment processing services are included in the commission charged on paid app downloads and in-app purchases.

142    Further, if it is necessary to identify a precise figure for the counterfactual commissions, Mr Holt considered that, based on a comparator analysis, an average effective rate of 15% is the most likely commission rate that app developers would have paid but for the contravening conduct.

143    Further, if Apple were to adopt the counterfactual commission at the lowest end of the range advanced by Mr Holt, being 10%, Mr Holt considered that it would remain profitable on an operating income basis even if the App Store’s market share dropped to 25%. Similarly, at a market share of 25%, the App Store’s return on assets would remain in excess of 100% and it would continue to generate an infinite return on capital employed. The App Store would therefore remain highly profitable at the counterfactual commission rate.

144    Further, Mr Holt said that adopting the conservative assumption that the size of the Android mobile app distribution market would remain the same following new entry absent the contravening conduct, the Play Store would remain profitable charging a commission of more than 10% and with a market share as low as 25%.

145    Let me now turn to the detail and begin with the Apple context.

General – Apple context

146    In this section of my reasons I will address the class applicants’ case concerning Apple on the question of overcharge including the question of counterfactual commissions.

147    Now it is convenient to address four topics in the following sequence.

148    First, I will address the question of App Store profitability, which of course I have addressed in more detail in my reasons in Epic v Apple. This topic is relevant to and is an input into the second to fourth topics.

149    Second, I will address the counterfactual scenarios and in this context principally address the appropriate retrospective hypothetical scenarios as they are relevant in the class action context. The future with and without test relevant to whether there have been any statutory contraventions by Apple has already been addressed in my reasons in Epic v Apple.

150    Third, I will then address Mr Holt’s comparator analysis. Now this topic could have been addressed before the second topic, but I have found it convenient to proceed in the sequence that I have outlined.

151    Fourth, I will finally address the output of this analysis which relates to the counterfactual commissions.

App Store profitability – Apple context

152    Now the absence of entry in the face of high profitability is some indication of barriers to entry and some indication of substantial market power.

153    Let me turn to this question of profitability, which I have discussed elsewhere, and begin with a summary of Mr Holt’s evidence in that regard.

154    Mr Holt said that the most useful metrics for assessing market power or whether profits are supra-competitive are asset-based metrics such as return on capital employed (ROCE) or return on assets (ROA). And he expressed the view that in a market characterised by competition, any excess returns above a company’s weighted average cost of capital (WACC) would be expected to be eroded over time.

155    Now as I have indicated in my reasons in Epic v Apple, Professor Davila concluded that the App Store exhibited high profitability in the period FY2017 to FY2019, with infinite returns on its ROCE. Professor Davila calculated the capital employed by looking to the assets attributable to the App Store less its short-term debt as these appeared on Apple’s balance sheet, concluding that the App Store’s returns were very attractive both in absolute terms and in terms relative to other companies. Mr Holt agreed with Professor Davila’s approach in this regard.

156    Further, Professor Wright said that the App Store’s profit levels enabled it to contribute to large fixed costs of Apple, being consistent with a conclusion as to it generating supra-competitive profits and an outcome that would not be possible if it was subject to close competition. Mr Holt accepted this position.

157    Mr Holt also said that Professor Davila’s calculations in relation to the App Store’s profitability were robust to alternative assumptions regarding the allocation of cash to the App Store’s balance sheet, the accuracy of the accounting data sources used by Professor Davila, and the allocation of a proportion of Apple’s research and development expenses.

158    Mr Holt adopted a cash retention method as an alternative approach to the allocation by revenue method. So, the App Store was presumed to retain the cash that it generates in any given year before remitting it to Apple on the first day of the following financial year. According to Mr Holt, the cash retention method still showed an ROCE in the range of 178% to 189%, which represented very high returns. It also showed an ROA of approximately 87%, which again represented very high returns.

159    Further, Mr Holt adopted two alternative assumptions, where billings and revenues were 50% lower than those on which Professor Davila relied, and expenses including opex were 50% higher than those on which Professor Davila relied. He found that even under these assumptions Apple remains strongly profitable. So, even if there were inaccuracies or discrepancies in the financial statements relied on by Professor Davila, Mr Holt’s view was that it was unlikely that there would be any material impacts on the ultimate conclusion that the App Store exhibits very high profitability.

160    Mr Holt concluded that despite being more conservative than the approach taken by Professor Davila, the cash retention method still showed an ROCE in the range of 178% to 189% and an ROA of approximately 87%, both of which represent a very high return.

161    Mr Holt said that the App Store would have remained profitable had it absorbed as much as 60%, or capitalised and amortised as much as 80%, of Apple’s R&D expense in FY2018. This conclusion was arrived at in circumstances where Mr Holt’s view was that such a large proportion of Apple’s R&D expenditure would not have been allocated to the App Store in light of the extent of Apple’s other businesses and the nature of the R&D expenditure.

162    Further, Mr Holt accepted the approach adopted by Professor Davila with respect to the allocation of common costs and observed that Apple itself has regard to the allocation of common costs in undertaking internal profitability analyses, regardless of the broader purpose for which that internal analysis is undertaken. Further, Mr Holt agreed with Professor Davila’s use of an allocation by revenue method to attribute assets to the App Store from a consolidated balance sheet.

163    Mr Holt concluded that the profitability outcome was not sensitive to a variety of cost allocation sensitivities, leading him to conclude that the App Store exhibits a high degree of profitability, regardless of the ultimate degree of cost allocation undertaken. He considered that supra-competitive profits would indicate that there is scope for new entry and the engendering of a reduction in the commission rates likely to be paid in the counterfactual.

164    Now Dr Majumdar was critical of Professor Davila’s approach to attributing value to intangible assets utilised by the App Store, although he did not seek to precisely identify those assets. He also did not suggest an appropriate methodology for valuing such assets or an appropriate value for those assets.

165    Further, it is to be noted that Dr Majumdar’s challenge to the reliability of Professor Davila’s ROCE analysis for the App Store was also underpinned by his opinion that the ROCE analysis was flawed due to the accounting analysis failing to accurately account for the App Store’s intangible assets in its asset base.

166    Now as the class applicants explained, Mr Holt accepted that intangible assets should be accounted for, but rejected Dr Majumdar’s assertion that the relevant intangible assets included assets other than intellectual property assets. He concluded that relevant intangible assets should relate to costs actually and necessarily incurred from the running of Apple’s business in general and which create an asset separate from any of those arising from the general running of the App Store, but that was not the case here.

167    Mr Holt said that one way to include intangible costs in the App Store’s asset base is to capitalise R&D costs, which is what Professor Davila did.

168    Further, Mr Holt’s sensitivity allowance demonstrated that his conclusions as to the high profitability of the App Store were, similarly, not sensitive to any realistic attribution of the appropriate value of intangibles to the App Store.

169    Mr Holt agreed with Professor Davila’s conclusion that the App Store exhibited high profitability in the period 2017 to 2019. Mr Holt concluded that applying Professor Davila’s assumptions to the data and estimating certain datapoints that are not available for the period, the App Store remained highly profitable throughout the relevant period. In his view the analysis showed that Apple’s returns were significantly and persistently higher than the cost of the assets or capital deployed in earning those returns.

170    Further, in Mr Holt’s view, a high level of profitability would normally attract market entry. And in his view this would have resulted in downward pressure on commission rates paid by iOS app developers.

Analysis

171    Now I agree with Apple that neither market power nor anti-competitive conduct can be established on the basis of profitability alone, and a firm that innovates and gains a competitive advantage may earn higher ROCE for the period that it is able to sustain that competitive advantage.

172    I have indicated elsewhere that App Store accounting margins, even if correctly measured, are not definitive concerning monopoly power. Further, I also accept that firms that possess significant intangible assets and invest heavily in intellectual property and innovation often have meaningful accounting profit margins.

173    As I discussed in my reasons in Epic v Apple, high accounting profits do not demonstrate the existence of market power, nor its exercise or misuse. And I also agree with Apple that whatever limited relevance profitability could have in relation to those matters, its role in the calculation of a but for price is questionable.

174    But in any event, I agree with Apple that Mr Holt’s profitability calculations are unreliable.

175    Now Mr Holt had regard to Professor Davila’s profitability calculations. He agreed generally with Professor Davila’s approach and carried out further profitability calculations based on alternative assumptions. But I agree with Apple that in doing so, Mr Holt did not grapple with some objections to Professor Davila’s approach.

176    His calculations did not address the point made by Apple’s accounting and economic experts, which was that in order to meaningfully consider the profitability of the App Store, it would be necessary to consider what a stand-alone app store would be required to pay Apple for access to the Apple assets that are necessary to operate for the App Store, and for access to Apple’s customer base.

177    Dr Majumdar said that if the App Store were to operate on a truly standalone or independent basis, it would need to pay Apple to use the relevant assets or it would have to incur the cost of developing its own equivalent assets. Whether or not Apple fully recovers its iOS costs is not relevant to whether Apple would grant this use for free to an independent or truly standalone App Store.

178    But as Apple points out, the problematic premise of the class applicants’ case is that the Apple respondents would grant access to its platform for free.

179    Now Mr Holt accepted that his calculations assumed that the App Store has access to Apple’s customer base for free, and he argued that in the counterfactual Apple would not charge a fee for access to iOS and iOS users. But as Apple points out, the commercial unreality of that assumption can be illustrated by reference to subsequent events.

180    As Apple points out and as I have mentioned elsewhere, in Europe, where Apple has been required to allow alternative app stores in response to the Digital Markets Act, it has imposed a “Core Technology Fee” of €0.50 for each “First Annual Install” of an alternative app store. This occurs the first time in a twelve-month period that a user installs the app on iOS. In other words, when required to allow alternative app stores, Apple has decided to charge, essentially on a per-customer basis, for access to its technology and iOS customer base.

181    Further, the class applicants say that Apple’s experts have not posited a value for the intellectual property that they say is lacking. But I agree with Apple that this is an attempt to reverse the onus which rests with Epic and the class applicants to establish their allegations of market power and to make good their contentions about profitability.

182    In any event, whilst internally generated intangible assets cannot be valued with precision absent a sale, it is possible to estimate the value of Apple’s internally generated intangible assets.

183    As Apple pointed out, Mr Samuel said that the market capitalised value of Apple’s equity is currently approximately USD 3 trillion. Its accounting value for equity in its financial statements as at 25 September 2021 was only USD 63.1 billion. In his opinion, the difference is mostly attributable to the value of its internally generated intangible assets. I note that the 30 September 2023 figure was USD 62.1 billion.

184    Both Professor Davila and Mr Holt treated the App Store as having free access to Apple’s customer base and internally generated intangible assets, with the exception of accounting for certain recent R&D expenses.

185    Moreover, whilst Mr Holt acknowledged that the assessment of the App Store’s profitability could have regard to any properly assessable cost of generating the intellectual property on which the App Store relies, he observed that there is insufficient information to properly assess either such rights held by Apple that is relevant to the App Store or the cost to Apple of generating such rights.

186    I agree with Apple that if a profitability assessment based on asset-based measures is to be useful for determining whether commission rates are supra-competitive or not, the asset base must be measured reliably.

187    And I agree with Apple that on any view, the value of Apple's internally generated intangible assets is very substantial, and Apple would not provide an app store that was not part of the Apple ecosystem, and generating revenue for Apple, with free access to them.

188    Let me say something about Mr Holt’s sensitivity analysis.

189    As the class applicants explained, Mr Holt calculated the App Store’s profitability under various counterfactual scenarios, and concluded that, on his calculations, the App Store would be highly profitable in any scenario in which it charged a commission of between 10% and 20% and enjoyed a market share of between 25% and 100%.

190    But I agree with Apple that even if Mr Holt’s calculations were reliable, his sensitivity analysis would not demonstrate that Apple has market power or has engaged in anti-competitive conduct.

191    In order for profitability metrics to provide meaningful insight, they should be able to distinguish between competitive and supra-competitive commission rates. But Professor Davila’s profitability metrics were insensitive to changes in the App Store commission rate and so could not distinguish between competitive and supra-competitive commissions.

192    Further, Dr Majumdar explained that he had a concern with Mr Holt’s sensitivity tests, because on Mr Holt’s calculations, a 10% commission rate would give rise to infinite ROCE and ROA and ROGM metrics that are substantially in excess of Professor Davila’s comparators. Dr Majumdar explained that, for those reasons, Mr Holt’s calculations do not provide an economically useful profitability metric. I accept that evidence.

193    Dr Majumdar explained that he did not regard Mr Holt’s capitalising and expensing of additional R&D expenditure for the App Store as a meaningful sensitivity test, for essentially the same reason that he considered Mr Holt’s other sensitivity tests unreliable. Rather than adjusting the commission rate, and thus revenue, capitalising and expensing of additional R&D expenditure holds revenue constant and adjusts costs. Dr Majumdar said that “it sounds like a reasonable sensitivity to allocate quite a lot of R&D to the App Store”, but it is not a useful sensitivity “if you’re still allocating it to a model that is demonstrably showing upward bias and then essentially just getting to the same outcome”. Again, I accept what Dr Majumdar said on this aspect.

194    Now Dr Majumdar said he would place limited weight on Apple’s internal documents when assessing the economic question of market power, because of the need to take into account intangible assets and a notional transfer price. He said that firms may compare the profitability of divisions on a consistent basis to assess trends over time, but said that this is different from an economic assessment of profitability. I agree with these observations.

195    Further, Dr Majumdar explained that, when commission rates said to be competitive are found to be supra competitive, that suggests a strong bias upwards in the profitability metrics. Having found that Professor Davila and Mr Holt’s metrics were indicating that allegedly competitive commission rates were supra-competitive, he considered why that might be and identified three potential causes of the upwards bias.

196    First, Professor Davila and Mr Holt placed no weight on rights that have economic value, being the rights to make iOS and the iOS user base available to developers and to charge for that service.

197    Second, they failed to account for important intangible assets and to consider the modern equivalent value of those assets rather than their accounting book value.

198    Third, there were economic challenges of cost allocation within an ecosystem.

199    As to this third matter, Dr Majumdar presented a hypothetical example to illustrate why cost and asset allocations between complementary products in an ecosystem are arbitrary from an economic perspective and explained that, because of the arbitrary allocation of costs, revenues and assets, Professor Davila’s estimated profitability metrics were not meaningful from an economic perspective.

200    Let me summarise some further aspects of Dr Majumdar’s evidence as drawn out in substance by Apple. I should say that on some aspects Mr Houston gave similar evidence to Dr Majumdar but it is not necessary to set out Mr Houston’s evidence for present purposes.

201    Dr Majumdar explained that in principle, to calculate the App Store’s profitability, one would need to consider the difficult question of the allocation of intangibles and the difficult question of what the transfer price is that should be set, and then do a sensitivity analysis. But Dr Majumdar’s evidence was that it is not possible to do a valid sensitivity test in the present context because that would require knowing the upper and lower bounds of the values.

202    In relation to the value of the relevant intangible assets, Dr Majumdar referred to Mr Holt’s sensitivity tests which allocated about $15 billion of R&D to the App Store; other figures were also referred to (up to $14.236 billion and up to $18.979 billion) as allocated to the App Store.

203    But contrastingly, Dr Majumdar said that “on the other hand intangibles could be hundreds of billions or even trillions on the basis of market capitalisation” and concluded that “the range is so broad that actually this sensitivity testing is incapable, to my mind, of giving us the relevant outcome”.

204    Now Dr Majumdar agreed that in considering the value of Apple’s intangible assets by reference to Apple’s market capitalisation, it would be necessary to disregard any amount that reflected market power if any.

205    But he said that even if one was to consider that some part of Apple’s market capitalisation is a reflection of the App Store’s market power, it is not plausible that the entire difference between Apple’s accounting value for equity and market capitalisation is attributable to that posited market power and thus it is necessary to bring the value of Apple’s intangible assets to account.

206    Dr Majumdar made a similar point in relation to properly measuring intangible assets. In relation to the proper measure of return on sales, he said that if the asset base cannot be properly measured, that impacts the return on sales measure because return on sales will include amortisation of the asset base. Dr Majumdar’s evidence was that, from an economic perspective, amortisation applies to intangible assets.

207    Further, Dr Majumdar’s evidence was that the profitability evidence is not reliable enough to draw any conclusions from. He said that the apparent strong upwards bias in the profitability metrics means that it would not be safe to rely on them in the assessment of whether commission rates are supra-competitive or would otherwise incentivise market entry.

208    Dr Majumdar emphasised that reliable profitability measures can be useful for distinguishing between competitive and supra-competitive commission rates and also for assessing the prospect of market entry.

209    He explained that if profitability measures are to shed light on whether commission rates are competitive, competitive commission rates should give rise to profitability levels that fall comfortably within a range obtained from properly defined benchmarks, whereas supra-competitive commission rates should lead to profitability levels that are substantially above the benchmark range.

210    Similarly, if profitability metrics are to be evidence of the prospect of market entry, the allegedly competitive commission rate should give rise to profitability metrics that fall comfortably within the range of comparators considered, whereas the allegedly supra-competitive commission rates should give rise to profitability metrics that lie far above the comparators considered.

211    Dr Majumdar explained that if a profitability test always finds high profits even at commission rates that are so low that it could not be claimed that they are above competitive levels, then the test fails economic principles. He gave the example that if the App Store set a commission rate of just 4%, Professor Davila’s analysis would find infinite ROCE, and ROA and ROGM higher than Professor Davila’s comparators.

212    In his report, Dr Majumdar pointed out that Professor Davila calculated ROCE of 3,170% at a 99.9% commission rate, but infinite ROCE at a 3.6% commission rate, suggesting that Apple would exercise less market power by charging a 99.9% commission rate than by charging a 3.6% commission rate, which is nonsensical.

213    Let me deal with some other aspects of Dr Majumdar’s evidence.

214    Now Mr Holt said that even if some indicators are not sensitive to the level of the commission, the overall excess profit would fall greatly in the counterfactual meaning that competition would, indeed, be doing its job.

215    But Dr Majumdar explained that as Mr Holt was instructed to assume market power, this is essentially a statement of theory and not a direct answer to his point that Professor Davila and Mr Holt’s profitability metrics were insensitive to changes between the alleged competitive and supra-competitive rates.

216    Further, Mr Holt said that the fact that, on his calculations, Apple would be highly profitable in the counterfactual was consistent with his opinion and might indicate that his counterfactual commission rate was conservative.

217    But Dr Majumdar explained that this is not an answer to the question of whether the profitability metrics are reliable or not and is not a reason to place weight on profitability metrics that are biased upwards.

218    Further, Mr Holt said that there is no evidence to suggest that the App Store’s profitability can be explained by the App Store being more efficient or innovative than its rivals.

219    But Dr Majumdar explained that it had not been properly analysed whether there has been sufficient innovation to explain the App Store’s profitability. But in any event he said that that was not an answer to his point that Professor Davila and Mr Holt’s profitability metrics were unreliable.

220    I accept Dr Majumdar’s evidence on the points that he has made.

221    Let me deal with some other matters.

222    Mr Holt said that his assessment of the App Store’s profitability was consistent with supra-competitive profits that would attract market entry. But as I have said elsewhere, economic profits are the relevant metric governing the financial incentive for entry. The economic rate of return is the only correct measure of the profit rate for the purposes of economic analysis. And of course neither Professor Davila nor Mr Holt calculated an economic assessment of the profitability of the App Store.

223    Further, it must be accepted that an incumbent firm earning high levels of return does not necessarily indicate that entry into the market would be profitable. To the extent that Apple’s past investments in innovation have contributed to Apple’s returns, a new entrant would either earn less revenue or would need to reproduce Apple’s past investments.

224    Further, as Apple explained, the likelihood of entry is affected by the potential profitability of the entrant, and not necessarily the potential profitability of the incumbent.

225    Now it was put to Dr Majumdar that a prospective market entrant would only consider the announced accounting profits of the incumbent. But he disagreed and said that he would expect a prospective new entrant to thoroughly research any market that it was seeking to enter, rather than relying solely on the incumbent’s accounts.

226    Further, as Apple rightly said, there are other conceptual difficulties with Mr Holt’s claim that the App Store’s profits as calculated by Mr Holt and Professor Davila are evidence which support the likelihood of future market entry.

227    Mr Holt did not consider whether a competitive reaction by Apple would reduce the profitability of any market entry.

228    Further, as Apple says, neither Professor Davila nor Mr Holt considered what a market entrant’s profits would be. They were focused on Apple’s accounting profits and did not seek to make any relevant adjustments, such as accounting for fixed costs, in order to determine how Apple’s accounting profits might be related to the profit that a market entrant could reasonably expect to earn.

229    Now the class applicants have said that Dr Majumdar did not precisely identify the intangible assets used by the App Store or an appropriate way to value those assets.

230    But as Apple says, Dr Majumdar’s evidence was that to undertake an economically meaningful calculation of the App Store’s profitability it would be necessary to consider what the App Store would spend to acquire the right to make iOS and iOS users available to developers and the right to charge developers for using iOS and the iOS user base.

231    Now I agree with Apple that although Dr Majumdar did not undertake that calculation himself, he cannot be criticised for that in circumstances where Professor Davila’s evidence was that, in the specific case of the App Store, calculating the profitability of the App Store as if it were a separate company would involve making assumptions that would render the exercise uninformative.

232    Now as I have said elsewhere, I accept that Apple has substantial market power in the relevant markets. But Mr Holt’s analysis as to the App Store’s profitability does not take the matter far.

The counterfactual scenario(s) – Apple context

233    Before dealing with the counterfactual commission question directly, let me make some observations relevant to the broader potential elements of any realistic counterfactual scenario or scenarios.

234    The class applicants’ case is that but for the contravening conduct, alternative app stores would have entered the iOS app distribution market and competed with the App Store for the provision of services enabling the distribution of iOS apps. This is said to be so for three principal reasons.

235    First, the presence of widespread multi-homing on both the developer and user sides would have reduced the barriers to entry created by indirect network effects and encouraged entry by new app stores.

236    Second, the likelihood of new entrants is supported by the presence of such app stores in the market prior to and despite Apple’s contravening conduct, and by the stated intention of other entrants to enter in the absence of the contravening conduct.

237    Third, there are numerous distribution platforms for applications on other devices, such as PCs, where competition is not constrained by Apple’s conduct, suggesting that a similar outcome would prevail in the iOS app distribution market in the absence of the contravening conduct.

238    Now any counterfactual is affected by indirect network effects and multi-homing.

239    As I have explained elsewhere, indirect network effects occur when the value that users on one side of the platform attach to the services provided by the platform increases with the number of users on the other side of the platform. An individual’s incentive to join a platform depends on the size of the user base on the other side of the platform. The optimal pricing strategy is influenced by indirect network effects, because demand on one side of the platform affects demand on the other side of the platform. For a new entrant to succeed, it would need to overcome the indirect network effects from which incumbents benefit. But new entrants are more likely to succeed if at least one of the platform markets is characterised by multi-homing.

240    Multi-homing reduces the cost of switching between platforms because it allows users to join a new platform without forgoing access to their current platforms. New app stores can successfully enter without the need to offer a full suite of apps, and it can be less costly for new platforms to attract users.

241    Given the substantial benefits of multi-homing to iOS device users, such as greater app availability, one would expect that but for the contravening conduct a significant number of iOS device users would multi-home across different app stores and distribution channels such as direct downloads. Similarly, on the iOS app developer side, one would expect that absent the contravening conduct a significant proportion of iOS app developers would multi-home.

242    Now I agree with the class applicants that prior to and since the advent of the App Store, other app stores have demonstrated a willingness and desire to exist in the iOS ecosystem. But due to Apple’s contravening conduct, they have been unable to. I have discussed elsewhere Epic’s position and strategy concerning the Epic Games Store on mobile device platforms.

243    Further, Apple has blocked other native apps that distribute third-party apps.

244    Further, as I have touched on elsewhere, there is evidence from the PC app distribution market that in circumstances where there is an absence of the contravening conduct, a significant number of app distribution platforms are present in the market. Further, many publishers operate their own digital distribution platforms. Further, there are app stores available on Android devices which are not pre-installed by Google or other device makers.

245    Further, as the class applicants have said, app stores are economically viable at small scale, making them less capital-intensive and less risky and more likely to enter the market.

246    Further, I agree with the class applicants that the presence of multi-homing means that even if iOS app developers and users continue to see value in the App Store and its differentiated features, they remain incentivised to download and use alternative iOS app stores because the cost of doing so is low and would not prevent them from continuing to use the App Store.

247    In general, I agree with the class applicants that in the absence of the contravening conduct, the iOS app distribution market would be characterised by a significant number of alternative app stores.

248    Now it was said by Apple that it would respond to alternative app stores becoming available on iOS by seeking to earn a return on its investments in its intellectual property, technologies, the App Store, iOS and the iPhone, to sustain its rates of innovation and product development.

249    And it was said by Apple that such a response might entail changing the structure and level of prices, or rules that apply to the App Store in such a way that Apple seeks to both maintain its return on investment and the quality of its service or reduce the level of investment if revenue from the App Store fell and could not be restored.

250    But I agree with the class applicants that removing the contravening conduct would not prevent Apple from earning a return on its investment in the App Store if it competed effectively and offered an attractive product to the market.

251    Moreover, I agree with the class applicants that in the counterfactual Apple would be constrained from imposing higher effective commission rates, whether through changing its pricing structure, levels or rules, since developers and users would simply switch to alternative distribution channels.

252    Further, I also agree with the class applicants that in the absence of the contravening conduct there would have been competing payment solutions providers in the iOS in-app payment solutions market, attracting market share away from Apple’s IAP by offering alternatives to app developers. In turn, app developers would have taken up the lower prices offered by the competing providers, and Apple would have faced competitive pressure to reduce its commission rate or lose market share. The increased degree of competition would have led to a reduction in average commissions on in-app purchases including subscriptions.

253    Let me begin by elaborating on Apple’s points.

Apple’s arguments

254    Now Apple says that the following matters weigh against the class applicants’ case on the question of the counterfactual(s) and any overcharge.

255    First, Apple says that its commission rates are already competitive. So, in 2021, 95.6% of all developers on the App Store who have made at least one paid transaction on the App Store have access to a 15% commission rate. That includes all small developers, all video-streaming developers and developers that offer subscriptions, who since 2016 automatically attract the 15% rate after the first year of auto-renewing subscriptions. That is the 15% rate applies to nearly all developers. Of course it is worth noting that the small business program was introduced in January 2021, so the 95.6% figure touted by Professor Hitt and Apple is hardly a complete answer to say the least.

256    Further, Apple says that a competitive price must be assessed on a quality-adjusted basis, and Apple has a quality advantage. Indeed, users are attracted to the App Store for its quality, privacy and security features. In assessing the counterfactual world, it says that one needs to consider whether the rivals will set lower prices on a quality-adjusted basis.

257    Apple says that it is already constrained by competition in the actual world, including because app developers and users multi-home through use of web apps, developer websites and alternative app transaction platforms. It says that in the actual world, developers can monetise in ways other than in-app purchase and paid downloads, and thereby avoid payment of commission to Apple. It says that its commissions are not supra-competitive.

258    Second, Apple says that services which are presently provided by Apple in relation to the App Store which are under-priced or not charged for may form part of Apple’s revised monetisation strategy. It says that the rational assumption is that it has priced certain services but not others on the basis that it thinks is optimal. It says that in a counterfactual, Apple would lose the efficiency benefit from choosing its optimal price, and therefore the pricing structure might change. It says that that must be brought to bear in any assessment of the counterfactual.

259    Third, it says that a further over-simplification in Mr Holt’s evidence is that in a realistic counterfactual, there would be different prices for different developers for different kinds of services.

260    He said that in the counterfactual, new entrants would likely offer differentiated services and charge different commission rates, to cause or attempt to cause both developers and consumers to move to their site. The commissions rates they charge may be lower or higher. And the mere fact that there is multi-homing does not mean prices necessarily decrease.

261    Apple says that a new app store would likely seek to offer a lower commission rate than Apple, but the attractiveness and prospects of that entry will depend on many things, including the extent to which that lower commission was capable of attracting developers that, in turn, would set lower prices for digital content, which might be necessary for them to attract app users.

262    Apple says that any analysis that does not consider or provide empirical evidence on how developers and consumers would choose where to transact in the but for world is speculative and cannot inform whether consumers and developers would be better off in the counterfactual.

263    Further, Apple criticised Mr Holt’s analysis.

264    Now Mr Holt’s key premise for the conclusion that commissions would decrease in the counterfactual is that in reality users do not multi-home. Mr Holt also said that, but for the impugned conduct, developers would multi-home. Mr Holt then said that multi-homing increases competition between platforms, and the introduction of multi-homing would therefore have caused a significant reduction in commission rates.

265    But Apple says that Mr Holt’s oversimplistic analysis is based on false assumptions. It fails to take account of the fact that both developers and users can and do multi-home already. End users may process the same app transaction on a different platform, regardless of whether they are using a different device or the same device.

266    So, Apple says that the relevant comparison is not between multi-homing and no multi-homing. Rather, the comparison is between multi-homing and the theorised possibility of increased multi-homing.

267    Apple says that Mr Holt did not grapple with this issue. He said that a benefit of multi-homing is that users can procure apps that are not available on an entrant’s store from other sources. But these benefits of multi-homing already exist, because users can, and do, access apps and digital content on a variety of platforms, including as a result of cross-wallet and cross-platform functionality. Similarly, Mr Holt said that multi-homing would mean that to the extent that a transaction can be carried out on different platforms, they can choose which platform to use for the transaction. But again, Apple says that this can occur now.

268    Further, Apple says that there would be product differentiation in the counterfactual world, and Apple would be at the top end of the quality spectrum in the counterfactual, as it is in the factual.

269    Now Mr Holt said that a range of 10 to 20% for the purported counterfactual commission rate can accommodate both lower and higher quality platforms.

270    But Apple says that given that Apple is, and in the counterfactual would be, a higher quality platform, even adopting Mr Holt’s suggested range of 10 to 20% it is likely that Apple would sit at the upper end of that range.

271    Further, developers need to use Apple’s software to develop their apps. The licence in clause 2.1 of the DPLA allows developers to install the Apple software including SDKs and APIs provided to them under the program, which means Apple’s development, testing, digital signing and distribution program, contemplated by the DPLA. The licence is subject to all the terms and conditions of the DPLA. Clause 2.6 provides expressly that no other use of the Apple software is permitted. Apple says that the class applicants do not grapple with Apple’s legal right to restrict the licence of its software in this manner, that is, to restrict the purpose for which the software is used.

272    Apple also says that the class applicants make the unrealistic assumption that Apple would broaden the licensed rights, without increasing the licence fee.

273    Let me address another topic.

274    The class applicants said that absent the alleged contravening conduct there would have been a credible threat and likely entry into the in-app payment solutions market.

275    But Apple says that there is no significant distinct demand from developers under reasonable pricing conditions to use alternative payment solutions.

276    Apple says that the fact that some developers have chosen to use alternative payment solutions on the Epic Games Store does not inform the analysis, because the evidence was that Epic’s payment system was of inferior quality. Developers including Ubisoft wanted their own payment processing system on the Epic Games Store because it was unable to integrate with EDP, which was a difficult and tedious system to use and needed improvement.

277    The class applicants said that other app developers including Facebook, Spotify, WeChat, TradeMe and Cho Tot have asked to use alternatives to Apple’s IAP.

278    The class applicants said that Paddle has received more than 1,000 enquiries from developers wishing to use Paddle’s system in preference to Apple’s IAP. But Apple says that mere registration of interest at a preliminary stage does not equate to a preference for Paddle’s system over Apple’s IAP.

279    Moreover, Apple says that whilst Paddle has received a number of enquiries from app developers about whether the Paddle platform can be used to support in-app purchases of digital content within iOS native apps, only a small number of enquiries are expected to convert to actual sales.

280    Further, Apple says that the evidence from the payment experts was that if developers have the ability to select a payment solution or billing system, small developers are more likely to have a preference for a “one-stop shop” that would involve a consistent user experience across mobile apps.

281    Now the class applicants relied on Epic’s historical consideration of the possibility that Epic may offer EDP as a stand-alone solution in support of the contention that EDP is a credible potential entrant in the counterfactual. But Apple says that it is not likely for the following reasons.

282    First, it says that EDP is an inferior payments system.

283    Second, it says that EDP is used in the Epic Games Store and for transactions in Epic’s Unreal Engine marketplace. It is not a separate product that Epic sells to other developers for use outside of the Epic Games Store.

284    Third, it says that Mr Sweeney expressly said that Epic has not made a decision to release EDP as a stand along payment solution.

285    Fourth, it says that there is no evidence of any work being undertaken by Epic to offer it as a separate offering to game apps.

286    Further, Apple says that the user experience of alternative in-app payment solutions would likely have an adverse effect on the user experience.

287    Apple says that users will lose material benefits from IAP including a better and more seamless payment system by which users can transact using credit card details linked to their Apple ID and without having to share those details with developers, the ability to view a comprehensive record of all transactions recorded against the user’s Apple ID, the ability to manage subscriptions in one place, including to cancel subscriptions and avoid “subscription traps”, without having to directly contact the developer, inability to engage in family sharing or utilise parental controls on purchases by children, and inability to get in-app refunds.

Analysis

288    I agree with the class applicants that but for the contravening conduct during the relevant period, there would have been a credible threat and likely entry of direct distribution methods and alternative distribution sources, comprising alternative app stores accessed by direct download.

289    In my view, alternative app stores would likely have entered the iOS app distribution market and would have competed with the App Store for the provision of services enabling the distribution of iOS apps. The class applicants have made the following points that I largely agree with.

290    First, as the class applicants have said, the likelihood of new entrants into the market is supported by the attempts of other app stores to enter the market, and by the stated intention of other entrants to enter the market in the absence of the contravening conduct.

291    Prior to and since the advent of the App Store, other app stores have demonstrated a willingness and desire to exist in the iOS ecosystem.

292    As I have indicated in my reasons elsewhere, Epic’s long-term strategy has been to launch the Epic Games Store on mobile device platforms. Since the Epic Games Store was launched in December 2018, it has not been possible for Epic to launch a mobile version on iOS devices because iOS devices users are not able to download the app onto their devices through a web browser. Epic has had the intention of launching a mobile version of the Epic Games Store on iOS since at least 2019, but could not do so due to Apple’s policies.

293    If Apple removed its restrictions, Epic would launch the Epic Games Store on iOS. Indeed, in response to Apple’s decision to allow sideloading of alternative app stores on iOS devices in the EU following the passage of the Digital Markets Act, the evidence at trial was that Epic will launch the Epic Games Store on iOS.

294    Apple has blocked other native apps that distribute third-party apps. There is evidence of intended entry or announcements from these platforms, in addition to the Epic Games Store.

295    Second, as I have also explained elsewhere, there are numerous distribution options for applications on other devices, such as PCs, where competition is not constrained by Apple’s conduct, suggesting that a similar outcome would likely have prevailed in the iOS app distribution market in the absence of the contravening conduct.

296    And as the class applicants point out, there is evidence from PC app distribution that in circumstances where there is an absence of the contravening conduct, a significant number of app stores are present in the market servicing Windows OS and macOS. Moreover, at least some of these app stores allow third-party app stores to be distributed through them.

297    Third, as I have explained elsewhere, there are several alternative app stores available on Android devices. This is notwithstanding that Google’s own restrictive conditions limit competition with respect to the Play Store.

298    I agree with the class applicants that the fact that these app distribution platforms have continued to operate in the market despite having very low market share is evidence that entry costs would not be so material as to prevent entry in the counterfactual. App stores are economically viable at small scale, making them less capital-intensive and risky and more likely to enter the market. These established third-party app stores would also be able to launch relatively expeditiously following removal of Apple’s anti-competitive conduct.

299    Fourth and relatedly, and consistent with the technical evidence that I have discussed elsewhere, there are no technical reasons why a third-party app store would not have been able to have been installed on iOS devices, or iOS app users could not have directly downloaded third-party apps onto iOS devices. As the class applicants correctly say, from a technical perspective, iOS devices are capable of supporting alternative app distribution methods because iOS devices at a hardware and foundational software level are similar to those other devices, such as those running macOS, that support other distribution models. iOS was built from the kernel of macOS.

300    Further, Apple has a sandboxing mechanism which preserves Apple’s capacity to maintain security over its devices irrespective of where an app is downloaded from. All apps have to be signed by an authorised party, typically Apple, and if Apple has not provided that signature the app will not run regardless of how it was installed onto the device.

301    Further, in terms of the information flow to Apple in circumstances where an app was distributed outside the App Store, the main difference would be that Apple would no longer have visibility over reviews posted by users. All other channels of monitoring would remain in place including on-device monitoring.

302    Similarly, and as I have discussed elsewhere, Apple’s changes in response to the Digital Markets Act show that it is technically possible for iOS devices to support the installation of apps obtained outside of the App Store, whether that be from alternative app stores or by way of direct download from websites.

303    For these reasons, I agree with the class applicants that in the absence of the contravening conduct during the relevant period, the iOS app distribution market would have likely been characterised by the credible threat and likely entry of a significant number of alternative app stores.

304    Now Apple said that the App Store is currently under strong competitive pressure from alternative app stores and other channels outside of iOS. So in the counterfactual, Apple says that it would sustain its commission rates. But I reject this claim, largely for the reasons given by the class applicants.

305    First, as the class applicants explain, platforms that do not offer services for the distribution of iOS apps do not constrain the App Store such that there would not have been credible and likely new entry of alternative iOS app stores in the counterfactual.

306    Users choose devices, and activities on those devices, based on usage experience and the context for using the device. User behaviour is different on smartphones and PCs, and the features of the device as well as the context determine its selection. Platforms that cannot offer services for the distribution of iOS apps therefore cannot meet a user’s expectations in terms of context for using the device and usage experience.

307    Platforms that offer services for the distribution of non-iOS apps do not operate as a real competitive constraint on the App Store to the exclusion of new entrants to the iOS app distribution market.

308    Second, as the class applicants explain, by the contravening conduct during the relevant period, Apple has not permitted the direct download of alternative app stores or other native apps on iOS.

309    Apple’s conduct has prevented those stores from distributing iOS apps. In the absence of the contravening conduct, app users would have been able to directly download iOS apps, including third-party app stores, from app developers providing competitive pressure on the App Store.

310    The evidence supports the view that, in the counterfactual, Epic would have distributed the Epic Games Store to iOS device users through direct downloading. Mr Sweeney said that Epic would invest in a marketing campaign to inform users that the Epic Games Store could be downloaded from Epic’s website using a web browser on iOS. Such an alternative distribution channel would have exerted a constraint on app store commission rates by driving them closer to a level that reflected the benefits that app developers and app users perceived relative to direct downloading.

311    Third, as the class applicants rightly point out, during the relevant period, web apps or app stores on other operating systems such as Android, or those on non-mobile devices such as Windows or gaming consoles, did not provide a significant constraint on the App Store.

312    Further, I agree with the class applicants that even if it be accepted that Apple is under competitive constraint and would be in the factual, in the counterfactual the competitive constraint on the App Store would likely have been greater.

313    Let me now descend into the detail on three particular points, being whether Apple’s commission rate is competitive, whether there would have been competition on commission rates in the counterfactual, and whether Apple would or could have imposed other fees in the counterfactual.

Apple’s commission rate in the factual

314    Now Apple said that the App Store’s commission rate has always been and is competitive and there would be no change in the counterfactual. But I do not accept Apple’s position. I have discussed this elsewhere, including in the comparator analysis in these reasons which follows this section.

315    Apple said that because commission rates have not increased over time and in some cases have decreased, total output has increased over time on both the App Store and on “rival” transaction platforms, and the quality of transactions has increased due to Apple’s investments and innovations. It was said that Apple would sustain its commission rate even if other iOS app stores charged a lower commission than the App Store. But these assertions do not withstand scrutiny.

316    First, as the class applicants say, the underlying premise of the assertion that Apple is offering a competitive commission rate in the factual is that Apple is competing in the markets as defined by Apple. But if I do not accept this premise as being the appropriate market by which to view the contravening conduct, Apple’s experts agreed in cross-examination that they have not attempted to provide me with an opinion as to what would happen in the counterfactual on the assumption that Apple is not competitive in the factual. That is, Apple’s case on existing competitive rates stands or falls on its market definition, and I have rejected that market definition.

317    Second, and relatedly, I agree with the class applicants that a lack of increase in commission rates in the past is not demonstrative of competition in a market that is characterised by only a single supplier. It merely indicates that Apple was not sufficiently incentivised to further increase its monopolist price. Further, Apple’s reduction in certain commission rates has been limited, as I have discussed elsewhere, and there is no evidence that it was as a consequence of competitive constraint.

318    Third, I agree with the class applicants that even if total output has increased, that merely reflects a growing device user base rather than competition within the relevant market. Moreover, if the increase has been brought about through innovation, as I have said elsewhere, even a monopolist will innovate if it sees it as profitable to do so.

319    Fourth, Apple also appears to contend that because it offers a 15% commission rate to developers accepted into the small business program, a counterfactual commission rate of 15% establishes that most developers have not suffered any harm as a consequence of the contravening conduct. But I agree with the class applicants that such a contention is without merit. The small business program was introduced in January 2021, some 3.5 years after the commencement of the relevant period. Further, the effective commission rate for each year of the relevant period was significantly higher than the 15% offered within the program. This indicates that the vast majority of paid downloads and in-app purchases of digital content continued to be taxed at the 30% commission rate. Further, eligibility for the small business program does not guarantee acceptance into it.

Price competition in the counterfactual

320    Now in my view in the counterfactual during the relevant period, there would likely have been competition, whether by actual or credible threat of new entry with additional constraints from directly downloaded apps, which would have put substantial downward pressure on the commissions paid by developers on paid apps and in-app purchases.

321    And one likely outcome of that competition would have been Apple reducing its commissions immediately so as to disincentivise those new credible entrants. The class applicants have made the following points which I agree with.

322    First, as the class applicants correctly contend, such new entry would have led to price competition as those app stores would have attempted to offer competitive commissions to attract developers and take advantage of network effects. And to the extent that new entrants are more efficient than the App Store, they would have been able to offer app developers a lower commission than the App Store.

323    Similarly, rival app stores would have had the incentive to attract app developers by keeping participation fees low and offering a competitive commission. In addition, this would have created pressure on the App Store to improve its efficiency. There would have been significant downward pressure on commission rates as a consequence.

324    Now one thing that new entrants could try to do is lower their prices and undercut Apple. Similarly, a new entrant app store would likely need or seek to set a lower price or offer a lower commission rate. Further, a competitive response in the counterfactual could potentially include Apple reducing the commission it charges or eliminating the commission entirely and replacing it with other sources of revenue. This would have been an economically rational response of Apple in response to the threat of new entry.

325    Second, as the class applicants point out, even the threat of entry would have likely had this effect. Apple would likely have engaged in a competitive response even if there was no actual new entrant for some time. Indeed, it is plausible that the threat of new entry would cause Apple to have engaged in a price reaction to try and make it uneconomic for a new entrant to come in or sustain itself over a period. It is legitimate for an incumbent which is earning supra-competitive profits to seek to deter entry by lowering prices to make entry to the market less attractive. Even anticipation of entry may cause a competitive response.

326    Third, I agree with the class applicants that evidence from the competitive processes in the PC app distribution market suggests that competitive price responses would have likely been rapid in the counterfactual.

327    Fourth, as I have endeavoured to point out in my reasons in Epic v Apple, Apple’s commission rate on the App Store was not set at 30% by reference to the costs that Apple expected to incur in relation to the creation and development of the App Store. Rather, the 30% rate was set in 2008 based on publicly available information about other platforms’ commission rates for digital content, absent any suggestion in Apple’s evidence that it considered that such platforms would be competing with the App Store.

328    Fifth, as the class applicants correctly contend, the potential for direct downloads in the counterfactual would have meant that app stores would need to ensure that they were offering sufficient value to avoid app developers bypassing them in favour of direct downloads.

329    Sixth, as the class applicants say, the removal of Apple’s conduct which prevents app developers from steering iOS device users to alternative payment solutions for digital in-app content would have increased pressure on Apple to lower its commission rate.

330    Seventh, Mr Houston said that ceteris paribus, Apple’s commission rates would have been no lower in the counterfactual. Mr Houston then observed that the ceteris paribus assumption is likely to be inappropriate given that, absent any competitive response by Apple, Apple would have lost revenue in the counterfactual due to increased competition. I accept the class applicants’ criticisms of Mr Houston on this aspect of his evidence without descending into the detail.

331    Eighth, I agree with the class applicants that competition on other dimensions, such as quality, would not have occurred to the exclusion of price competition. There were other means of competition in the counterfactual, beyond improvements in quality, including price. Competition is likely to have occurred on a range of dimensions, including price competition. Although a number of factors beyond price could well be relevant to competition, that does not undermine the important role that price competition would likely have played in the counterfactual. Moreover, price responses can occur more quickly in response to increased competitive intensity than a change in quality. But I do accept that in some scenarios this is not the case. Moreover, it may be the case that a firm in anticipation of competitive pressure may move first on matters of quality. But the class applicants’ position is a fair generalisation.

332    In order to compete, entering services would have needed to offer attractive attributes along dimensions that included price. If other iOS app transaction platforms were to exist in the but-for world, they would likely offer differentiated services and charge different commission rates.

333    Now in my view the counterfactual market would likely have been differentiated, with some app distribution channels offering higher quality while others offered lower quality, and with some channels offering higher commission rates while others offer lower rates. But this differentiation in the market does not alter the position that in absolute terms there would have been an overall market adjustment of commission rates downwards.

334    Ninth, as the class applicants correctly point out, to establish that there would have been price competition in the counterfactual does not require establishing that any reductions in commission would have been passed through to consumers.

335    Price competition would likely have occurred without depending on the degree to which the lower commissions would have been passed through to consumers. As the class applicants have said, this is because competition between suppliers of app distribution services and in-app payment solutions occurs at the developer level, that is, competitive rivalry between competing app stores, direct distribution channels and in-app payment solutions for developers. Whether a commission reduction is absorbed by developers or whether that reduction is passed through to consumers is not meaningful to price competition in the counterfactual.

336    Tenth, Apple’s suggestion that there may have been higher commissions in the counterfactual, using China as an example, is problematic to say the least. There are unique features of the Chinese market that make it an inappropriate comparator for the competitive counterfactual. I have discussed this elsewhere.

Would Apple have imposed other fees?

337    Now it was said that Apple could have responded to alternative app stores becoming available on iOS by seeking to earn a return on its investments in its intellectual property, technologies, the App Store, iOS and the iPhone, to sustain its rates of innovation and product development. It was speculated that such a response might have entailed changing the structure and level of prices or rules in such a way that Apple seeks to both maintain its return on investment and the quality of its service or reducing the level of investment if revenue from the App Store fell and could not be restored.

338    But even if it were permissible, it is unclear whether the fee is one charged to all developers in the counterfactual, or only developers who do not distribute via the App Store or only developers who distribute a competing app store by direct download. A developer may wish to offer direct downloads, or be available on a competing app store. Such a developer will pay a USD 99 annual developer fee for development of the app to Apple. Once developed the app is available for direct download. There is no explanation of how a variable fee could be enforced when the app is directly downloaded or made available on a competing app store. The basis of any fee is not identified.

339    The class applicants have made the following points with which I largely agree.

340    First, Apple has led little evidence as to what steps it might or would seek to implement in the counterfactual. To the extent it has led such evidence, the evidence was speculative and largely directed towards potential additional fees which Apple might have levied on developers, as distinct from alternative app stores.

341    Second, it is inappropriate to assess the counterfactual by reference to a scenario in which Apple simply deploys a different set of anti-competitive measures to have the same or similar effect to the contravening conduct.

342    If Apple imposes some charge that makes entry of alternate app stores not viable, that would have the same or similar effect as that part of Apple’s anti-competitive conduct. Any user access fee designed to replace the high profits generated by the contravening conduct would not be a realistic counterfactual, as it would distort competition by simply moving the gateway from one level to another.

343    And as the class applicants have pointed out, in the context of the profitability concurrent evidence session, Apple’s expert, Dr Majumdar, conceded that “one shouldn’t allow market – if you like, a premium for market power to be capitalised – to be included in the transfer price. I think we agree on that point.”

344    Third, as the class applicants point out, removing the contravening conduct would not have prevented Apple from earning a return on its investment in the App Store if it competed effectively and offered an attractive product to the market.

345    Fourth, as Professor Wright said in the concurrent evidence session, the counterfactual must be informed by the practice of other operating systems. MacOS, Windows and Android do not currently impose additional fees beyond commissions for app distribution that would be prohibitive to credible and likely new entry. Professor Wright said that each of these operating systems embodied their intellectual property, but they are not charging in a way that there cannot be entry of alternative app stores. Further, Apple’s pricing on the Mac app store was a relevant factor in considering what Apple would do in terms of pricing in the counterfactual if there was a competitive alternative iOS app store. Apple does not impose an additional charge on alternative app stores or direct downloads, or developers utilising those means, on macOS.

346    Further, and notwithstanding Apple’s protestations to the contrary, in my view the macOS scenario represents a reliable natural experiment in relation to any developer fee for a non-Apple distributed app.

347    Fifth, Apple also does not impose an additional charge on free apps downloaded from the App Store. Although free downloads are a very important part of the services Apple provides, it has never charged a commission, at least over the relevant period for free transactions. Nor has Apple ever charged for app updates, or for in-app purchases of physical goods.

348    And I agree with the class applicants that the likely reason Apple has never imposed a fee for free apps, app updates and in-app purchases of physical goods is that it recognises that there is significant value to Apple from having a broad and varied catalogue of apps available for iOS, even where Apple is not directly monetising those apps. Free apps represent a significant source of value that Apple provides through the App Store and they build up the value of Apple’s service which enables them to be profitable. Apple benefits significantly in economic terms from the indirect network effects, the details of which I have discussed elsewhere. I will come back to the topic of free apps later as Apple sought to deploy it to its advantage for a different purpose.

349    In the circumstances, in the counterfactual, Apple would not have been likely to stop investing in the App Store or impose a materially different fee.

350    Sixth, as the class applicants suggest, one way in which Apple could recoup lost commission revenue in the counterfactual with competition, including even if it were to eliminate commissions almost entirely, would be to increase search ads revenue on the App Store. Indeed there is already significant demand for such search ads, that developers pay in addition to any relevant commission, because in relation to the App Store, there is a significant amount of competition for attention, and one of the ways in which developers compete for attention is to pay for ads.

351    Finally on this aspect, I agree with the class applicants that even if an additional fee had been charged to developers in the counterfactual, it would not likely have been so high as to discourage new entrants into the app distribution or in-app payment solutions markets.

Credible threat of entry into the in-app payment solutions market

352    I agree with the class applicants that in the absence of the contravening conduct during the relevant period, there would have been the credible threat and likely entry into the in-app payment solutions market, as indicated by the following.

353    First, as the class applicants say, in the absence of the contravening conduct there would likely have been substantial competition in the iOS in-app payment solutions market. In the counterfactual there would have likely been competing payment solutions providers in the iOS in-app payment solutions market, attracting market share away from Apple’s payment solution by offering alternatives to app developers. In turn, app developers would have taken up the lower prices offered by the competing providers, and Apple would have faced competitive pressure to reduce its commission rate or lose market share. The increased degree of competition would have led to a reduction in average effective commissions on in-app purchases including subscriptions.

354    Second, as I have discussed elsewhere, there is evidence of app developers choosing alternative payment solutions on other app stores where such choice is provided. Despite Apple’s claims of a trivial number of developers using alternate payment providers on the Epic Games Store, game developers are generally large developers with large volumes of transactions and, therefore, revenue, as per the evidence in the Google proceeding. Other app developers have also requested to use alternatives to Apple’s IAP.

355    Third, I agree with the class applicants that existing payment solutions providers would have been a credible threat of entry in the counterfactual.

356    I have referred to Mr Owens’ evidence elsewhere who said that the size of the potential customer base would have been meaningful from a revenue standpoint for any in-app payment solutions business, indicating that there would have been significant incentives for existing solutions providers to enter the market. So, Mr Owens said that if Apple had allowed alternative payment solutions providers to facilitate in-app purchases, Paddle would have significantly increased its business, and provided developers with an alternative solution for a lower price and with better features. If Apple permitted alternative payment solutions, these standalone payment providers would be able to enter the market rapidly, including because some of them already facilitate the sale of physical goods on iOS native apps or offer payment solutions on web-apps or PC native apps. Mr Owens said that Paddle was able to develop an effective payment solution for iOS native apps within a matter of weeks and, if permitted by Apple, would be able to make that product available to app developers in Australia in a matter of weeks.

357    I have also made the following points elsewhere.

358    Epic also offers its own payment processing solution, being EDP, in the Epic Games Store while also giving developers the choice to use an alternative payment solutions provider.

359    If Epic could distribute the Epic Games Store, it would offer all developers who distribute through the Epic Games Store the option to choose their own payment system for in-app purchases. Were Epic able to launch a mobile version of the Epic Games Store on iOS, then Epic would maintain the same payment solution policy that it currently has for the Epic Games Store and EDP on PCs. Moreover, Mr Sweeney even said that he and others at Epic have considered the possibility of launching EDP as a standalone solution that facilitates in-app purchases in apps downloaded outside the Epic Games Store. Although Epic has not decided to do so, it is nonetheless a credible potential entrant into the market in the counterfactual.

360    Fourth, in the absence of Apple’s conduct, developers could also rely on their own payment solutions including those that suit their individual needs.

361    Fifth, as I have explained elsewhere, given the wide range of payment solutions that are being used by merchants selling non-digital products within apps, one would expect merchants selling in-app digital content on iPhones and iPads to similarly use a range of payment solutions if permitted to do so. Further, the payment solutions used for digital content do not differ from those used for physical goods. Moreover, there are a number of examples where merchants selling physical goods in-app on iOS have selected various payment solutions providers. IAP is not available to these merchants in such a context. Such merchants use different solutions that are optimised for their own individual businesses. They deliver the individual user experience that they want for their customers.

362    Apple’s contravening conduct prevents developers of iOS apps from constructing tailored solutions in the factual, but it is likely that they would have done so in the counterfactual, generating demand for alternative in-app payment solutions.

363    Sixth, I agree with the class applicants that the integration benefits of Apple’s IAP propounded by Apple are not such as to negative the credible threat and likely use of alternative in-app payment solutions in the counterfactual.

364    As the evidence which I have discussed elsewhere reveals, third party payment solution providers can and do assume responsibility for complying with tax obligations, ensuring compliance with payment card industry security standards. They offer a system for controlling refunds. They support customers’ purchases and in-app managing of subscriptions and refunds. They offer customer support. They provide analytical services. They protect against accidental purchases and they provide fraud protection, privacy and security features.

365    In summary, in my view it is likely that price competition would have occurred in the counterfactual. And in the absence of a competitive response, Apple would likely have lost revenue because it would have been exposed to price competition from alternative app stores, sideloaded apps and alternative in-app payment methods.

Comparator analysis — Apple context

366    Let me now turn to the comparator analysis itself.

367    Now different methods may be used to quantify the impact of any anti-competitive conduct on prices.

368    Mr Holt said that comparator-based approaches compare the price that resulted in the presence of the conduct to a price that resulted in the absence of the conduct. He said that the three broad categories of comparator-based approaches are as follows.

369    First, the time-series approach which compares prices for the affected market during the impact period to prices during a control period.

370    Second, the cross-sectional or yardstick approach which compares prices for the affected market during the impact period to prices in other comparable markets, such as an alternative geographic market. Where no alternative geographic markets are suitable, it may be possible to compare the price of the affected product to a related unaffected product.

371    Third, the difference-in-differences approach, which applies statistical techniques to compare the evolution of the affected market to a control market, requiring data for the affected market prior to the conduct occurring.

372    Mr Holt explained that the technique used to carry out the comparison depends upon the data available.

373    Now in addition to the comparator-based approaches, Mr Holt also considered the cost-plus method, but determined that it would be less reliable than using a comparator method in these proceedings. He explained that it was not an approach that he would necessarily support, as it requires many assumptions to be made including as to required rates of return and the basis upon which to identify the plus element of the cost-plus method.

374    Due to data availability and the characteristics of the markets considered, in his view the yardstick comparator method was the best approach for assessing the counterfactual commissions.

375    Further, Mr Holt said that it was not possible to use other methods to determine what commission rate would be applied in a competitive counterfactual because Apple has engaged in the contravening conduct globally since the inception of the App Store.

376    It was not possible to refer to the commission rates that were applied to developers in a competitive period prior to the relevant period since Apple’s conduct has subsisted since the inception of the App Store.

377    I might also note that Mr Holt’s evidence was that it was also not possible to use other methods to determine what commission rate would be applied in a competitive counterfactual because Google has engaged in some or all of the contravening conduct since the launch of the Play Store. Let me return to Apple and the App Store.

378    In Mr Holt’s view there was no time period and no overseas market which could be observed absent the contravening conduct. In his view this left the yardstick approach as the only option to use empirically in this case.

379    Mr Holt concluded that there was no unaffected period or unaffected geographic segment which could be observed, leaving the yardstick approach the only empirical method available.

380    Professor Asker agreed with the descriptions of the three potential methodologies identified by Mr Holt. Moreover, in the concurrent evidence session, Professor Asker seemed to agree that the yardstick analysis with the PC app distribution market was informative.

381    In undertaking his yardstick comparison, Mr Holt examined the commission rates charged by online platforms that operate in two sided markets. In selecting relevant comparators, Mr Holt identified the following criteria as salient.

382    First, industry and business model similarity, where comparators offer similar services in a similar industry, and using a similar business model, their costs and the demand they face are more likely to approximate that of the App Store.

383    Second, market competitiveness, where markets that are competitive are more likely to exhibit prices that are a benchmark for the prices that would have prevailed absent the contravening conduct.

384    Third, pricing structure, where comparators like the App Store which have a fixed fee and a per-transaction commission structure are more comparable as a benchmark for the commission component of that pricing structure.

385    Let me turn to Apple’s arguments concerning the correct methodology.

Apple’s arguments

386    Apple says that Mr Holt did not properly apply the yardstick approach. It has made the following points.

387    Apple says that the evidence is not capable of establishing any overcharge, let alone its quantum. The lack of rigour in the analysis is evident from, among other things, the fact that Mr Holt concluded that the competitive commission range and rate remained unchanged over a four-year period. But a yardstick analysis does not let one pick their preferred period and ignore everything else that has occurred in other parts of the relevant period.

388    In any event, Apple says that Mr Holt’s evidence was not directed to the relevant question of what Apple would have charged in the counterfactual. Instead, he addressed what a notional market participant would charge. Further, he made no allowance for the consideration of the quality of different platforms or for the payment of any fee to Apple. Further, Apple says that there is no evidentiary basis for saying that there is a real chance that Apple would not charge a fee at all or only a modest one.

389    Apple says that even if Mr Holt’s evidence were accepted, it would mean that there has been no overcharge for at least the vast majority of the class who are developers. Mr Holt’s most likely counterfactual rate is 15%.

390    Apple said that 29,981 out of some 30,000 developers in Australia including Dark Ice are eligible to pay a rate of 15%. Even for the 19 Australian developers above the small business threshold, there are further opportunities to pay commission of 15% by other means such as through the video partner program and subscription renewals. This proposition seems to have been sourced to Ms Wong’s evidence. But the class applicants make two points.

391    First, her evidence did not address any iOS app developer who supplied their apps on the Australian storefront of the App Store during the relevant period whose accounts were registered with an overseas address and who paid higher rates.

392    Second, the small business program was only introduced in January 2021 and well into the relevant period, which began on 6 November 2017.

393    Moreover, Apple says that Mr Holt has made no allowance for the fact that third party stores, and developers distributing apps on those stores, would be required to pay a fee to Apple in the counterfactual to use its tools and technologies, and access its user base.

394    Apple says that Mr Holt’s only response to his failure to allow for the payment of such a fee is that Apple does not present a concrete counterfactual scenario. That is no answer to a failure to consider a necessary integer of the theory of loss.

395    Apple says that the onus lies on the class applicants to prove to the requisite standard the counterfactual for which they contend and the overcharge said to arise by reason of the identified conduct. The point has added significance because of what it reveals about false assumptions of competitive entry. If there is no reasonable commercial likelihood of entry, it follows that entry on the balance of probabilities has not been proved.

396    Let me elaborate on some of the points made by Apple.

The quality of the comparators

397    Now Mr Holt said in his report that the yardstick approach compares the price for the affected market to prices in other markets that are deemed to be reasonably comparable to the market at issue.

398    During oral evidence, Mr Holt described this approach as one which looks at the price of an alternative product similar in nature to the affected product but unaffected by the anti-trust concern, and uses it as a benchmark to construct what may have happened in the event of competition.

399    But Apple says that Mr Holt did not identify comparators which were in fact reasonably comparable or similar in nature to the App Store. And Apple says that Mr Holt’s purported use of the yardstick approach involved cherry-picking four comparators in relation to PC app distribution.

400    Now Mr Holt accepted that the quality of an app store can be affected by a range of factors and that differences in commission rates charged by different app stores may reflect differences in their quality compared to other app stores.

401    So, Mr Holt recognised that developers may want curated app stores that are attractive to different types of consumers that could be relevant to which app store they distribute through. He said that innovation could lead to increases in quality. But in selecting the four PC app stores that he used as comparators, Mr Holt did not form any view as to whether the quality of those app stores is greater or lesser. He also said that he had not formed a specific view as to the relationship between quality and commission rates.

402    Apple says that Mr Holt did not consider quality issues identified in respect of the Epic Games Store, and he accepted that those matters could well be an explanation for the lower commission rate.

403    As to the Epic Games Store, insofar as it involves the sale of third-party apps it is loss making. Further, at least in 2022 it had only about 0.1% of the number of apps compared to the App Store. Further, almost all the apps available from the Epic Games Store are game apps. Further, its curation of content is very limited. Further, it permits other app stores to be downloaded through it. Further, it takes no responsibility for the security of, or content available through, those other app stores.

404    Apple says that an app store that has been loss making for 6 years and is anticipated to remain loss making for at least three more years is not a reliable basis for inferring what a counterfactual commission charged for the App Store would have been.

405    Further, Apple asserted that the Microsoft Store is viewed as a low-quality platform that offers a limited number and variety of apps. Further, Apple asserted that it suffers from significant security issues. Further, Apple asserted that the Microsoft PC store suffers from poor search functionality, which makes it difficult for users to find the apps they want. Now I did not have specific direct evidence on such matters supporting such assertions.

406    As to itch.io, there were also quality issues in relation to the platform.

407    Apple says that the attempted comparisons are meaningless because a useful comparison on price is not possible when divorced from quality. Apple says that there was no attempt made by Mr Holt to apply commercial reality to the comparator exercise.

408    Apple says that as Professor Hitt explained, it does not make sense to compare prices between products unless they are meaningfully comparable, and there is research on price indices that supports the idea that products need to be equivalent before comparing their prices.

409    Similarly, Professor Asker said that in any market setting where there is product differentiation, accounting for the difference in quality is important. Mr Holt assumed that price is the primary dimension of competition. But without accounting for quality differences, price comparisons are without merit, and to do so would be to ignore the ceteris paribus principle.

410    Apple says that a proper application of the yardstick approach would have entailed an assessment of the quality of the chosen comparators.

The commission rates charged by Steam

411    Apple says that Mr Holt’s adoption of the lowest rate, being 20%, rather than the highest rate, being 30%, charged by Steam is flawed. It says that that is so for the following reasons.

412    First, it says that Mr Holt did not consider how many of Steam’s developers pay the lower commission rates of 20% and 25%.

413    Second, it says that had Mr Holt undertaken that analysis, it would have revealed that at least 96.3% of developers on Steam pay the 30% commission rate. When presented with that evidence, Mr Holt said that his main focus was “in the round” evidence, whatever that means. He also referred to the possibility that large developers had negotiated favourable rates. That can only refer to the 3.7% of developers paying a commission of less than 30%, and Mr Holt accepted that he had not undertaken any analysis of Steam’s actual effective commission rate.

414    Third, it says that in the circumstances, there is no foundation for Mr Holt’s selection of 20% of the upper bound of his suggested commission range, and hence for the selection of 15% as the most likely commission rate.

415    Further, Apple says that there are flaws with Mr Holt’s selection of 20% as an upper bound.

The omission of console stores and Android stores

416    Apple says that Mr Holt omitted app transaction platforms for Android and console stores when calculating the counterfactual commission rate. Apple says that that omission is unjustifiable because those app stores are relevant comparators.

417    As to the console stores being Nintendo Switch, Xbox or PlayStation, Apple says that Mr Holt did not undertake any analysis of whether console platforms compete with one another because on his instructions as to market definition he considered that they were not relevant comparators. But Apple says that if the console operators compete with one other, then the commission rates they set are, by definition, competitive commission rates, and should have been taken into account.

418    Apple says that console app stores should be included as comparators because of the likelihood that they compete with each other, and also because even if they are outside the market they are likely important out of market constraints. Mr Holt’s omission of console app stores implies the improbable economic outcome that these platforms have been exercising substantial market power for many years.

419    Professor Hitt also considered console stores as relevant comparators for the same reasons as Mr Houston, and also because consumers who used Fortnite on iOS often transacted on those consoles for content that can be used across those platforms. Consumers are able to access consoles for content on at least Fortnite.

420    Mr Holt said that it is unlikely that app stores operated by console operators can provide a measure of what the competitive commission rate would be. But in the part of the report where that opinion is expressed, Apple says that Mr Holt offered nothing to support his opinion.

421    Apple says that in addition to Microsoft Store for Xbox, other comparator app transaction platforms for game consoles, which Mr Holt ignores in his analysis, include the following.

422    Nintendo eShop launched on 6 June 2011 and is currently available for the Nintendo Switch family of devices. The Nintendo eShop mostly offers game app transactions, but developers of certain entertainment apps, such as YouTube, also transact through the Nintendo eShop. The Nintendo eShop charges developers a 30% commission rate.

423    PlayStation Store launched in 2006 for PlayStation 3, which is a home video game console developed by Sony Computer Entertainment. Consumers also have access to Sony PlayStation Plus, a subscription-based streaming service that gives users access to a streaming library of PlayStation 4, PlayStation 5, and classic PlayStation games. The PlayStation Store charges developers a 30% commission rate on game downloads and in-app purchases.

424    In addition, in certain instances, Sony’s cross-platform policies charge developers a percentage of their earnings on other platforms based on PlayStation gameplay. For example, Epic’s agreement with Sony regarding Fortnite revenue sharing stipulates that Sony’s share of Fortnite revenues depends on the [REDACTED] [REDACTED] [REDACTED] [REDACTED] [REDACTED] [REDACTED] [REDACTED]

425    As to the Android app stores including the Play Store, Mr Holt excluded them as comparators from his yardstick analysis on the basis that the Android app store market might not be competitive. But again, Apple says that Mr Holt accepted that if Android is a competitive market, then those platforms should be included as part of his yardstick approach.

426    Apple says that competitors of the App Store include the Play Store and other Android app stores, and their commission rates should have been included as relevant comparators.

427    Apple says that even if the Play Store did not charge a competitive commission rate, that does not mean other platforms for Android apps, such as Amazon Appstore, Samsung Galaxy Store and Huawei AppGallery, do not charge competitive commissions. Accordingly, Mr Holt should have used commission rates charged by those Android app stores as comparators in assessing the alleged overcharge in this proceeding.

428    Mr Holt accepted that Android has multiple app distribution channels. Despite this, Mr Holt said that other Android app stores may not be competitive. The justification was that the Play Store enjoys a dominant position on the Android platform. But Apple says that that does not justify its exclusion. Even with a high market share, it remains a relevant comparator.

429    Moreover, whatever the position of the Play Store, it says nothing at all about other Android platforms. Indeed, if those other platforms have a low market share and are looking to expand, economic theory says that they will offer highly competitive rates to win market share. Yet these other platforms were ignored by Mr Holt.

430    Mr Holt acknowledged that non-Play Store Android stores are viable. But he did not include them as benchmarks in his analysis of the counterfactual commission rate. Moreover, Mr Holt said that he excluded Android platforms from his comparator analysis because Android app stores are highly concentrated.

431    But he did not provide any rationale as to why concentration justifies the exclusion of other non-Play Store Android stores from his analysis, especially considering that Android platforms compete with transaction platforms on other devices.

432    Apple says that consideration of the following platforms suggests that the commission charged by Apple on the App Store is highly competitive. The Amazon Appstore has a [REDACTED] headline rate with [REDACTED] in some circumstances. The Play Store has a 30% headline rate with 15% in some circumstances. The Huawei App has a 30% headline rate with 20% to 50% in some circumstances. And the Samsung Galaxy Store has a 30% commission.

433    As to the Mac app store, Mr Holt sought to justify its exclusion based on a single internal email within Apple in 2011, being 6 years prior to the start of the relevant period, which stated that Apple takes account of the consistency of the commission across the different App Store platforms that it runs. But Apple says the email does not justify the exclusion of the Mac app store as a comparator. Apple says that the Mac app store is a PC store that should have been included as a comparator in assessing the counterfactual in this proceeding.

434    Further, Mr Holt referred to a range of other platforms which appear to be far less comparable to the App Store than any of the platforms he excluded. Mr Holt referred to platforms relating to taxi and ride sharing services, being Uber, accommodation, being Airbnb, and dating services, being Match Group. But Apple says that the reality is that any comparison with these platforms is problematic.

435    As to Uber, which charges 25% commission, the drivers are ostensibly equivalent to the developers, and the passengers equivalent to end users. But Uber provides wholly different services to the App Store, including, for example, car insurance which appears to be provided for the mutual benefit of both drivers and users. Mr Holt’s other platforms such as Airbnb, eBay, and Match Group have analogous differences.

Comparator payment processors

436    Apple made the following points concerning Mr Holt’s analysis.

437    Mr Holt identified a range of pure payment processors who charge a fee of between 1.75% and 5% plus a fixed fee per transaction, with the sole exception of Square, which does not charge a fixed fee. After Mr Holt attempted to take into account the fixed fee, by looking at an average transaction sized fee of AUD 12.03, the commission rate was calculated to be 2.20% to 9.16%.

438    Stripe, Square, Braintree, Adyen and Paddle offer services that allow them to provide equivalent functionality to that offered by Apple’s IAP, including payment gateway features, storage of payment information, security features, data protection, fraud detection, digital inventory management and one-off payments and transactions.

439    The fees for these payment solutions providers range from 1.75% to 5% plus a fixed fee of up to AUD 0.50 per transaction.

440    Apple says that the evidence does not establish the nature and extent of the services offered by the payment operators identified by Mr Holt with the possible exception of Paddle which, according to Mr Owens, provides some services beyond mere payment processing but nowhere near the array of other services offered by Apple's IAP functionality.

441    Mr Holt said that in the counterfactual, the portion of Apple’s commission attributable to its payments solution would be in the range of 2.2% to 9.16%, and that “the central scenario” is likely to be closer to 4% or 5%.

442    But Apple says that there is nothing in Mr Holt’s report capable of justifying those conclusions. In any event, Mr Holt did not offer an opinion on whether the charges from the payment processors were higher or lower than the equivalent charges on the App Store. That would first require an opinion as to the costs charged by Apple.

443    Moreover, Apple says that there is no evidence that allows a comparison of the quality of the payment processing offered by Apple and by third parties.

Analysis

444    Generally speaking I have not accepted Apple’s arguments.

445    Now it is necessary to rely on comparators in other markets that have sufficiently similar characteristics to the iOS app distribution market and the iOS in-app payment solutions market.

446    In selecting relevant comparators, Mr Holt identified various criteria as salient.

447    In Mr Holt’s view, app distribution platforms on PCs are the most appropriate comparator for the purposes of a counterfactual commission benchmarking analysis. Mr Holt gave evidence to the following effect as the class applicants point out, and which evidence I accept.

448    First, the services offered by PC app distribution platforms are similar to those provided by the App Store, and they usually offer users a large selection of PC apps which can be directly purchased before the user can download the app from the platform.

449    Second, there are multiple distribution platforms for PC applications, including Microsoft Store, the Mac app store, Steam, and the Epic Games Store. There are, or have been during the relevant period, additional smaller competitors. There has also been market entry by the Epic Games Store.

450    Third, as with the iOS app distribution market, in the absence of Apple’s contravening conduct, and the Android mobile app distribution market in the absence of Google contravening conduct, developers of PC apps multi-home, and many developers release their apps via multiple app stores. Mr Holt said that PC app stores do not face artificial barriers to entry on PC operating systems. That is, unlike in the iOS app distribution market and the Android mobile app distribution market, app distribution is not restricted.

451    Fourth, like the counterfactual, PC users can download apps directly from a website, rather than only having to access a PC app from a PC app store. This adds an additional competitive constraint on those platforms.

452    Fifth, despite likely having a similar cost structure as a stand-alone App Store and Play Store, PC app stores exhibit much lower profitability compared to the App Store and the Play Store.

453    Sixth, PC app stores are the app platform with the most proximate competition. In relation to the App Store, Mr Holt said that PC users who wish to acquire an app from an alternative source can do so using an alternative app store on the same device and do not need to switch to a different device, in contrast to iOS and game consoles. In relation to the Play Store, Mr Holt contrasted the frictions that apply to sideloading or direct distribution on PCs and Android smart mobile devices, and said that removing Google’s contravening conduct would reduce the frictions of sideloading on Android to a level that is likely to be more similar to the cost of direct distribution on PCs.

454    Seventh, an important factor to bear in mind is that although the PC market is an appropriate comparator, it is not a market in which there exists or has existed an incumbent firm enjoying the supra-competitive profits of Apple.

455    In Mr Holt’s opinion, the PC app stores market when considered with the relevant evidence supports the conclusion that the counterfactual commissions for app distribution and in-app payments would have been materially lower in the counterfactual. This is for the following broad reasons, as the class applicants correctly contend.

456    First, the PC app market shows rivalry between app stores, such as Microsoft, Epic and Steam, and provides examples of those rivals reacting contemporaneously to entry into the market by reducing prices from their headline rates.

457    Second, the PC app market demonstrates that headline rates can differ significantly from the rates agreed with individual developers in a competitive market. On this point Mr Holt referred to evidence that Steam allows discounts to developers against its published rates and allows developers to generate codes or keys for sale outside the Steam storefront, which can then be used to download games for Steam with no commission paid. Mr Holt also referred to the evidence of Mr Sweeney that Epic negotiated a commission rate of between 5% and 15%. Further, side negotiated deals had occurred between Steam and developers which resulted in the commission being less than the headline rate, and Professor Asker acknowledged that Steam allows developers to generate codes or keys for sale outside Steam, for which Steam can charge no commission, as a feature of how it helps developers promote their game. I will say something more about Steam in a moment.

458    Third, the PC app distribution market demonstrates the importance of direct distribution to competition in the counterfactual. In that market, several developers launched their own storefront and Microsoft estimated that 83% of PC app revenues were by direct download. Mr Holt’s evidence that his range of 10% to 20% for the counterfactual commissions is conservative in light of the low costs of direct distribution to developers and the significant amount of direct downloading that would have likely occurred in the counterfactual.

459    Mr Holt considered the market position and pricing structure of PC app stores.

460    In respect of Steam, prior to and during the relevant period, Steam was the largest platform for digital distribution of PC games outside China. In 2020, it had 62.6 million daily active users. Steam charges a percentage fee for app distribution and also a USD 100 fee to each new developer who submits an application distribute on Steam.

461    Until December 2018, Steam charged all developers a 30% fee on app distribution. But five days before the launch of the Epic Games Store, and following the movement of several large developers to other PC app distribution platforms and direct download avenues, Steam reduced its commission rate to 25% for developers’ revenue between USD 10 million and USD 50 million, and to 20% for any revenue over USD 50 million. I will return to Steam in a moment.

462    In respect of the Epic Games Store, it was launched in 2018. In 2020, it had 31.3 million daily active users. Epic’s fee structure is to charge a percentage fee on developers’ gross revenue and also a recoupable submission fee per game submitted to the platform for distribution by a developer. Epic charges a commission of 12%. Where an app developer uses its own payment solution, Epic does not charge a commission.

463    In respect of the Microsoft Store, the Microsoft Store is an app store for Windows PCs. The Microsoft Store was included by default with Windows 10 and 11.

464    Since 2019, the Microsoft Store has charged at most a 15% commission for apps that are not gaming apps. Between 5 March 2019 and 14 January 2020, Microsoft reduced its commission to 5% for sales of qualifying apps on applicable platforms when customer acquisition was directly driven by the publisher’s promotion activities. From 28 July 2021, Microsoft allowed developers who sell through the Microsoft Store and use third-party providers for in-app payments to keep their entire revenue from non-game apps. From August 2021, Microsoft reduced the commission for Windows PC game apps from 30% to 12%, bringing it into line with Epic’s commission on the Epic Games Store. Microsoft’s commission for Windows non-game apps has remained 15%.

465    In respect of Itch.io, which is an open marketplace for independent digital creators with a focus on independent video games, Itch.io has a default 10% commission rate, although users can choose to pay less.

466    In each proceeding, Mr Holt also considered other possibilities for inclusion as potential individual comparator PC app stores. Mr Holt did not include those platforms in his analysis of the commission rates charged by individual PC app stores, on the basis that they accounted for very few sales among developers, discontinued their storefronts, and/or operated as an intermediary for other app stores.

Console stores and Android app stores

467    Now Apple suggests that other relevant comparators are app stores on Android and on games consoles. But this assertion is problematic in substance for the reasons given by the class applicants.

468    First, games consoles are not relevant comparators since each console only allows digital downloads via built-in app stores. They are therefore not a competitive counterfactual rate.

469    Second, Android app stores are also not relevant comparators since there is a high degree of concentration amongst those stores indicating a lack of competitive rivalry.

470    Mr Holt considered the app distribution platforms operated by major console manufacturers being Sony PlayStation, Microsoft Xbox and Nintendo. But in his view, these platforms do not serve as reliable comparators. The owners of each of the console platforms operate the only digital app store on their respective operating system. So, they are an unreliable indicator of what the commission rate would be in a counterfactual in which developers and users have a choice between multiple different app stores and direct distribution.

471    Now in relation to games console and Android app stores, Apple says that Mr Holt disregarded those app platforms because he claimed that they are operating in markets in which they are monopolists or near monopolists. But I agree with the class applicants that this is not a fair representation of Mr Holt’s evidence.

472    Further, as the class applicants put it, the owners of each of the game console platforms each operates the only digital app store on their respective operating system. Due to their exclusivity, console app stores are similarly positioned to the App Store in the actual world in that they are the sole digital distribution channel on their respective devices. As such, console app stores at best provide an indication of what the commission rate might be if there were only one app store on iOS, that is, the actual situation the App Store finds itself in.

473    In relation to Android, Mr Holt explained that he did not consider that Android app stores, including the Play Store, provide a credible benchmark for understanding competition in the counterfactual world or assessing the counterfactual commission in light of the high degree of concentration and the lack of observable competitive interaction between Android app stores. Mr Holt contrasted this with his observations in the PC app stores segment.

474    Let me deal with some other matters before addressing Steam.

475    Mr Holt’s yardstick analysis is not limited to four comparator app stores.

476    Now in cross-examination, it was put to Mr Holt that he used a yardstick approach with four comparators being the Epic Games Store, the Microsoft Store, Steam and itch.io to calculate his counterfactual commission and he has not sought to check it with any different methodology, such as a cost-plus assessment.

477    But that was rejected by Mr Holt. Mr Holt considered the different methods that may be used to assess the counterfactual commissions, and explained why, in this case, due to data availability and the characteristics of the markets considered, the yardstick comparator method is the most reliable.

478    Further, Mr Holt’s analysis of the counterfactual commission was based on the profitability analysis, economic theory, analysis of the market conditions in the iOS app and in-app distribution markets and his analysis of PC app distribution market as a whole, and his finding that the App Store would have continued to be profitable at counterfactual commission rates of between 10% and 20% and with reduced market share, and evidence of Apple, when lowering its commission rate, lowering it to a 15% commission rate for developers in certain circumstances.

479    Further, Mr Holt emphasised that in his yardstick analysis he considered the evidence across the board in relation to the distribution of PC apps, not just the four app stores identified by Apple.

480    Further, I agree with the class applicants that Apple’s criticism of Mr Holt for focusing on comparing four individual PC app stores with the App Store does not satisfactorily address Mr Holt’s evidence where he explained:

Now, it’s important to recognise what I have done. I have not tried to find the best comparator for Apple. I have not tried to identify a single comparator that, in my view, represents the whole market. What I’ve done is identify, in the round, the best evidence that points to what would likely have happened in the counterfactual across the iOS app distribution market and in-app payments solutions markets. So there’s no suggestion that I am relying on any particular app store.

Steam

481    Let me say something about Apple’s criticisms concerning Steam. In doing so, I should say that I have largely accepted the class applicants’ answers to these criticisms.

482    Now Mr Holt was also criticised for his use of Steam’s 20% headline commission rate as a comparator. Mr Holt said that Steam’s commission rates vary between 20% and 30% depending on revenue. Higher revenue developers are charged less. Mr Holt said that the lower rates are a more reliable approximation of the outcome of a competitive market than the higher fees because of two matters. As he said, the larger developers are those whose custom is likely to be the subject of rigorous competition. Further, the larger developers are more likely to have the resources to launch a viable app store as evidenced by the launch of stores operated by Ubisoft, which had chosen to withhold apps from purchase on Steam due to Steam’s business model, Electronic Arts, Epic, and Rockstar Games.

483    Now it was said that Mr Holt was speculating on why Steam chose its pricing structure, which in effect provides for non-linear pricing or a volume discount, and that if seeking to apply a competitive benchmark, Mr Holt ought to have considered the higher commissions. But I agree with the class applicants that Mr Holt’s evidence on Steam went beyond supposition or speculation. He took into account Steam’s headline commission pricing structure. Mr Holt said that the more plausible explanation of the 20% commission rate that Steam offered large developers is that Steam considered it necessary to compete on price with alternative PC app stores rather than a desire to encourage demand on the Steam storefront itself.

484    Further, Mr Holt identified the importance of the distinction between Steam’s headline rates on the one hand and net rates after discounts negotiated with developers on the other. He said that Steam allows developers to generate codes or keys for sale outside the Steam storefront, which can then be used to download games for Steam with no commission paid. Further, there are side-negotiated deals between Steam and developers, although the evidence was vague on this topic.

485    In any event, as Mr Holt rightly observed during cross-examination:

… it doesn’t necessarily matter the number of them. It matters the overall revenue to which the discounts apply. I would expect in general that it’s the largest developers, the ones who account for the most significant proportion of revenue, who will be most attractive for the individual platforms to negotiate with because it’s those very develop that they really want to attract to their platforms.

486    Now Apple says that at least 96.3% of developers on Steam pay a 30% commission rate.

487    So, Apple says that the commission range on Steam is 20% to 30%, and 96.3% of developers pay 30%. It says that Steam charges 30% on the first USD 10 million to 100% of developers. It is only when that threshold is exceeded that the lower commission applies. Apple says that this makes the selection of 20% as an upper bound even more difficult to justify. Mr Darke SC also referred to the 96.3% figure during Apple’s closing submissions.

488    But I agree with the class applicants that this assertion is not established on the evidence and fails in any event to grapple with discounts given by Steam on its headline rates.

489    In evidence was a screen-print of an article published by “Video Game Insights” dated 28 February 2022 and titled “There are 44,000 game developers on Steam. Who are they?”. The article states that “[a]t the time of writing this, there are 44,000 developers who have released or announced a game on Steam” and “[t]his article explores who these developers are, what type of games they make and how much revenue they generate.”

490    The article includes a graph that purports to summarise the number of developers on Steam that have earned lifetime gross revenue within certain bands. Using the numbers presented in the graph, 3.7% of developers have earned lifetime gross revenue of “$1m+”. As the class applicants say, it would seem that the 3.7% figure is the premise for Apple’s assertion that at least 96.3% of developers on Steam pay a 30% commission rate.

491    The document is not a business record of Steam. And the document cannot be used to establish the truth of the 96.3% figure.

492    Further and in any event, the article says nothing about the commission rates paid by developers to Steam. It merely purports to summarise the gross lifetime revenue earned by developers on Steam. In fact, the article does not mention the word “commission” or any equivalent term.

493    So, the article provides no admissible evidence about what headline commission rates developers in fact pay to Steam, including the impact of any discounted commission rates agreed between those developers and Steam, and the keys that Steam allows developers to generate for sale outside the Steam storefront, which can then be used to download games for Steam with no commission paid.

494    Accordingly, I agree with the class applicants that the document cannot support Apple’s claim that at least 96.3% of developers on Steam pay a 30% commission rate.

495    Further, as the class applicants point out, Apple’s submission that Steam charges 30% on the first USD 10 million to 100% of developers similarly fails to grapple with the prospect that Steam has agreed any discounted commission rates with any developers for revenues under USD 10 million and excludes the impact of Steam keys. There are side-negotiated deals between Steam and developers.

496    In any event, Apple only focuses on the number of developers that might be eligible to pay Steam a commission rate of 30%. As Mr Holt made clear, it is the revenue earned and commissions paid by developers that is the relevant integer for the purpose of assessing Steam’s effective commission rate.

The quality question

497    Now Apple said that the class applicants’ evidence in respect of the overcharge issue did not account for quality or other differences between relevant platforms. Google made a similar submission.

498    Now as part of a criticism of the comparators which have been taken into account by Mr Holt, it was suggested in cross-examination that Mr Holt had not analysed, or made any assumptions about, the quality of apps in the PC market relative to the App Store.

499    But Mr Holt’s analysis did account for differences in product quality. As the class applicants point out, His evidence was that with rival entry into the iOS app distribution market, a range of options could materialise. Some platforms may be high quality and some may be low quality. Moreover, he expected that platforms may wish to create differentiated products, some of which might be niche entrants.

500    During the concurrent evidence session involving the relevant economic experts, Mr Holt said that:

… [I]n a competitive environment, there can be a number of issues that might arise. There could be increases in quality, there could be new innovations, there could be increased choice, there could be, you know, many sort of developments that could happen. I’ve essentially abstracted from the specific detail as to what business models and what level of quality individual entrants would take, and formed an overall view as to what likely would happen in the market.

501    In response to questions from me, Mr Holt further said that:

… [W]hat I’ve measured is the effective average that would apply across the market. That could include some app stores charging higher amounts, some charging lower; that allows for differentiation on quality. It could allow for some app categories to have some on higher rates, some to have lower. There’s nothing in my analysis that precludes any degree individualised outcomes of that nature.

502    Under cross-examination from Mr Darke SC, Mr Holt addressed the comparative quality features of the Microsoft Store. Mr Holt also explained that product differentiation, in terms of quality, would be a feature in the counterfactual.

503    As the class applicants point out, in cross-examination, Mr Holt was asked to assume that the Microsoft Store, the Epic Games Store and itch.io are low quality PC app stores. It was put to Mr Holt that low commission rates set by low quality app stores are unlikely to trigger any material response from other app stores, including the App Store, and are unlikely to provide a useful guide to the realistic counterfactual commission rate on iOS.

504    Mr Holt pointed out that, whilst he has not formed a view on quality of specific stores, this was not necessary in his analysis, which was aimed at understanding the likely implications of changing the competitive conditions, including entry.

505    Further, as the class applicants point out, Mr Holt said that even on the assumptions put to him, Apple’s proposition was disproved by the facts because, under conditions of entry and competition, there has been significant downward reductions in headline rates amongst PC app stores. So, Steam’s price reaction was to the announcement by Epic of the imminent entry of the Epic Games Store. That is, it was the fact that there was more competition that led Steam to reduce its commission rates, rather than anything specific to the level of quality that was being assessed.

Comparator analysis in relation to in-app payment solutions

506    Now Mr Holt’s opinion was that the component of his counterfactual commission for app distribution which concerns in-app payment is between 2.2% and 10% of the purchase and likely to be around 4% or 5%.

507    And as the class applicants point out, Mr Holt identified five comparators being Stripe, Square, Braintree, Adyen and Paddle.

508    Now each of them provides payment solutions to merchants in a competitive environment. And each offers services that allow them to provide equivalent functionality to that offered by Apple’s IAP, including payment gateway features, storage of payment information, security features, data protection, fraud detection, digital inventory management and one-off payments and transactions.

509    The fees for these payment solutions providers range from 1.75% to 5% plus a fixed fee of up to AUD 0.50 per transaction.

510    Mr Holt compared the pricing offered by these comparators by considering the commission rate developers would have paid for the average transaction in Australia during the relevant period on the App Store, which he identified was AUD 12.03, and during the period between November 2017 and December 2021 on the Play Store, which he identified was AUD 10.82.

511    In reaching his opinion, Mr Holt also had regard to the following.

512    First, the commission rate in fact set by Apple in South Korea and the Netherlands, where legislative and regulatory requirements require Apple to untie its payment services from its distribution services. This rate was between 3% and 4%.

513    Second, the reduction of 4 percentage points to the commission rate that Google offers developers when they use their own billing option under Google’s user choice billing program.

514    Now Professor Asker originally said that because Google provides a 3 to 4% discount on its service fee for Google Play Billing, this is Google’s implied fee for its billing system. And he originally said that this is comparable to the fees charged by rival providers of payment solutions and is therefore competitive. But ultimately he accepted that the global blended cost across all forms of payment, according to the documents, seems to be about 6%.

515    In those circumstances, Mr Holt concluded, correctly in my view, that Google’s implied fee is not competitive. Indeed, as I have indicated elsewhere, it would be uneconomic for a developer to use a payment solution other than Google Play Billing in jurisdictions where Google offers the 3 to 4% discount since their overall costs would rise to 31 to 32% from the Google Play Billing fee of 30%.

Conclusion

516    Generally speaking, and notwithstanding its imprecision, the comparator analysis performed by Mr Holt does tend to have some merit.

Counterfactual commission(s) – Apple context

517    Now on the class applicants’ case, the difference between the effective commission rates paid by Dark Ice and iOS app developer group members during the relevant period and the counterfactual commission represents the extent of the overcharge obtained by Apple by reason of the contravening conduct, being commissions paid by Dark Ice and iOS app developer group members, which would not otherwise have been paid by them.

518    Now as the class applicants have explained, Mr Holt’s evidence was that having regard to the relevant comparators, a competitive level of commission in a competitive iOS app distribution market would be in the range of 10 to 20%, with a midpoint of 15%.

519    He said that PC app distribution platforms provide the closest comparators for the purpose of assessing what a competitive outcome would be in the absence of the contravening conduct.

520    On his analysis, Steam, the largest digital platform for distributing games on PCs, charges a 20% commission for the largest developers. Epic, which has entered recently and is competing to grow its market share as an app distribution platform, charges a 12% commission rate. The Microsoft Store charges 15% for non-games, and 12% for Windows games. Finally, itch.io charges a default commission rate 10% and continues to be a viable app distribution platform.

521    He said that the counterfactual commission of 10 to 20%, with a midpoint of 15%, includes a component attributable to the use of an in-app payment solutions in the range of 2.2 to 10% of the relevant purchase.

522    Now he said that the counterfactual commission is expressed as a range, which allows a degree of product differentiation within the market. Further, although 15% is expressed as a midpoint in this range, Mr Holt considered that, based on a comparator analysis, it is also the most likely commission rate that iOS app developers would have paid but for the contravening conduct.

523    Now as to the other integer for any calculation, Mr Holt relied on Professor Wright’s calculations of the effective or average commission rate.

524    Professor Wright calculated that, during the relevant period, Apple’s effective commission rate based on transaction data for Australia was 29.1% in 2017, 28.8% in 2018, 28.3% in 2019, 28.1% in 2020, 26.8% in 2021, and 26.6% in 2022. In cross-examination, Professor Hitt confirmed that he did not dispute that Professor Wright calculated Apple’s effective commission rate correctly.

525    The “effective commission rate” as calculated by Professor Wright was the weighted average commissions paid by app developers for paid downloads and in-app purchases, with the weights being the gross revenue from digital content and paid downloads made at that rate. It incorporates only paid downloads and in-app purchases, but not free downloads nor free in-app purchases.

526    Now I tend to agree with Apple that the effective commission rate for each year in the relevant period is not useful to compare against the counterfactual. No developer currently pays an effective average commission. Rather, they pay many different commission rates on different transactions, and an assessment of the individual commissions could well lead to the conclusion that many of them are competitive relative to benchmarks.

527    Further, Apple says that free transactions comprise the majority of transactions on the App Store and it has criticised both Professor Wright and Mr Holt on the basis that they failed to account for these transactions when calculating Apple’s effective commission rate. But Mr Nicholas De Young KC for the class applicants identified the error in Apple’s position. I agree with him. It is unnecessary to descend into the detail of this point.

528    Now Mr Holt said that he has not assumed any fee, such as for access to Apple’s intellectual property-protected tools, technologies and services, and user base, because that would be an unrealistic counterfactual. But Apple says that there is no realistic prospect of Apple simply granting access to its intellectual property-protected tools, technologies and services, and user base for free.

529    Accordingly, Apple says that even if the rates charged by Apple could be equated with Mr Holt’s notional market rate, it would be necessary to add an access charge to Mr Holt’s notional counterfactual commission range of 10 to 20% and most likely rate of 15%.

530    Now of course the class applicants dispute that any significant additional charge would have been made, and I am inclined to agree with them.

531    Let me deal with a more general topic. I agree with the class applicants that Mr Holt did not arbitrarily select the counterfactual commission. He arrived at his counterfactual commission range having regard to all of the evidence.

532    The lower end of Mr Holt’s range, being 10%, allows an app store to cover the expected cost of providing payment services using the most expensive option available, whilst the upper end of his range, being 20%, reflects the response of Steam to increased competitive pressure on its commission rates for its most valuable customers.

533    Further, on one view Mr Holt’s range of 10% to 20% may be conservative in light of the low costs of direct distribution to developers and the significant amount of direct downloading that would have likely occurred in the counterfactual.

534    Further, Mr Holt’s opinion that 15% is the most likely commission rate was expressed having regard to all of the relevant material available. He noted the following considerations.

535    First, in the PC app distribution market, Microsoft Store charges a 12% commission for Windows game apps and 15% for non-game Windows apps while the Epic Games Store charges a 12% commission rate.

536    Second, in the factual, there is evidence that when exhibiting a price response, each of Apple and Google have dropped their commission rates to 15%.

537    Third, Google’s internal documents reveal that there was “no rationale” for the 30% commission rate while a 20% revenue share would bring “Play rev share in-line with upper end of desktop gaming stores”. Similarly, Apple’s internal documents demonstrate that the 30% commission rate was set without regard to competitive constraints.

538    Fourth, the effective commission rate for many app stores is less than the headline rate. For example, the Galaxy Store has a headline rate of 30% but there is evidence that it has negotiated significantly lower rates with individual developers, including Epic, which it offers rates of between 5 and 15%. Further, there is evidence that Steam and the [REDACTED] offer discounts to developers.

539    Finally, Apple made the point that in terms of counterfactual commissions, there may be a distinction to be drawn between game apps and other types of apps. It says that if there was anticompetitive conduct in the markets alleged, it may not have had uniform consequences across all app types in terms of overcharge.

540    And it said that there are some distinct conditions and methods of monetisation, which in turn affect the effective commission rates that operate in respect of game apps and in different transaction platforms.

541    I tend to agree with Apple that the class applicants have not provided any sound basis for me to conclude that if there was some form of contravening conduct and even overcharge in respect of some types of apps, that this can be extrapolated to a finding that there was an overcharge that applied equally across all apps.

542    Further, the commissions payable by developers may vary depending on their circumstances and may vary and have varied over time. App developers may pay no commission on transactions associated with some of their apps. They may pay a rate of 15% on certain transactions which take place through the App Store in respect of other iOS apps which they develop.

Conclusion

543    I do not propose to deal further in these reasons with the counterfactual commission and overcharge questions at this time. I will need to hear further from counsel in the light of the specific contraventions that I have found against Apple.

General – Google context

544    In this section of my reasons I will address the class applicants’ case concerning Google on the question of overcharge including the question of counterfactual commissions.

545    Now it is convenient to address four topics in the following sequence which mirrors the sequence of the discussion in the corresponding Apple sections set out earlier.

546    First, I will address the question of the Play Store profitability, which of course I have addressed in more detail in my reasons in Epic v Google. This topic is relevant to and is an input into the second to fourth topics.

547    Second, I will address the counterfactual scenarios and in this context principally address the appropriate retrospective hypothetical scenarios as they are relevant in the class action context. The future with and without test relevant to whether there have been any statutory contraventions by Google has already been addressed in my reasons in Epic v Google.

548    Third, I will then address Mr Holt’s comparator analysis. Now this topic could have been addressed before the second topic, but I have found it convenient to proceed in the sequence that I have outlined.

549    Fourth, I will finally address the output of this analysis which relates to the counterfactual commissions.

Play Store profitability – Google context

550    Now as I have said in my Epic v Google reasons, Google Play as a business division and correspondingly the Play Store exhibited high profitability over the relevant period in an accounting sense. Let me begin with a summary of Mr Holt’s evidence, Professor Easton’s criticisms and Mr Holt’s responses. The class applicants’ summary for the most part accurately sets this out.

551    Mr Holt agreed with Professor Davila’s conclusion that the Play Store achieved highly attractive profitability during the relevant period. I have addressed Professor Davila’s evidence elsewhere, but it is convenient to summarise it at this point.

552    Professor Davila addressed the profitability of Google Play as a proxy for the Play Store by reference to its gross profit margin and operating profit margin, concluding that, from the period from 2012 to 2022, Google Play achieved highly attractive profitability, with a return on gross profit (ROGP) in the range of 69.1% to 85.6%, after indirect value from other parts of Google.

553    Professor Davila also concluded that since 2016, Google Play’s return on assets exceeded 100%, and its ROCE between FY17 and FY21 was infinite, that is, that Google does not need to invest in the Play Store to achieve returns.

554    Professor Davila’s calculations were achieved by allocating a portion of Google’s balance sheet to the Play Store in the absence of Play Store’s balance sheet being available.

555    Now Mr Holt applied a sensitivity analysis to Professor Davila’s calculations, assuming that revenues were 30% lower, and the cost of sales and operating expenses were 30% higher, than those in the documents relied upon by Professor Davila.

556    Mr Holt performed a sensitivity analysis from which he concluded that, even if he allowed for a conservative assumption that Google’s revenues were in fact 30% lower than the figures relied on by Professor Davila, the Play Store would remain highly profitable.

557    Mr Holt agreed with Professor Davila’s approach to calculating the Play Store’s costs and profits. He concluded that Google Play and correspondingly the Play Store exhibited high profitability over the relevant period, and the Play Store’s returns were significantly and persistently higher than the cost of capital it deployed in earning those returns, being matters consistent with the proposition that the Play Store’s market power allowed it to earn supra-competitive returns which, absent the contravening conduct, would have attracted market entry which would, in turn, have placed downward pressure on commission rates paid by Android app developers and resulted in a reduction in the Play Store’s profitability.

558    Now as I have already indicated, one recognised metric for assessing profitability involves considering whether a company’s ROCE is significantly and persistently higher than its WACC. Such a finding is suggestive, albeit not determinative, of a market lacking effective competition, particularly where those excess returns are not eroded over time.

559    Mr Holt’s profitability analysis was based on the primary analysis undertaken, and conclusions reached, by Professor Davila, who considered Google Play’s and correspondingly the Play Store’s profitability by reference to its ROGP, ROCE and ROA.

560    Mr Holt agreed with Professor Davila’s conclusion that the Play Store exhibited very high profitability over the relevant period according to each of the metrics employed by Professor Davila.

561    Now Mr Holt also undertook a range of sensitivity tests to inform his opinion as to the robustness of Professor Davila’s findings, applying sensitivities which he considered to be extreme rather than realistic. These sensitivity analyses included the following variations or possibilities as the class applicants correctly summarised.

562    First, there was an adjustment of the Play Store’s cash allocation method. On the assumption that the Play Store would need to hold sufficient cash to cover 60 days’ worth of its expenses, Professor Davila calculated the Play Store’s ROA as ranging between 148% and 272% across the relevant period and its ROCE as infinite. Mr Holt calculated the effect on these profitability metrics of the Play Store holding the cash for the full financial year, in effect overstating the income of the Play Store, increasing its assets and reducing its profitability, resulting in the Play Store’s ROA ranging between 70% and 84% across the relevant period and its ROCE ranging between 135% and 288%.

563    Second, there was an adjustment of revenue and cost of sales. Professor Davila considered the Play Store’s profitability by reference to internal Google materials, upon which he constructed and analysed profit and loss statements for Google Play and correspondingly the Play Store. Notwithstanding his opinion that the underlying Google data was very clear and that Professor Davila’s approach was already conservative, Mr Holt adjusted Professor Davila’s calculations to assume revenues 30% lower, and a cost of sales and opex that was 30% higher, than those in the documents on which Professor Davila had relied, reflecting continuing high profitability across the relevant metrics and demonstrating, in his opinion, that the Play Store’s profitability did not depend on those revenues.

564    Third, an increase was made to the allocation of R&D to the Play Store. Whilst Professor Davila assumed a level of opex including R&D for the Play Store that was proportional to its revenue relative to Alphabet, Mr Holt allocated an additional R&D expense, as a proxy for both its use of Alphabet’s intellectual property and other intangible assets. Mr Holt modelled this increase as, alternatively, an expense incurred in the relevant year or amortised over four years. The application of this sensitivity demonstrated high profitability in each case.

565    Fourth, there was the exclusion of various liabilities or revenues and the inclusion of various Android costs.

566    Fifth, Mr Holt also applied additional sensitivities which were designed to either reflect a reduction in Play Store’s revenue or an increase in its asset base, in each case reducing its profitability.

567    Following the application of his sensitivities, Mr Holt concluded that the Play Store continued to demonstrate high profitability, and that Professor Davila’s conclusions on profitability were robust to various changes in the assumptions underpinning his analysis.

568    Now Mr Holt accepted that adjustments to the Play Store’s operating profit or assets should not be looked at in isolation. But Mr Holt’s evidence was that the size of any relevant adjustments would not impact his primary conclusion as to the profitability of the Play Store, even allowing for his sensitivity analyses.

569    Now as the class applicants explain, Professor Easton criticised Mr Holt’s allocation of R&D over the period from 2012 to 2022, which he observed equated to only 3% of total Alphabet R&D expenses and only 26% of total Android expenses, suggesting that an allocation of 25% of Alphabet’s R&D to the Play Store would have produced values of ROA and ROGP consistent with firms identified by Professor Davila as relevant comparators and, by implication, reflect a competitive profitability. But according to Mr Holt the selection of the figure of 25% was for illustrative purposes only.

570    Further, Mr Holt said that in order for the Play Store’s profitability to fall below 10%, which Mr Holt considered to be a normal WACC, the Play Store would need to bear an aggregate reduction in operating profit of approximately USD 45.3 billion over the relevant period or an average annual increase in assets of USD 75.1 billion over the same period, neither of which were considered by Mr Holt to be realistic.

571    Now as the class applicants explain, Professor Easton disputed that 10% is an appropriate benchmark for such analysis. But Mr Holt said that even if the WACC for the Play Store was double that of Alphabet, which he considered to be highly unlikely, having regard to the likely non-diversifiable risk of the Play Store, the resultant WACC of between 10.2% to 15.8% is still far below his estimated ROCE for the Play Store.

572    Further, Professor Easton criticised both Professor Davila and Mr Holt for relying on Alphabet’s accounting records in order to assess Play Store’s profitability on two bases. First, he said that under US generally accepted accounting principles, R&D costs and costs related to other intangibles are not capitalised in external financial reports. Second, he said that under the generally accepted accounting principles, companies are required to record some assets at their historical costs, which might be substantially lower than the market value or replacement cost of the asset. But Professor Easton did not seek to identify the relevant intangible assets nor value them, nor did he seek to identify the relevant assets of the Play Store which he said may be affected by higher replacement costs or which have a higher present-day value.

573    Now whilst Mr Holt did not seek to specifically identify the intangible assets in question, he applied his R&D sensitivity as a proxy for all intangibles assets that were not recorded on the balance sheet.

574    Mr Holt said that it is not clear that there would be any important distinguishing feature of Google’s brand as between Google and its potential rivals, and questioned whether there was an appropriate brand intangible.

575    Further, Mr Holt considered that his R&D allocation accounted for intellectual property, which typically includes brand in any event.

576    Mr Holt’s evidence was that it was not appropriate to include the Play Store’s customer base in the range of intangible assets to be assessed, due to the circularity of including the customer base in circumstances where that customer base has been obtained as a result of the contravening conduct.

577    Further, Mr Holt also explained that Android development costs should not be included in the Play Store asset base.

578    Further, as the class applicants point out, Professor Easton criticised Mr Holt for not including costs associated with the development of the software used to launch the Play Store, but according to Mr Holt the non-inclusion of such costs is consistent with the proposition that the useful economic life of that software is likely to have ended prior to the relevant period.

579    Further, it was Mr Holt’s position that the non-inclusion of these costs is also supported by the fact that Apple, which also relies heavily upon software for the operation of its App Store, assumes a useful economic life of internal-use software of five to seven years in its financial reporting.

580    Further, it was put to Mr Holt that the difference between Alphabet’s market capitalisation and the book value of its asset base reflects the value of Alphabet’s intangible assets, but this was rejected by Mr Holt.

581    He explained that market capitalisation is based on a forward-looking assessment by investors of future revenue and, as such, does not provide any foundation for reaching any conclusion as to the value attributable to any relevant intangible assets.

582    Now Mr Holt also gave evidence that the Play Store would remain highly profitable in various counterfactual scenarios.

583    Mr Holt undertook an analysis of the Play Store’s likely profitability in various counterfactual scenarios, conservatively utilising the Play Store’s revenue for the 2017 financial year, being the year in which the Play Store exhibited the lowest profitability throughout the relevant period, and allowing for the additional allowance of the R&D as described in the sensitivity analysis, on the basis, again conservatively, that it was capitalised in the year in which it was incurred.

584    Mr Holt concluded that the Play Store remained highly profitable across the ROGP, ROA and ROCE metrics in all of his scenarios which considered market shares of between 25% and 100% and commissions of between 10% and 20%.

585    Professor Easton said that Mr Holt’s counterfactual analyses suggest that the Play Store is most profitable in scenarios where the commission rate is at its lowest and that the profitability metrics exhibited little variability at different assumed market shares.

586    Mr Holt accepted that the trends identified by Professor Easton appeared counterintuitive, but he said that this reflected the mathematical nature of the profitability ratios, having regard to the revenue streams earned through the Play Store, independent from the revenue earned by way of commissions. The class applicants have correctly summarised the points Mr Holt made.

587    First, Mr Holt said that although his revenue calculations in response to varying commissions, considered against alternative market shares, were consistent with economic intuition, these revenue movements were reflected differently in the context of his profitability analysis, once the relevant asset base and other revenue streams are factored into the analysis. That is, where a change in the market share affects the numerator and denominator of the profitability ratio in equal proportions, the profitability indicator is largely insensitive to changes in market share. Accordingly, it was his view that there is nothing unusual in the profitability remaining stable in the face of declining market shares.

588    Second, Mr Holt said that the increase in profitability in the face of reduced commissions, at constant market shares, is a function of the Play Store’s revenue streams. Where the reduction in commissions is limited to apps and games and does not affect Play Store’s other revenue streams, this has the effect of re-weighting the other revenue streams.

589    Mr Holt’s position was that where those other revenue streams are highly profitable in their own right, the increased net weight of those revenue streams can actually maintain or increase the overall profitability of the Play Store, even in the face of reducing commissions.

590    Now as the class applicants have said, Professor Easton made several challenges to Mr Holt’s assumptions. He challenged the assumption that indirect revenue streams from other parts of Alphabet would increase in the counterfactual, where the Play Store decreases its commission rate. And he challenged the assumption that most of the Play Store’s costs are variable.

591    Mr Holt accepted the criticism of the counterfactual model displaying an increase in the Play Store’s indirect revenue whilst commission rates decreased. As I have said, Mr Holt explained the reason for this occurring and adjusted his calculations based on the indirect revenue calculated by Professor Davila. But according to Mr Holt the revised model continued to reflect very high profitability across all metrics.

592    As to the variability of the Play Store’s costs, Professor Easton said that as a reduction in commissions reflects a reduction in revenue, rather than a reduction in transactions, the Play Store’s costs would not be expected to reduce, on the basis that its costs were reflective of the volume of transactions, rather than by reference to its net revenue.

593    Similarly, Professor Easton said that Mr Holt’s model incorrectly assumed that the Play Store’s allocation of Alphabet’s assets reduces as a result of reduced revenue, even though those assets remain owned by Alphabet, meaning that Mr Holt simply re-allocated the assets without proper explanation. He further said that Mr Holt’s model did not account for the effect on the Play Store’s operations of the reduction in costs and assets.

594    Now Mr Holt described the assumptions that he had made in relation to his treatment of the Play Store’s costs in the absence of sufficient granular Play Store data, and explained that his conclusions as to the appropriate characterisation of the various costs of the Play Store were determined having regard to both internal Alphabet documents and his professional judgment.

595    Further, and as the class applicants have said, Mr Holt adjusted his counterfactual analysis to respond to various issues raised by Professor Easton and concluded that the Play Store remained highly profitable across all of the applicable profitability metrics in each of the counterfactual scenarios.

596    Finally, Mr Holt considered that the Play Store’s high level of profitability makes it likely that market entry would have occurred in the counterfactual scenario resulting in a lower commission rate being paid by Android app developers.

Analysis

597    Now before proceeding further, generally speaking it will be more convenient to refer to Google Play given that the accounts are for a business division. But what I say can be taken as a proxy for the Play Store.

598    Now I largely agree with Google that there were various flaws in Professor Davila’s and Mr Holt’s profitability analysis.

599    Mr Holt adopted Professor Davila’s profitability and balance sheet estimates for Google Play and correspondingly the Play Store, notwithstanding that those estimates were flawed. But because Mr Holt did not properly seek to remedy the deficiencies in Professor Davila’s analysis, his analysis was also problematic for the following reasons, which Google has given and with which I largely agree.

600    First, Professor Davila failed to estimate the profitability of Google Play and correspondingly the Play Store on a standalone company basis. I have discussed this elsewhere.

601    It should be readily apparent that such analysis is necessary if the purpose of the inquiry is to understand the true economic profitability of Google Play and whether its profits are supra-competitive. This is because Google Play presently forms part of a conglomerate company, Alphabet, and benefits from the cost and capital efficiencies of conglomeration. The underlying profitability of Google Play needs to be isolated from the benefits of conglomeration in order to ascertain whether its profitability is truly attractive or supra-competitive, because otherwise, Google Play as a business unit will appear more profitable than it truly is.

602    In my view Mr Holt’s analysis suffered from the same deficiency, because he too made no real attempt to identify Google Play’s profitability on a standalone basis. Whilst he tried to qualify his concession in this respect by pointing to a calculation he performed with respect to allocating Android costs, which in the context of his sensitivity analyses did not remedy the deficiency, he accepted that aside from this, he did not seek to identify Google Play’s profitability on a standalone basis.

603    Second, as I have explained elsewhere, Professor Davila’s profitability analysis was based on Google Play’s internal management P&Ls, which cannot be viewed as a proxy for profitability on a standalone basis. This is because Google Play’s management P&Ls do not capture the full costs incurred by Google in order to support Google Play.

604    So, as I have explained elsewhere, Google Play relies on and is supported by significant investments made by Alphabet across all product areas within the Google Services segment of which Google Play forms one small part. But those significant investment costs do not appear on the Google Play management P&Ls. Take the costs of operating Android OS for instance, without which Google Play cannot exist. These costs are not allocated to the Google Play management P&L, notwithstanding that they are necessary to its operation.

605    Moreover, there are other costs that are not included or reflect only an allocation in a management accounting sense being general marketing, brand, R&D costs such as those relating to artificial intelligence and costs associated with supporting payments and ads platforms.

606    In other words, the management P&L does not include all costs necessary to operate Google Play that would be incurred if Google Play were an independent standalone company. No part of Mr Holt’s analysis cures this deficiency in Professor Davila’s analysis, including Mr Holt’s sensitivity analyses.

607    I have explained all of this in detail in my reasons in Epic v Google.

608    Third, as I have explained elsewhere, Professor Davila’s profitability analyses and ROA and ROCE ratios did not reflect all the costs and assets necessary to operate Google Play on a standalone basis. It follows that had Professor Davila assessed the profitability of Google Play on a standalone company basis, his gross profit and operating profit calculations for Google Play would have been different, and likely significantly so. The same result follows for the profitability estimates and ratios used by Mr Holt.

609    Fourth, as I have explained elsewhere, Professor Davila’s estimates did not include all the intangible assets necessary to operate Google Play’s business on a standalone basis, including because internally generated intangible assets like brand and intellectual property rights are not on Alphabet’s balance sheet and Professor Davila made no attempt to estimate these assets. Likewise, Mr Holt did not seek to directly estimate Alphabet’s internally generated intangible assets, nor did he seek to adjust Professor Davila’s profitability estimates for Google Play to account for these assets.

610    Now in cross-examination, Mr Holt tried to suggest that his sensitivity analysis with respect to R&D expenses was a proxy for an allocation of Alphabet’s internally generated intangible assets to Google Play. He even went so far as to suggest that his R&D expenses allocation sensitivity analysis was a proxy for Google Play’s share of internally generated intangible assets including a company’s brand. But I agree with Google that this evidence was problematic for several reasons.

611    Mr Holt’s R&D sensitivity analysis involved allocating a very small percentage of Alphabet’s R&D expenses to the Google Play P&L in each financial year, based on Google Play’s operating expenses as a proportion of Alphabet’s operating expenses. He also provided an alternative version that capitalized the R&D expense and amortised it over a period of 4 years. But Alphabet’s R&D expense in any particular financial year cannot possibly be a proxy for its internally generated intangible assets that existed during that year and that had been built up over several previous years.

612    Further, even if current year R&D expenses theoretically could be a proxy for presently existing internally generated intangible assets, there is no basis for Mr Holt’s suggestion, which I have mentioned above, that it would approximate the value of all such assets, including the Alphabet brand. Mr Holt’s claim to this effect in cross-examination was also undermined by his own report, in which he said that he considered R&D expense to be a reasonable proxy for the cost of intellectual property. There was no mention of it being a proxy for brand. Mr Holt tried to suggest in his joint report with Professor Easton that he considered that intellectual property generally includes a company’s brand. But this is not made clear in his first report.

613    In addition, as Professor Easton pointed out, R&D is a bad proxy for intangible assets like the Google brand, because there is simply no connection between R&D and the Google brand.

614    Mr Holt also tried to justify his non-inclusion of Google’s brand as an internal intangible asset of Google Play by saying that there is no evidence of the Google brand being relevant to Google Play. I reject this evidence.

615    So, in my view Mr Holt’s reliance on Professor Davila’s profitability and balance sheet estimates for Google Play as a basis for concluding that Google Play exhibited very high profitability is flawed.

616    Let me turn to Mr Holt’s sensitivity analysis. I agree with Google that each of the sensitivity analyses conducted by Mr Holt failed to cure the central deficiency in Professor Davila’s analysis, namely, that he did not calculate profitability on a standalone company basis. And most of the sensitivity analyses were adjustments that on any view ought to be made in any event.

617    As Google explained, Mr Holt’s first sensitivity made changes to the cash allocation method used by Professor Davila. This analysis increased the amount of cash allocated to Google Play relative to the allocations used by Professor Davila. But as Professor Easton observed, this sensitivity analysis was arbitrary and did not account for the substantial costs and assets that would be necessary to operate Google Play. In other words, it failed to grapple with the deficiencies in Professor Davila’s analysis.

618    Mr Holt’s second sensitivity was to adjust Professor Davila’s estimate of Google Play’s revenue, and costs of sales and opex, by lowering revenue by 30% and increasing costs of sales and opex by 30%. The supposed rationale for this analysis was to test whether Professor Davila’s estimates were sensitive to material adverse variances in the event that the management P&Ls were inaccurate. This sensitivity failed to address the deficiencies in Professor Davila’s analysis, because Mr Holt performed no analysis of whether this sensitivity could approximate all or even a substantial proportion of the costs necessary to operate Google Play.

619    Mr Holt’s third sensitivity was an adjustment based on allocating a proportion of Alphabet R&D expenses to Google Play. But for the reasons articulated above, Mr Holt was wrong to suggest that this sensitivity is a proxy for the internally generated intangible assets that were not included in Professor Davila’s analysis.

620    Mr Holt’s fourth, fifth and seventh sensitivities excluded various liabilities or revenues from Google Play that inadequately address the deficiencies in Professor Davila’s analysis. Mr Holt included as one of his sensitivities the exclusion of indirect revenues generated by Google Play for Alphabet’s benefit. But Professor Davila’s inclusion of indirect revenues in his profitability analysis was inappropriate in the first place.

621    Further, in relation to the inclusion of certain long-term liabilities, Mr Holt did not explain why this was appropriate other than to say that they were included for illustrative purposes. The class applicants’ submission that two of Professor Davila’s long-term liabilities were excluded by Mr Holt is not to the point when he has failed to explain why any should be included in the first place. Similarly, while Mr Holt removed browser ads revenue, he failed to explain why he did not exclude all advertising revenue.

622    Mr Holt’s sixth sensitivity related to Android costs. Mr Holt increased Professor Davila’s allocation of Android costs by 50% without addressing whether the revised allocation reflects the Android costs required to operate Google Play. The reasons why Professor Davila’s allocation of Android costs was insufficient have been addressed earlier. Mr Holt’s sensitivity did not address this deficiency, because his 50% increase still did not even allocate to Google Play a proportion of Android costs based on its proportion of revenue generated via Android OS.

623    In summary, I agree with Google that Mr Holt’s sensitivity analyses did not remedy the flaws in Professor Davila’s profitability analysis or provide the analysis with robustness.

624    There were other problematic features of Mr Holt’s analysis as Google rightly contended.

625    First, Mr Holt did not directly account for the intangible assets necessary to operate Google Play and his purported R&D proxy for this asset category was problematic.

626    Second, Mr Holt said that in order for Google Play’s ROA to be considered not above normal, that is, at or below 10%, Google Play would need to bear an aggregate reduction in operating profit of approximately USD 45.3 billion or an average annual increase in assets of USD 75.1 billion over the same period. But Mr Holt did not provide any basis for the 10% as a relevant benchmark for comparing Google Play’s estimated ROA.

627    Further, it is inappropriate to consider the numerator and denominator in isolation, for example, incorporating additional costs and capitalising those costs without considering how that would affect the estimated expenses and so the estimated operating profit of Google Play being the numerator of ROA, and the estimated assets being the denominator of ROA.

628    If the numerator is adjusted for increased expenses, it is likely that the denominator would be increased for the assets used to create those expenses, so the required reduction in expenses or increase in assets would be much lower than the figures put forward by Mr Holt.

629    In addition, it is clear from both the evidence of Professor Easton and Mr Cramer that substantial costs and assets are necessary to operate Google Play, which Mr Holt failed to account for. So, there is no basis for Mr Holt to speculate that the aggregated additional sums based on Professor Easton’s criticisms would not exceed USD 45.3 billion in costs or USD 75.1 billion in assets. I do not accept Mr Holt’s evidence in this respect.

630    Let me deal with another matter.

631    I agree with Google that the analysis of Professor Easton demonstrates that Mr Holt’s analysis of the profitability of Google Play under various counterfactual scenarios is counterintuitive. I have mentioned this already above, but let me touch on it again now.

632    In summary, Mr Holt’s model is flawed because it shows that Google Play’s ROGP would increase as its rate of commission decreased, assuming its volume and value of Google Play transactions remained the same.

633    For example, Mr Holt determined that Google Play’s ROGP would be 77% if it charged a service fee of 20% and maintained a 50% market share, but 89% if it charged a 10% service fee and maintained a 50% market share. Again, that result is not driven by any assumption that Google Play would win a greater share of transactions if it charged a lower service fee, because Mr Holt held Google Play’s market share constant.

634    Indeed, Mr Holt’s model suggested that, with an assumed 50% market share, Google Play would realise a higher ROGP if it charged a service fee of 1% than if it charged a service fee of 10, 15, 20, 30 or 50%. They are the product of flaws in Mr Holt’s model.

635    Mr Holt attempted to explain the counterintuitive trends in his model as being due to the economics of the Play Store. I do not accept Mr Holt’s explanation. The counterintuitive trends are driven by the assumptions that Mr Holt adopted.

636    First, Mr Holt assumed that most of the costs and assets of Google Play would vary proportionally merely with changes in Google Play’s revenue after a change in the commission rate. But there is no basis for assuming that Google Play’s costs and asset base would fall merely because its service fee changes in contrast to its market share. Costs typically vary with the level of activity, not with a change in commission rate. So, it is unclear that costs such as payroll would be lower if the commission rate were reduced but the market share, that is, level of activity, stayed the same. This assumption of the model is unsupported, which renders the model unreliable.

637    Second, the specific cost assumptions in Mr Holt’s model appear to be wrong in any event, even when regard is had to the document on which they were based. And although Mr Holt adjusted his counterfactual model in his supplementary report, the revised model remained problematical, partly for the reasons given by Professor Easton.

638    Finally, I agree with Google that Mr Holt’s conclusion that Google Play earned supra-competitive profits based on a comparison of Google Play's purported WACC against its ROCE is flawed in circumstances where he inappropriately used Alphabet’s WACC, as estimated by Bloomberg, to estimate Google Play’s WACC which assumed, erroneously, that the non-diversifiable risk of Google Play is the same as that of Alphabet.

639    In light of the foregoing, I accept that Google Play and accordingly the Play Store has high profitability at least in an accounting sense. But there are difficulties in drawing other conclusions.

The counterfactual scenario(s) – Google context

640    Before dealing with the specific question of counterfactual commissions, it is necessary to consider the broader potential elements of any realistic counterfactual scenario or scenarios and how the class applicants have put their case.

641    The class applicants say that absent the contravening conduct, alternative app stores would have competed with the Play Store for the provision of services enabling the distribution of Android apps.

642    It is said that by imposing the obligations under the MADA, Google required OEMs to pre-install the Play Store on Android smart mobile devices. This hindered competition for the exclusive pre-installation of alternative app stores on Android smart mobile devices and reduced the incentive for OEMs to install other app stores on their devices. I have not found a competition breach in this respect.

643    Further, it is said that whilst the terms of each MADA did not preclude OEMs from installing other app stores in addition to the Play Store, the use of RSA3s and MIAs incentivised OEMs not to do so. I have found a competition breach in this respect.

644    Further, it is said that these arrangements softened the incentive for Android smart mobile device users to install and use app stores other than the Play Store.

645    Further, it is said that Android app developers were prevented from distributing app stores or similar apps through the Play Store, meaning that developers who wished to distribute alternative app stores, or Android smart mobile device users who wished to download such apps, were limited to sideloading. I have not found a competition breach in this respect.

646    Further, it is said that but for the contravening conduct, competition would have exerted considerable downward pressure on commission rates, with app stores competing on commissions so as to incentivise Android app developers to drive Android smart mobile device users to the platform that was most profitable to those developers.

647    Further, it is said that but for the Project Hug, GVP agreements and AVP agreements conduct, rival app stores would have been able to compete more effectively. I have found a competition breach concerning this conduct.

648    Further, it is said that but for the contravening conduct, the charges paid by Android app developers for in-app purchases, including subscriptions, within the Play Store by way of commissions to Google would have been significantly more competitive. I have found a competition breach concerning the in-app payments solution question. And on this particular aspect, the class applicants make the following points, with which I agree.

649    First, there would have been competing payment solutions offering Android app developers more favourable terms.

650    Second, Android app developers would have likely taken the opportunity to use alternative payment solutions and take advantage of the cost savings.

651    Third, Google would have experienced competitive pressure to reduce its commission. If Google were to choose not to reduce its commission, Android app developers would have been likely to switch from Google Play Billing.

652    Now I have not found that Google engaged in any contraventions save and except concerning Project Hug, the GVP agreements, the AVP agreements and the RSA3s and also concerning the Android in-app payment solutions market. That being the case, the relevant lens is narrower concerning the relevant counterfactual.

653    Let me turn to Google’s position first.

Google’s arguments

654    Let me address a number of points raised by Google which are focused on the counterfactual scenarios relevant to the contraventions that I have in fact found.

655    First, let me consider the question concerning the RSA3s and Google’s position.

656    The class applicants say that absent the terms of the RSA3s, OEMs will not be disincentivised from pre-installing alternative app stores, including the Epic Games app. Further, they say that absent the terms of the RSA3s, rival app stores including the Epic Games Store would enter the alleged Android mobile app distribution market and compete with Google Play, including by offering developers lower fees.

657    But Google says that this contention is speculative and unsupported by the evidence.

658    It says that the fact that the RSA3s did not foreclose rival entry is proven by Epic’s conduct. Epic’s announcement in early 2024 that it would launch the Epic Games Store by the end of 2024, including on Android, suggests that rival app stores can, in fact, enter the market and compete with the Play Store. As such, the existence of the RSA3s has not precluded competition, and there is no evidence or analysis of any material impact on competition.

659    Further, it says that the absence of RSA3s would not have materially changed the status quo, particularly having regard to the data on activation rates. Removing the Premier tier provisions would have only made 7.7% of Android phones available for a second app store. Put another way, the presence or absence of RSA3s does not make any difference in respect of over 92% of Android devices in Australia.

660    Similarly, it says that the presence or absence of RSA3s would not have made any difference to OEM incentives or the ability of Epic and others to enter the market and compete against the Play Store. Google said that the RSA3s were not enough to move the needle in those circumstances.

661    Further, it says that the relatively short period of time for which RSA3s were in place further diminishes any potential market impact they could be argued to have had.

662    I do not need to say much about the RSA3s and the relevant counterfactual consisting of their absence. This is because I did not find any anti-competitive effect or likely effect in any event as I discussed in my reasons in Epic v Google.

663    Second, Google says that there would be no meaningful change in the counterfactual absent Project Hug and the GVP agreements because they had only a modest likely effect. So removing them in the counterfactual would make no meaningful change.

664    Third, let me say something concerning the Android in app payment solutions question and the counterfactual and Google’s position.

665    Google says that any proper counterfactual analysis in respect of the Google Play Billing requirement first requires the precise identification of what, if anything, would change in the counterfactual. I have referred elsewhere to this requirement as the payments tie; there is also of course the anti-steering rule.

666    Now Google says that the disconformity between the boundaries of the product that defines Epic’s market and the boundaries of Google Play Billing creates analytical difficulties when it comes to the counterfactual.

667    Epic’s and the class applicants’ case is that, in the future without the payments tie, developers could choose whether or not to use Google Play Billing.

668    But Google says that this proposition obscures the problem that it is not immediately obvious what it would mean for a developer to forgo Google Play Billing in circumstances where it is a tightly integrated aspect of the Play Store, comprising both payment processing aspects and non-payment processing aspects.

669    Google says that any assessment of competition in the counterfactual must identify the commercially realistic likelihoods. Those likelihoods cannot be identified, and the question of competitive effects assessed, without some attention to the aspects of the Play Store’s current product that would be unbundled in the counterfactual, including because that will affect the assessment of whether there would, in fact, be material demand from developers to obtain an alternative to Google Play Billing.

670    Google says that one reason there would likely be no material demand from developers for a payment solution other than Google Play Billing in the counterfactual is because developers could not otherwise obtain the full suite of features and functions that Google Play Billing provides.

671    It says that one consequence of this is that any developer who chooses to forgo Google Play Billing would likely need to acquire a whole range of services in addition to payment solutions from different suppliers. Even then, developers would likely be unable to replicate all the features and functions that they obtain from Google Play Billing, which would affect their willingness to obtain a payment solution other than Google Play Billing.

672    Further, Google says that it has not been demonstrated that there is or at any relevant time was a real commercial likelihood that there would be material demand for alternative in-app payment solutions in a future without the payments tie.

673    Further, Google says that there is no evidence that any alternative provider would supply payment solutions in the counterfactual at a competitive price.

674    Ultimately, Google says that there is no reliable evidence that would permit me to conclude that, in a world without the payments tie, competition would be increased.

675    Further, Google says that a further difficulty with Mr Holt’s analysis relates to his theory of causation. Mr Holt’s analysis depended upon the assumption that, in the retrospective hypothetical counterfactual, there would have been more price competition between app stores in terms of their headline commission rates. Google says that that is not a sound assumption on the evidence. Indeed, Google says that what the evidence shows is that, in the factual, smaller Android stores do not typically compete on headline commission rates.

676    Google says that what was inherently unlikely to happen in the retrospective hypothetical counterfactual was a reduction in headline commissions for all developers. Different developers are differently situated and so their commissions and counterfactual commissions may be different.

677    But I may as well note here, and as the class applicants have said, Mr Holt expressed the view that there would likely be more competition including on price. But his evidence was not limited to competition only at the headline rates. And given that headline rates are often negotiated down, the more relevant lens was to consider effective commission rates.

Analysis

678    Now I agree with the class applicants that in the counterfactual, there would have been likely entry or a credible threat of entry of new rival app stores and effective or substantially more effective competition from existing app stores within the Android app distribution market, particularly given the freeing up concerning alternative payment solutions which may have encouraged more entry into the Android app distribution market as well as the payments solution market. But I will need to hear further from counsel on these matters.

679    Further, I agree with the class applicants that as a consequence of either new entry into or effective competition within the Android app distribution market or the credible threat of that entry or competition, there would likely have been price competition in the counterfactual and the Play Store’s service fee would likely have fallen.

680    I have said elsewhere that Google’s 30% fees were in one sense set not by reference to competitive forces. And as the class applicants point out, they were always vulnerable to competition in the absence of the contravening conduct. I have addressed this in my reasons in Epic v Google.

681    Indeed, Google’s documents reflect that prior to and during the relevant period, Google may have had an appetite to consider dropping its service fee to as low as 5% in response to competitive pressure and bargaining by counterparties, although not necessarily across the board.

682    Further, in the absence of the contravening conduct, it is likely that Paddle would offer an Android version of Paddle with all the features found on other platforms and a 10% transaction fee for transactions below USD 10 million, and a 5% transaction fee plus USD 0.50 for all transactions above USD 10 million. But I accept that Paddle does not presently offer all of the relevant features of Google Play Billing.

683    Further, the prices offered by PC app stores fell by up to 20 percentage points following the entry of the Epic Games Store. As Mr Holt concluded, in the PC app distribution market “price competition has been intensifying in recent years and has driven commission rates down, with some coming down to between 12% to 15%.”

684    Further, I agree with the class applicants that any competition on quality in the counterfactual would not have been to the exclusion of price competition. Generally speaking, competitors can implement price competition faster than competition on quality, stemming loss of market share. But I do not accept that price competition can and would be implemented faster in all scenarios, but it is not a bad generalisation to make as the class applicants assert.

685    But I do accept that in some scenarios, complex price changes may take some time to implement. Moreover, I do accept that in some scenarios, a firm which anticipates future competitive pressure may seek to prioritise offering more quality value first rather than implementing a price change.

686    Generally, I agree with the class applicants that there is likely to have been greater price competition in the counterfactual such that the Play Store’s service fee was likely to have fallen by a significant amount.

687    Now Google has sought to resist such an idea and has made the following points, most of which are underwhelming to say the least.

688    First, Google says that the Play Store’s headline service fee is already fully competitive. In my view that assertion does not properly reflect the evidence. Moreover, parallel headline rate setting, even if it occurs, is hardly any reflection of vigorous competition.

689    Second, Google’s reference to the Chinese rate of 30 to 55% as somehow justifying the competitiveness of its own rate was not even a good jury point.

690    Third, Google said that there was no increase in the Play Store’s service fee following the introduction of the RSA3s and Project Hug. But I would just note here that this is not much of a point if the fee is well above what would otherwise be a competitive price if there was meaningful competition.

691    Further, Google would have been under greater competitive constraint in the counterfactual and commissions are likely to have fallen.

692    Further, the post-2019 conduct committed participating developers to refrain from offering their apps and services at a lower cost on alternative app stores. The sim-ship and feature parity clauses in the relevant GVP agreements and the AVP agreements associated with the implementation of Project Hug also prevented developers from offering their Android apps on alternative app stores earlier than on the Play Store or with different features. Absent these agreements, Google considered there was a risk that the Epic Games Store and Samsung Galaxy Store would have obtained exclusive titles and would have posed a greater competitive constraint on the Play Store. In the counterfactual, therefore, Google would have been under greater pressure from these app stores. Now I did not find an anti-competitive effect or likely effect concerning the AVP agreements but I did find an anti-competitive purpose. So in the retrospective hypothetical I must assume their absence.

Alternative fees charged by Google

693    Let me turn to the question of whether and what alternative fees would or might have been imposed by Google in the counterfactual. I will also deal with the notion that in the counterfactual, Google would or might have reduced investment in the Play Store or the Android ecosystem.

694    Google says that it would have responded to alternate app stores and in-app digital content payment processors competing in the relevant market(s) by reducing investment in the Play Store or the Android ecosystem, or both, or adopting alternate monetisation options. But for the reasons that follow, which have been largely given by the class applicants, I would reject such claims save that there may be alternative monetisation mechanisms but not such as to prop up commissions at the current levels.

695    Now Google has never imposed a fee for free apps, app updates and in-app purchases of physical goods, and would be unlikely to do so in the counterfactual. This is because Google recognises that while only a limited number of developers pay for the Play Store, all other developers contribute to the ecosystem by providing app diversity and innovation and in this way enhance the value of the ecosystem. By attracting users, Google generates revenue through its proprietary GMS apps, advertising, and increasing the value of Android OS.

696    Google has revenue streams outside of the Play Store that benefit from Android’s openness, including serving ads on third party apps and websites. Mr Feng accepted that such revenues outside the Play Store incentivised Google to invest in the Play Store. Mr Feng further accepted that a vast majority, around 80%, of Google’s revenue comes from advertising, and that this is maximised by making the Android environment an attractive platform for OEMs and consumers.

697    Later in his evidence, Mr Feng explained why Google had never previously charged for free downloads:

The reasons would be that if you charge for free downloads, it substantially increases the barrier to entry for innovation. It will substantially reduce the diversity of apps that are available to users, and developers really hate the idea.

698    Users would be unlikely to acquire Android devices if Google did not have an app store with a diverse range of attractive apps. Google receives a benefit from having developers or users on Android, whether they are paying users or not, if they use other Google services.

699    Clearly Google obtains a wider benefit from having popular apps on Android insofar as it incentivises OEMs from adopting Android with Google advertising apps pre-loaded such as YouTube and Google Search. And it is in Google’s interests for there to be a large and growing number of Android users, and for there to be a large body of Android app developers seeking to distribute their apps to users.

700    There are significant network effect benefits from the Play Store distributing free apps insofar as it supports the general attractiveness of the OS enabled smartphones. Indeed there are various ways that the GMS apps generate revenue for Google after the consumer has stopped having anything to do with the OEM.

701    I agree with the class applicants that in circumstances where Google obtains such significant economic benefits from having an app store with a diverse and high quality range of apps available, it would not be economically rational for Google to impose an alternate fee structure or reduce its investment which would have the effect of reducing the attractiveness of the Play Store or the Android system more generally.

702    Further, I agree with the class applicants that Google would have been unlikely to reduce its investments in the Play Store even if had been exposed to a counterfactual in which the contravening conduct was absent, there was significant downward pressure on its Play Store service fee and Google did not impose any additional fee.

703    Now Mr Feng said that if Google had to reduce the service fee in a way that materially reflected revenues, it would need to recover those revenues by alternate monetisation options in order to continue investment and funding to build the Play Store and Android. But Mr Feng did not identify any figures to support these assertions, nor any documents or records which would establish them. As the class applicants point out, he also accepted that any ultimate decision about sources of revenue would not be his, and that his opinion that Google would stop investing in the Play Store was based on not much more than a simple analysis that “as a for profit entity, company invests in areas where we think we can get a return. So if the profitability is substantially less, revenue is less, we would need to find a way to – to improve the revenue in order to invest”.

704    Further, as the class applicants point out, Ms Kochikar also implied in her affidavit that a reduction in the service fee would reduce the amount Google was able to invest in the Android ecosystem without alternative monetisation mechanisms to ameliorate the revenue impact. But like Mr Feng, she conceded that she was not in a position to quantify the amount by which any specific reduction in the service fee would impact Google’s investment in the Android ecosystem but maintained that it would have an impact despite her not being able to quantify it. Nor had she identified any documents in her affidavit which recorded such an amount.

705    Further, I agree with the class applicants that the most commercially realistic likelihood in the retrospective counterfactual is that Google would likely have materially lowered its commission rates and, to satisfy any need or desire to maintain existing revenues, it may have adjusted its monetisation strategy for the Play Store by increasing advertising revenue rather than imposing alternate fees to ameliorate any lost revenue occasioned by a reduction in the headline service fee.

706    It seems that the advertising functionality on the Play Store is an alternate way of monetising the distribution of apps on the Play Store. It is another way that the Play Store could derive revenue and it was a realistic option that Google could change its monetisation strategy in this way. Now Google said that any increase in ads on the Play Store may be regarded as a quality degradation, but this point may be debated.

707    Further, I agree with the class applicants that it is inappropriate to assess a retrospective counterfactual in which Google devises separate and novel anti-competitive measures. If one is going to impose some charge that makes entry of alternate app stores not viable, then that would have the same or similar effect as the relevant impugned conduct. I cannot assume that that would be part of any realistic counterfactual.

708    To the extent that any such measure involved the addition of new charges to replace revenues lost due to competition, such measure would be circular and problematic as the class applicants say.

709    Further, as the class applicants point out, Google’s willingness to levy an alternate fee would be significantly constrained by the capacity of a rival app store to distribute its store by alternate means than the Play Store in the retrospective counterfactual, including through sideloading or negotiated distribution through OEMs. It is likely that these alternative distribution channels would place significant competitive pressure on any alternate fee that Google may wish to levy.

In-app payment solutions question

710    Let me say something about the in-app payment solutions question.

711    I agree with the class applicants that in addition to new entry into the Android app distribution market, in the counterfactual there would have been the credible threat of entry, and likely entry, into the in-app payment solutions market. The class applicants have made four general points which I agree with and the evidence for which I have set out elsewhere.

712    First, there would have been app developer demand for payment solutions other than Google Play Billing in the counterfactual.

713    Second, a range of suppliers of such payment solutions including Stripe, Adyen and Braintree are already extant and operating within Australia. One such provider is Paddle.

714    As I have indicated elsewhere, Mr Owens, the CEO of Paddle, said that the size of the potential customer base would have been meaningful from a revenue standpoint for any in-app payment solutions business, indicating that there would have been significant incentives for existing solutions providers to enter the market.

715    Mr Owens said that if Google had allowed alternative payment solutions providers to facilitate in-app purchases, Paddle would have significantly increased its business, and provided developers with an alternative solution for a lower price and with better features. Mr Owens’ evidence was that if Google permitted developers to use Paddle for purchases of digital content made within native apps downloaded from the Play Store, without requiring those developers to pay Google a fee which made it more expensive for them to use Paddle than to use Google Play Billing, Paddle would develop a version of Paddle for apps downloaded from the Play Store.

716    Further, given the wide range of payment solutions that are being used by merchants selling non-digital products within apps, one would expect merchants selling in-app digital content on Android smart mobile devices to similarly use a range of payment solutions to the exclusion of Google Play Billing if permitted to do so. Further, the payment solutions used for digital content did not differ from those used for physical goods.

717    Third, there is also evidence of app developers choosing alternative payment solutions on other app stores where such choice is provided. This is supported by the evidence of Mr Owens. Mr Owens said that following development of Paddle’s iOS facility and accompanying marketing campaign, Paddle received more than 1,000 inquiries from developers who wanted to use the system in preference to Google Play Billing. That enthusiasm would presumably be replicated if an Android solution was similarly announced.

718    Fourth, it is not technically difficult for third-party payment solutions to be integrated with the Play Store, nor for developers to design their apps to support solutions other than Google Play Billing. The evidence before me demonstrates that it would be technically feasible for the Play Store to be compatible with alternative payments solutions and that it is already done in countries like South Korea. Indeed, as the class applicants point out, approximately 5,000 developers used alternative payment providers on the Play Store prior to the policy change in 2021.

719    Furthermore, if Google permitted alternative payment solutions, these standalone payment providers would be able to enter the market rapidly, including because some of them already facilitate the sale of physical goods on Android native apps or offer payment solutions on web-apps or PC native apps.

Comparator analysis — Google context

720    Now as the class applicants have explained, under the comparator analysis, the Play Store commission was compared by Mr Holt to a comparator in another market, where the comparator could be expected to have behaved in a similar manner to the market in question, absent the anti-competitive behaviour.

721    Mr Holt said that this was an appropriate approach to estimating commission overcharge in circumstances where there is no other geographic market, or period of time, in which the Play Store was subject to competitive pressures.

722    The class applicants said that the most relevant comparators can be identified based on assessing their offering of similar services in a similar industry and using a similar business model, assessing their competitiveness or, at least, a greater degree of competitiveness than the present markets, and assessing their similar pricing structures.

723    Further, they said that PC app distribution platforms are comparable to the services provided by the Play Store in several important respects, including in respect of the manner in which app stores offer access to apps, which can be purchased online and downloaded.

724    Further, they said that Apple’s App Store is not a useful comparator, as the App Store has a monopoly on the distribution of iOS mobile apps and does not allow other app stores to be distributed through the App Store.

725    But they said that PC app distribution platforms exhibit a substantially greater degree of competition than the Android app distribution market, with competition between competing app distributors such as Steam and Epic which distribute a range of apps published by multiple developers, and also with app publishers which operate their own distribution platforms for their own apps, such as Electronic Arts and Ubisoft.

726    Now Mr Holt explained that there are three broad categories of comparator-based approaches to determining a counterfactual commission. I have addressed these earlier in the Apple context.

727    Mr Holt concluded that the yardstick approach as applied to a comparator to the Android app distribution market was most likely to yield reliable results.

728    Mr Holt selected the PC app distribution market as an appropriate comparator on the basis that it exhibited similar industry and business models to the Android app distribution market, it displayed a high degree of competition and it had comparable pricing structures in the sense of charging percentage commissions. Mr Holt’s approach was consistent with that adopted by Ms Edwards-Warren.

729    In his response to Professor Holt’s report, Professor Asker agreed with the descriptions of the three potential methodologies identified by Mr Holt. But Professor Asker said that the counterfactual commission could be determined by benchmarking against the commission rate charged by Android app stores in a different region, China, where the contravening conduct has not been present, and against the commission rate charged prior to the implementation of some but not all of the contravening conduct.

Google’s arguments

730    Now Google says that Mr Holt’s yardstick analysis is flawed and unreliable.

731    First, it says that Mr Holt did not consider differences between PC app stores and Android app stores that may explain differences in pricing between the two. So, he did not consider the relevance of the fact that operating systems and games are monetised for PCs in a different manner from Android.

732    Differences of that kind may explain price differences between the service fees charged by PC app stores and Android app stores, but Mr Holt did not consider that possibility.

733    Second, it says that Mr Holt made no attempt to identify which PC app store competitors are the best comparators to the Play Store. That step is necessary because there is actual price dispersion in app distribution on Android already, with not all competitors charging the same service fees.

734    So, as part of the Samsung collaboration between Epic and Samsung, Epic agreed to pay Samsung a revenue share of 15% on Fortnite in app purchases for the first 18 months and 12% after that. That compares to the Play Store ’s relevant service fee of 15% for the first USD 1 million in revenue, and 30% after that.

735    Given that one can see from these figures that Android app stores do not charge the same service fee, any counterfactual pricing should reflect this same price dispersion, including by identifying the competitor who is the best proxy for the Play Store.

736    Professor Asker said that Google’s service fee is already comparable, and often lower, than the service fees charged by Steam, which is the only one of the four comparators selected by Mr Holt that Professor Asker said might be a useful comparator.

737    Third, it says that Mr Holt inappropriately omitted some PC app stores from his analysis that charge higher service fees than the Play Store.

738    So, Mr Holt omitted the Amazon Appstore, which charges a [REDACTED] service fee. Mr Holt provided no reasoning for his exclusion of the Amazon Appstore as a comparator from his analysis.

739    Google says that in Mr Holt’s yardstick analysis he inappropriately omitted relevant comparators from his yardstick analysis. This is a topic that I have discussed elsewhere. But there are two further points in this regard.

740    Google says that Mr Holt omitted a PC app store, GOG, which has a headline commission rate of 30%. Google says that he did so based only on the evidence of Mr Allison to the effect that GOG is known for entering into custom revenue share models with different developers. Google says that that was the only evidence he was aware of to suggest GOG charges any developers less than 30%. But that was an insufficient basis to exclude GOG from his yardstick analysis and it should have been included.

741    The mere fact that an app store may offer some developers custom rates does not mean that they are not a relevant comparator for the Android mobile app distribution market, including because some Android app stores take the same approach.

742    But this is incomplete. As the class applicants point out, Mr Holt also excluded GOG as it accounted for few sales among developers who offered their games on GOG.

743    Further, Google says that Mr Holt omitted the Amazon Appstore from his analysis, even though there is no suggestion that its commission structure was affected by the impugned conduct, and notwithstanding that it distributes apps both on Android and on PC, in the case of Android apps capable of running on PCs.

744    Fourth, and related to the second and third points that I identified earlier, Google says that Mr Holt did not consider differences in quality between app stores. Mr Holt’s analysis did not account for the reality that higher quality products appropriately command higher prices. For this reason, in any yardstick analysis, the comparison must be based on quality-adjusted prices, not nominal prices alone.

745    Google says that Mr Holt’s yardstick analysis is incapable of being used to assess the counterfactual commission that the Play Store would have charged because it does not account for quality differences between app stores or seek to determine which app store is the best comparator for the Play Store. Google says that any yardstick analysis in a differentiated product market must take into account differences in quality because the exercise is directed at determining quality-adjusted counterfactual prices.

746    Google says that given the quality and catalogue of the Play Store, its most obvious comparator amongst PC app stores is Steam.

747    Further, Google says that Mr Holt included the Epic Games Store as one of his comparators to the Play Store. But there are significant differences between the Epic Games Store and Steam. Whereas Steam offered 50,046 titles in 2021, the Epic Games Store inventory of games is much smaller, being only 1,548 titles in 2022. That is consistent with Steam being a higher quality app store than the Epic Games Store, if only because of its much larger range of titles.

748    Google says that economic reasoning indicates that, in these circumstances, the Epic Games Store would need to offer a lower service fee than Steam. That being so, Mr Holt was wrong to treat the Epic Games Store fee as indicating a market price for PC app distribution. Epic charges less than Steam because its product is not as good as Steam.

749    Google says that in circumstances where the Play Store has 2.5 billion monthly active users and around 3 million apps, the pricing charged by the Epic Games Store for PC app distribution is a poor comparator to the Play Store.

750    Fifth, Google says that Mr Holt’s analysis was unable to predict different counterfactual service fees depending upon which, if any, of the discrete elements of the impugned conduct is considered. A reliable economic framework to calculate counterfactual prices should be capable of isolating the price impact of each specific element of the challenged conduct.

751    Sixth, it says that Mr Holt ignored feasible yardstick and benchmarking analyses that are more informative. Professor Asker considered two such benchmarks or yardsticks.

752    Professor Asker analysed service fees charged by third-party app stores in China where the Play Store is not made available by Google and where third-party app stores are more numerous.

753    Further, Professor Asker considered the Play Store’s service fees before and after the commencement of certain elements of the impugned conduct, being Project Hug and the RSA3s. Professor Asker determined that the Play Store’s fees decreased after that conduct, suggesting that that conduct did not lead to higher prices, including because Mr Holt did not identify any evidence to suggest that the decreases in the Play Store’s service fee would have been larger absent the impugned conduct.

754    Seventh, Google says that Mr Holt’s approach did not explain why various sources of apps and content have commission rates of 30% or even more, in multiple contexts that very clearly do not correspond to any market power or lack of competition. Mr Holt did not deal with this at all.

755    Further, Google says that Mr Holt’s analysis inappropriately extrapolated the commission rates charged by his selected comparators at the end of the relevant period to the entire relevant period. In doing so, Mr Holt overrode his yardstick results and selected different outcomes, which is inconsistent with the yardstick methodology.

756    Let me say something more concerning Google’s submissions as to the use of Steam’s commissions.

757    Mr Holt gave evidence that the top of his counterfactual range was derived from Steam’s commissions. But Google says that Mr Holt selected the rate that Steam only gives to developers with revenues of over USD 50 million on Steam.

758    Google says that there are two problems with Mr Holt’s approach.

759    First, it says that insofar as one is doing a yardstick analysis for the whole of the Android mobile app distribution market, and using Steam as a comparator in that analysis, it makes no sense to focus only on Steam’s lowest rate offered to the small minority of developers with revenues over USD 50 million. If one is using PC app stores as the yardstick, one must take them as one finds them.

760    Second, it says that Mr Holt did not, in fact, understand how Steam’s commission structure worked. Mr Holt believed that Steam charged a 20% flat commission for developers with revenues above USD 50 million, but that is not correct. The true position is that Steam charges a 30% commission to all developers, regardless of revenues, on the first USD 10 million of revenues they earn on Steam, a commission of 25% on revenues earned between USD 10 million and USD 25 million, and a commission of 20% on revenues above USD 50 million.

761    It says that the error is significant because reasoning consistently, the top of Mr Holt’s counterfactual range should have been above 20%, as it is clear the blended average commission for developers earning revenues on Steam above USD 50 million will be larger than 20%.

762    But more fundamentally, Google says that when the correct Steam commissions are used the Play Store’s commissions appear to be lower than Steam’s for the vast majority of developers who pay a service fee to Google.

763    It says that on Professor Asker’s estimate, based on the global Play Store revenues of US developers that publish on the Australian Play Store, 99.6% of developers who pay a service fee to Google would have a lower service fee on the Play Store than on Steam. Further, Professor Asker found that for developers who paid a service fee to Google, there would be a 1% difference between the average service fee they paid to Google as compared to what they would have paid to Steam based on Steam’s commission structure.

764    It says that this implies that if Steam is the correct comparator for the Play Store, the Play Store’s commissions were competitive.

765    In summary, Google says that Mr Holt’s yardstick analysis should not be accepted.

Analysis

766    Generally speaking, I have not accepted Google’s arguments.

767    Now I would make similar points here concerning Mr Holt’s comparator analysis that I have made earlier in the Apple context concerning comparators with the App Store. I will not repeat that discussion.

768    Now Professor Asker and consequently Google advanced a number of criticisms of Mr Holt’s PC app distribution yardstick approach. I will focus on Professor Asker’s criticisms for the moment.

769    First, Professor Asker said that Mr Holt did not include in his analysis two PC app stores, being Good Old Games (GOG) and the Amazon App Store.

770    Second, Professor Asker said that Mr Holt had not presented sufficient evidence to exclude the Mac app store as a relevant app store within the PC app distribution market.

771    Third, Professor Asker said that Mr Holt failed to account for differentiation within the PC app distribution market by focusing on the commission rates charged by the market as a whole rather than focusing on the participant in that market that most resembles Google.

772    But as to the first proposition being that Mr Holt failed to take into account two relevant PC app stores, as the class applicants point out, Mr Holt said that GOG was considered but excluded from his analysis of the PC app distribution market on the basis that GOG’s headline commission rate is negotiable, and GOG is “known for entering into custom revenue share models with different developers”. So, Mr Holt considered that it was unclear what GOG’s effective commission rate was, and therefore it would not be prudent to include GOG in the overall analysis.

773    Similarly, as the class applicants point out, Mr Holt excluded the Amazon App Store from his analysis of the PC app distribution market on the basis that it was not a distribution platform for PC apps but rather was used to distribute Android apps on Android smart mobile devices, Amazon’s proprietary “Fire” devices and Windows devices capable of running Android apps.

774    In any event, I agree with the class applicants that Mr Holt’s analysis of additional documents discovered by Google relating to the Amazon App Store suggests that the effective commission rate on the Amazon App Store across all devices between 2018 and 2021 supported his position.

775    As to the second proposition in respect of the Mac app store, as the class applicants point out, Mr Holt said that the Mac app store commission rate of 30% is not a reliable datapoint for the purposes of determining a competitive commission rate because Apple has a policy of setting a uniform commission rate across all of its app stores rather than in response to competitive pressure. And as the class applicants say, there is also little evidence to suggest that the 30% commission rate on the Mac app store was set in response to any competitive benchmark.

776    In my view, it was appropriate for Mr Holt to exclude the Mac app store as a reliable datapoint for the purposes of establishing the counterfactual commission.

777    As to the third proposition being that Mr Holt failed to accommodate differentiation in the PC app distribution market and instead should have focused on the commission rate charged by the participant in the market that most closely resembles Google, which Professor Asker said is Steam, this proposition is problematic.

778    As the class applicants point out, Mr Holt’s approach of determining a 10 to 20% range for the counterfactual commission was based on his understanding that there was likely to be differentiation in the counterfactual market which would mean that some Android app developers would offer lower-quality, lower-priced distribution platforms while others focused on higher-quality app stores which charged commensurately higher commission rates. Indeed, as was pointed out, when markets consist of differentiated products, that is, products with differing quality, the market equilibrium will sustain multiple prices, with firms selling higher-quality products charging higher prices.

779    So, in arriving at the counterfactual commission, I agree with the class applicants that it would be misleading to focus on the commission charged by a single entity within the comparator market rather than the commissions charged by a range of stores within that market, since, in the absence of the contravening conduct, many of the transactions that occurred on the Play Store would likely have occurred on competing platforms.

780    I also agree with the class applicants that it is unlikely that Steam is the relevant counterpart to the Play Store in the PC app distribution market.

781    It is incorrect to assume that the quality of the Play Store in the counterfactual would be the same as in the actual, since the latter is at least in part a product of the contravening conduct. Absent the contravening conduct, rival Android app stores would have had greater ability and incentive to innovate and invest in the dimensions of quality that Professor Asker cited.

782    As the class applicants say, to the limited extent that it may be helpful to identify the most direct comparator to the Play Store in the PC app distribution market, this is likely to be the Microsoft Store. It is the only PC app store that does not primarily focus on games and so carries a variety of applications. The Microsoft Store also exhibits many other similar characteristics to the Play Store. Moreover, Professor Asker’s trenchant criticism of the lack of quality of the Microsoft Store lacked foundation.

783    Contrastingly, Steam focuses only on games. In contrast to the Play Store, Steam offers community features such as access to user forums to discuss games, a hub for sharing user-created content or tools relating to games, a community market for users to buy and sell in-game items, and a platform for users to share content such as user guides to games, artwork and videos. Community features such as allowing users to customise games and the ability to trade virtual items also benefits developers, as such features are associated with a price premium and a slower decline in prices over time.

784    Let me say something more about Steam.

785    Now Professor Asker said that the Play Store’s fee is comparable to and often lower than the commission charged by Steam. But these assertions are problematic for the following reasons as the class applicants pointed out.

786    First, Mr Holt’s approach of determining a 10 to 20% range for the counterfactual commission accommodates differentiation in the counterfactual market in that some Android app developers would have offered lower-quality, lower-priced distribution platforms while others focused on higher-quality app stores which charged commensurately higher commission rates.

787    In considering the comparator PC app distribution market, it would be wrong to focus on the commission charged by a single entity rather than the commissions charged by a range of stores within that market.

788    Second, and in any event, Google has not established that Steam is the counterpart to the Play Store in the PC app distribution market.

789    In contrast to the Play Store, Steam offers community features that are not present on the Play Store. Indeed, Professor Asker accepted in cross-examination that there are significant differences between the Play Store and Steam. Whilst Steam is an app store focused on gaming apps, the Play Store provides distribution services for both gaming and non-gaming apps. Further, Steam offers a forum for discussing games whereas the Play Store does not offer any equivalent service or feature. Further, Steam offers a hub for sharing user-created content and tools relating to games whereas the Play Store does not offer any equivalent service or feature. Further, Steam offers a community market for users to buy and sell in-game items whereas the Play Store does not. These community features of Steam that are absent on the Play Store are associated with a price premium and a slower decline in prices over time.

790    Moreover, as Mr Holt said, it is incorrect to assume that the quality of the Play Store in the counterfactual would be the same as in the actual, since the latter is at least in part a product of the contravening conduct. Absent the contravening conduct, rival Android app stores would have had greater ability and incentive to innovate and invest in the dimensions of quality that Professor Asker cited.

791    Third, Professor Asker said that Steam is more informative of the market counterfactual than other app stores because his preferred approach would consider the prevalence and the extent to which any of these PC app stores are actually used.

792    The implicit assumption is that Steam has market dominance in the PC app distribution market over other app stores within that market. But the evidence does not support that proposition.

793    The high point of the evidence cited by Professor Asker is an internal Google presentation from 2021 which asserts that Microsoft Store lags Steam in popularity as a games launcher. But this does not disclose the relative market shares of the two entities. It concerns the purported popularity of each as a games launcher rather than as a PC app store. Further, it relates to a single pinpoint date in time rather than establishing the position over a time-series. And it fails to have regard to the market position of any other app store in the PC app distribution market.

794    Fourth, the assertion that Steam is the only reliable app store comparator within the PC app distribution market contradicts Professor Asker’s contention elsewhere that, in having regard to the PC app distribution market, Mr Holt’s approach was flawed because he failed to have regard to GOG and the Amazon App Store.

795    Fifth, whilst Professor Asker sought to draw attention to Steam’s headline rates, it is not in dispute that Steam offers discounts from its headline rates.

796    Faced with that reality, Professor Asker and Google sought to suggest that those discounts are limited. But the evidence establishes that prior to 2023 and throughout the relevant period there were no limits to the discounts offered by Steam.

797    Moreover, since at least November 2018, Steam has offered a 25% commission to developers with revenue between USD 10 million and 50 million, and a 20% commission rate to developers with revenue above USD 50 million.

798    These two factors mean that, as Mr Holt said and as the class applicants correctly contended, Steam’s effective rate is likely to be materially lower than 30%. Mr Holt was therefore correct not to extend his counterfactual commission range to 30% by reason of Steam’s top headline rate.

799    Sixth, Professor Asker’s attempt to establish that the Play Store’s service fee is comparable to and often lower than that of Steam has no foundation. Professor Asker purported to establish this proposition by examining a sample set of US-based developers that published on the Australian Play Store in 2021. He relied on this evidence to suggest that the vast majority of developers pay less fees on the Play Store than on Steam.

800    But there is no evidence that this is a representative sample. Further, Professor Asker only accounted for headline rates on both Steam and the Play Store. He also attributed to the Play Store’s rates a 15% rate for all apps and in-app purchases revenue less than USD 1 million. This is despite the fact that the data on which he relied related to 2021, and Google’s small business developer program which introduced the 15% rate for revenue less than USD 1 million only commenced in July of that year. So, Professor Asker incorrectly applied the 15% rate to the entirety of the 2021 year, skewing Google’s average fee lower than it was.

801    Let me identify other difficulties with Google’s submissions regarding Steam’s commission rates, as the class applicants have identified.

802    First, Google says that it makes no sense to focus only on Steam’s lowest rate offered to the small minority of developers with revenues over USD 50 million. But this criticism reflects Google’s failure to grasp the concept of an effective commission rate.

803    It is the revenue earned and commissions paid by developers that is the relevant integer for the purpose of assessing Steam’s effective commission rate. Large developers earning significant revenues are highly relevant to effective commission rates. This is apparent from both Apple’s and Google’s effective commission rates during the relevant period, which were between 29.1% and 26.6%, in the case of Apple, and 29.8% and 28.5%, in the case of Google, notwithstanding that a significant number of their developers are eligible to pay a 15% commission.

804    Second, Google alleged an error in Mr Holt’s reasoning regarding Steam’s headline commission rates that it says was exposed during cross-examination. Google says that Mr Holt incorrectly believed that Steam charged a 20% flat commission for developers with revenues above USD 50 million, when in fact Steam charges a 30% commission to all developers on the first USD 10 million of revenues earned on Steam, a 25% commission on revenues between USD 10 million and USD 25 million, and a 20% commission on revenues above USD 50 million. But the point goes nowhere.

805    Google cross-examined Mr Holt on a part of his reply report, but Mr Holt’s primary expert report regarding Steam discussed the progressive nature of Steam’s headline commission rates. Steam lowered its prices for large developers in November 2018. It retained the 30% commission on all revenue under USD 10 million, but reduced it to 25% for revenue between USD 10 million and USD 50 million and to 20% for revenue over USD 50 million.

806    Further, the substance of Mr Holt’s evidence was that Steam offers a lower headline commission rate, being 20%, to developers of titles for which Steam faces the most competitive pressure on commission rates, which are also the developers with the greatest ability and incentive to distribute directly, and the developers that have the greatest bargaining power. That evidence is unremarkable.

807    In any event, Mr Holt said that while Steam’s headline commission rates are relevant to his assessment of the counterfactual commission, he considered them alongside the overall market evidence in the round, including negotiations with developers that go beyond headline rates and the prevalence of direct distribution.

808    Mr Holt said that Steam’s rates are one indication of how competition might arise. It is only one of the many dimensions over which he would expect competition to take effect.

809    Third, Google relies on Professor Asker’s evidence to support its claim that the Play Store’s commission rates appear to be lower than Steam’s for the vast majority of developers who pay a commission to Google. Google says that Professor Asker estimated that 99.6% of developers who pay a commission to Google would have a lower commission on the Play Store than on Steam, and found that there would be a 1% difference between the average commission rate that developers paid to Google versus what they would have paid to Steam.

810    But as Google acknowledged, Professor Asker derived his estimate by applying Steam’s headline commission rates to the global revenues earned by US-based developers that publish on the Australian Play Store and earned revenues via the Play Store in 2021.

811    Accordingly, Professor Asker’s evidence said nothing about the commission rates that developers in fact pay to Steam or Steam’s effective commission rate.

812    It also said nothing about the position of developers not included in his sample of US-based developers.

813    Further, Professor Asker’s analysis was limited to the year 2021. It therefore also said nothing about the first 3 years of the relevant period, including the period before 1 July 2020, when Google reduced its service fee from 30% to 15% for the first USD 1 million earned each year by each developer.

814    Google’s suggestion that Professor Asker’s analysis somehow means that 99.6% of group members would get zero in these proceedings is wrong.

815    Fourth, Google says that Steam charges a 30% commission rate to all developers on the first USD 10 million of revenues they earn on Steam. That assertion is unsafe for the reasons discussed above.

816    For these reasons, Professor Asker’s calculations are unreliable and must be disregarded, along with any contention that Google’s average fee at any time during the relevant period was equal to or lower than that of Steam.

817    Let me move to the China context.

Professor Asker’s approach – is China relevant?

818    Professor Asker said that the counterfactual commission can be determined by benchmarking against the commission rate charged by Android app stores in a different region, China, where the contravening conduct has not been present. Indeed, in the concurrent session, and consistently with the case opened by Google, Professor Asker went so far as to suggest that Google app stores in China were a more useful comparator than the PC app distribution market.

819    Now I agree with the class applicants that there were several difficulties with Professor Asker’s use of the commission rates in China; indeed, of all the impressive evidence that Professor Asker gave before me on many topics, his use of Chinese rates was a low point.

820    Professor Asker’s benchmarking against commission rates in China assumed that the Chinese market for Android app distribution is comparable to that in Australia in all relevant respects save for the absence of the contravening conduct. But there are aspects of the Chinese market for Android app distribution which are idiosyncratic to China and which render that analysis of doubtful probative value to say the least, as the class applicants explained.

821    As Google itself recognises, Android in China is a parallel universe. As a consequence of Google’s absence, app distribution services in China are primarily provided by OEM device manufacturers that operate and pre-install their own app stores and actively divert users to those stores, such as Xiaomi and Huawei. OEMs also impose technical restrictions on the download of third-party app stores, restricting the capacity of those third-party stores to compete. This conduct is analogous to at least some of the contravening conduct.

822    This is a significant difference between the Chinese Android market and the counterfactual in Australia. If the Play Store was present in China, OEMs would have an incentive to ensure their systems were compatible with the Play Store. The absence of the Play Store from the Chinese Android market is clearly relevant to Professor Asker’s China analysis.

823    Professor Asker’s benchmarking against commission rates in China assumed that the Chinese market for Android app distribution would have been comparable to that in Australia in the counterfactual. Professor Asker said that the Chinese market is a reliable comparator because it is competitive. But that proposition has not been established on the evidence. Government intervention in the Chinese market suggests that it is highly unlikely that it is a competitive market.

824    Indeed, in response to concerns I raised that the Chinese comparison was contaminated by government regulation or government interference, Professor Asker conceded that he did not dispute at all that the Chinese government has an interest in all sorts of mobile architecture.

825    Despite this, Professor Asker accepted that he had not considered in his reports various aspects of government interference and regulation in the Chinese market, including the requirement for app stores to be licensed.

826    The regulatory framework faced by technology companies in China differs from that in Australia, including by reason of the so-called Great Firewall, which restricts users’ access to content, and regulatory measures which limit foreign companies’ access to the Chinese market in favour of local entities.

827    Similarly, the cost structure of operating an app store, including due to licensing and censorship requirements imposed by the Chinese government, differs from that in other jurisdictions, including Australia.

828    Moreover, there is evidence of relatively weak enforcement of intellectual property rights in China, which further reduces incentives for app stores to compete for developers in circumstances where apps can be scraped or cloned from other stores.

829    When this was put to Professor Asker in cross examination, he admitted to being aware of scraping in the Chinese market as it was mentioned in a presentation on China referred to extensively in his evidence. Moreover, he accepted that it would be implausible to assume that, in the counterfactual, app stores would copy, without developers’ authorisation, apps from other app stores. Yet, by drawing an analogy with China without accounting for the effect of scraping and cloning in that market, that is the implied assumption that Professor Asker made.

830    The realistic possibility that these features of the Chinese market have a causal relationship with the commission rate charged in that jurisdiction renders his China benchmarking analysis unreliable.

831    Further, Professor Asker’s analysis failed to grapple with the distortionary effect of Play Store’s absence from the Chinese market. The Play Store and other GMS apps have been blocked form China since 2010.

832    In all the circumstances, Professor Asker’s suggestion that based on Android in China, the counterfactual world Mr Holt proposed would likely be associated with higher service fees, rather than lower service fees, is not made good.

Summary

833    The difficulties with the benchmarking approaches adopted by Professor Asker lend themselves to the conclusion that they are not reliable measures by which to calculate the commission in the counterfactual. Contrastingly, the method put forward by Mr Holt is the most reliable means of measuring the counterfactual commission rate.

834    Generally speaking and notwithstanding its imprecision, the comparator analysis performed by Mr Holt does tend to have some merit.

The counterfactual commission — Google context

835    The class applicants say that absent the contravening conduct, the counterfactual commission would be in the range of 10% to 20%.

836    They say that adopting the conservative assumption that the size of the Android mobile app distribution market would remain the same following new entry absent the contravening conduct, the Play Store would remain profitable charging a commission of more than 10% and with a market share as low as 25%.

837    Whilst Google imposed different commission rates for different types of transactions at different points during the relevant period, Mr Holt calculated the effective commission rate for paid apps worldwide as ranging between 26.3% to 30% and for in-app purchases in Australia as ranging between 25.9% to 29.9%.

838    The difference between the effective commissions paid by Dark Ice and Android and app developer group members during the relevant period and the counterfactual commissions represents the extent of the overcharge obtained by Google by reason of the contravening conduct, being commissions paid by the second applicant and Android app developer group members, which would not otherwise have been paid by them.

839    The class applicants say that in all of the circumstances, a competitive level of commission in a competitive Android mobile app distribution market would be in the range of 10% to 20%, with a midpoint of 15%, being the counterfactual commission. Google is not alleging that the counterfactual commission rate during the relevant period should be reduced by reason that the contravening conduct occurred prior to the relevant period.

840    They say that the extent of the overcharge imposed by Google is the difference between the counterfactual commission and its effective commission during the relevant period.

841    The overcharge imposed by Google on Android app developers equates to the difference between the effective commission rate imposed by Google and the rate that Android app developers would have paid, but for the contravening conduct.

842    Now Google criticises Mr Holt’s evidence for failing to provide a basis to estimate the counterfactual commission having regard to only a subset of Google’s contravening conduct. I tend to agree with Google on this aspect and will need to hear further from counsel. At this stage I am not in a position to properly analyse Mr Holt’s 10% to 20% range. His analysis will need to be modified to address the actual contravening conduct of Google that I have found, and the significance of its absence in any applicable counterfactual scenario.

843    But notwithstanding, let me briefly touch on a few points.

The Play Store commission in the actual is not competitive

844    Now Professor Asker suggests that the Play Store commission in the actual is competitive on the basis that, while Google has made UCB and DOB available in some jurisdictions, there has been little take-up of these options.

845    Professor Asker asserts on this basis that the Play Store is offering a competitively priced, high-quality billing solution and the removal of the contravening conduct would have no positive effect for Android developers. But I agree with the class applicants that the argument is flawed.

846    As Mr Holt observed, to the extent that the rebate offered by Google through UCB and DOB is lower than the cost to developers to employ alternative payment solutions providers, the total cost to developers of using those alternatives will be higher than if the developer continues to use Google Play Billing.

847    Indeed, Google’s own internal modelling shows that the cost to Google, let alone third-party providers, of providing Google Play Billing ranges from 3.7% to 6.1%. Some modelling suggests costs as high as 7.5%.

848    As the class applicants say, and as I have indicated elsewhere, it is likely that the cost to developers of using third-party providers will be higher still, since there will be additional costs associated with integration of third-party billing providers into the Play Store and app infrastructure.

Comparator analysis — payment solutions

849    Now as the class applicants have said, a number of other alternative payment solutions providers already offer payment solutions that have equivalent functionality to Google Play Billing, and, but for the contravening conduct, would have presented an alternative option to Google Play Billing.

850    Mr Holt calculated a hypothetical effective commission rate from the fee structures of the alternative payment solution providers, applied to the average value of paid transactions on the Play Store during the relevant period, of between 2.2% and 9.62%, with most providers sitting between 4% and 5.5%.

851    Mr Holt concluded that, in his opinion, the portion of the commission charged for purchases made via the Play Store and attributable to its payment solution would have been in the range of 2.2% to 10% of the relevant purchase, with the central scenario likely to be 4% to 5%.

852    I am not in a position at this stage to properly address this.

Is the counterfactual commission calculated by Mr Holt arbitrary?

853    For reasons similar to what I said in the Apple context, Mr Holt did not take an arbitrary approach at the level warranting Google’s criticisms. Of course any approach taken in one sense would always have an element of arbitrariness associated with it.

Is there a single answer concerning overcharge that applies to all apps?

854    Google has made the following observations with which I agree.

855    It is unclear whether the conduct that I have found affected all types of apps in the same way, including all apps developed by developer group members, such that it can be concluded not only that overcharging in fact occurred, but that it occurred in respect of all apps, and thereby impacted upon all types of developers and app users.

856    And it is unclear whether this effect was the same across all apps, so that it is appropriate for me to arrive at a single answer in relation to overcharge which applies equally to all types of apps, developers and app users. The commissions payable by developers vary depending on their circumstances and have varied over time. I will need to hear further from counsel.

Summary

857    I do not propose to deal further with the counterfactual commission and overcharge questions at this time. I will need to hear further from counsel. These questions and their resolution need to be more focused and to take into account the actual contraventions that I have found against Google.

Conclusion

858    In both class actions in my view the commissions payable by developers would likely have been materially less absent the contravening conduct of Apple and Google that I have found, that is, in any applicable counterfactual scenario.

859    But at this stage I am not in a position to quantify the difference between the commissions actually paid and the commissions that would have been paid in any applicable counterfactual scenario.

860    As I say I will need to hear further from counsel.

Associate:

Dated:    12 August 2025