The CMA’s Steering Fee Framework for Apple and Google: Cost, Value and a Nod to the CAT

On 30 June 2026 the Competition and Markets Authority (CMA) published proposed Steering conduct requirements (CRs) for Apple’s and Google’s mobile platforms under the Digital Markets, Competition and Consumers Act 2024 (DMCCA). For those following the mobile investigations, the choice of instrument is itself worth pausing on. In mobile, the CMA had so far preferred the voluntary commitments route, having secured promises from both firms in February 2026 and final commitments from Google and Apple on 1 April 2026. The move to a formal CR on steering signals that this issue was always going to be treated differently, because steering goes to the heart of the platforms’ revenue model in a way that a light-touch undertaking was never likely to resolve.

Much of the early commentary has focused, understandably, on the mechanics of steering, namely link design, the interstitial screens a user meets when moving out of an app, and the side-by-side presentation of payment options. The part of both decisions that interests readers is the Steering Fee framework, and in particular the way the CMA draws on the Competition Appeal Tribunal’s (CAT) judgment in Kent v Apple [2025] CAT 67 to support its provisional reasoning on fees. This appears to be the first time a collective proceedings judgment has been used to inform a DMCCA decision, and it deserves a closer look. It also invites a dialogue with the recent discussion of collective redress on this blog.

Why steering, and why a fee at all

Anti-steering restrictions are the rules through which a platform operator prevents app developers from telling their own customers about cheaper ways to buy digital goods and services outside the platform’s payment system. In practical terms a developer has been unable to place a link inside an app pointing the user to a lower price on the developer’s own website (paragraphs 1.5-1.6 of the Google and Apple proposed CRs). The treatment of steering provisions under the DMA has been analysed in this blog previously. Apple has banned steering in the United Kingdom. Google has restricted it, and on 24 June 2026 announced revised Play Store terms that permit steering subject to conditions and introduce new fees.

Removing the prohibition alone would achieve little. The CMA provisionally considers that absent safeguards going beyond the removal of the ban, each firm would retain the ability and incentive to impose contractual, technical, design or commercial restrictions that would materially reduce the viability and attractiveness of steering (paragraph 4.2 of each decision). The design of the proposed CR therefore seeks to ensure that developers are not only permitted to steer but are able to do so effectively in practice, without undue friction, deterrence, discrimination or commercially unworkable terms. This is the reasoning that carries the whole exercise, and it explains why the CMA regulates the practical conditions of steering rather than merely its existence.

The fee is the pivot. A platform that must permit steering, yet remains free to charge whatever it likes for a steered transaction, can price the alternative channel back out of existence. So the CR provides for a Steering Fee (paragraph 9 in Section 3 of the both proposed CRs), and the interesting question becomes how that fee is to be set. Here the CMA has chosen a principles-based framework rather than a fixed number (Apple decision, paragraphs 4.77 to 4.80; Google decision, paragraphs 4.78 to 4.81), resting on three principles set out in paragraph 10 of the proposed CR: 1) a cost-based principle, 2) a value-based principle, and 3) an administrative simplicity principle.

The cost-based principle, and the Kent v Apple as corroboration

The cost-based principle (paragraph 10(a) in Chapter 3 of both proposed CRs) would allow a platform to recover the costs it genuinely incurs, including a reasonable return on investment, rather than a share of the developer’s revenue. The CMA provisionally considers that a cost-based fee is likely to be significantly lower than Apple’s current App Store fees, given the low-cost nature of operating a high-volume, platform-based app store and Apple’s high observed profitability.

It is at this point that Kent v Apple enters the reasoning. The CMA states that its profitability analysis appears consistent with the CAT’s findings on Apple’s costs in the context of an excessive pricing evaluation, noting the Tribunal’s finding (see this piece by the author for a detailed breakdown) that there was a significant and persistent difference between the price of App Store services and the cost of providing them (Apple decision, paragraph 4.111, citing Kent, paragraph 610). The CMA does not adopt the CAT’s conclusions as its own and recognises that the judgment is subject to an ongoing appeal. The CMA takes from the CAT corroboration of its own analysis of profitability which was reached independently in an adversarial setting but points in the same direction as the CMA’s.

The second borrowing is arguably more useful. Apple submitted that it does not, in the ordinary course, systematically allocate costs to specific products or services, and so was unable to provide data on the annual costs of operating the App Store in the United Kingdom (Apple decision, paragraph 4.93). The CMA’s answer is that Apple cannot rely on a stated inability to identify costs as a reason to avoid the cost-based principle, and the proposed CR requires Apple to record the relevant data (paragraph 12(d) in Chapter 3 of both proposed CRs). In support, the CMA points to the CAT’s treatment of the very same argument. The CAT declined to accept that complexity in allocating operating expenditure across product lines should prevent an excessive pricing evaluation, observing that to accept such an argument would allow a firm’s own business-model choices to provide an unjustifiable shield against scrutiny by courts and regulators (Apple decision, paragraph 4.116, citing Kent, paragraphs 599 and 602). The CMA thus enlists a judicial finding on method to answer a methodological objection of its own.

The value-based principle

The value-based principle (paragraph 10(b) of both proposed CRs) sits alongside the cost-based one. It would allow a platform to capture a share of the surplus created where its offer to developers is particularly differentiated, valued, or the result of ongoing innovation. This is the principle the platforms prefer because it opens the door to a fee above cost. This principle may draw similarities with the economic value under excessive pricing case law under Article 102 TFEU and Ch. II CA98 (see the Pfizer/Phenytoin saga in the UK for more in this). The CMA accepts that the App Store provides value that a pure cost measure may not capture, through app review, security and privacy standards, standardised interfaces and a trusted download environment which some smaller developers benefit particularly.

The CMA builds two limits into the principle. First, any estimate of value should be adjusted for the effects of Apple’s substantial and entrenched market power, because in a competitive market some of those returns would be competed away (paragraph 13(d) in Chapter 3 of both proposed CRs). Second, the assessment should reflect the two-way exchange of value, since developers also confer value on the platform (paragraph 13(e) in Chapter 3 of both proposed CRs). Without third-party apps, users would value the platform far less, with knock-on effects for device sales. Value therefore has to be netted, not simply asserted.

Kent v Apple reappears here too. The CMA notes that Apple’s current fees may not have been set by reference to benchmarks or value, and states that this is consistent with the CAT’s judgment, again flagged as subject to appeal (Apple decision, paragraph 4.129). The CAT found that Apple had set an arbitrary fee (Kent, paragraph 631(11)) and was not satisfied that, although Apple clearly provides value to developers through the App Store, the level of those benefits could reasonably explain the very high profitability of the App Store (Kent, paragraph 673). The CMA also rejects each of the benchmark comparators Apple proposed, on the basis that an appropriate comparator must provide similar services, be exposed to effective competition, and reflect the two-way value exchange, and that none of Apple’s benchmarks meets all three conditions.

The Google decision proceeds in parallel and reaches for Kent in the same way, noting that Apple and Google operate similar fee structures and that the CAT found the fee set by Apple to be an arbitrary one (Google decision, paragraph 4.28). That the CMA is willing to carry an Apple-specific judgment across into the Google analysis, on the strength of the similarity in fee design, is itself telling. It treats the CAT’s findings as having a reach beyond the specific defendant, at least at the level of the economic logic they express about platform pricing and cost scrutiny. This provide wind in the sails for similar argumentation in Barry Rodger v Alphabet, Or Brook v Alphabet and similar cases to be heard by the CAT.

Administrative simplicity, and how the principles interact

The third principle, administrative simplicity (paragraph 10(c) of both proposed CRs), is directed at practical exercise. It requires that the fee structure and related terms are straightforward to understand and implement, and do not introduce unnecessary technical or administrative burdens for developers. The CMA observes that Apple’s current UK fee structure is relatively simple, a headline rate with a reduced rate for small-business developers, while fee structures introduced in other jurisdictions have been criticised as overly complex. The message to both firms is that complexity introduced under cover of a steering fee will attract scrutiny rather than deference.

As to how the three principles fit together, the CMA proposes that the steering fee should be no higher than the level determined by applying them in combination (paragraph 10 of Google and Apple proposed CRs), and expects a properly defined cost-based and value-based assessment to yield similar outcomes. The firms are left a degree of flexibility in application, provided they can demonstrate that their approach takes proper account of each principle, and one suggested method is to use the cost and value assessments as a cross-check on each other (Interpretative Note 9). The CMA also guards against double counting, by prohibiting the recovery of investment and innovation returns twice over once through the cost-based principle and again through the value-based one (paragraph 11 of both proposed CRs).

The mechanics, briefly

For completeness, the fee framework sits within a wider set of protections. The CRs address the design of steering links and the information screens users encounter (paragraphs 6 to 8 of both proposed CRs), and they include non-discrimination protections so that developers who steer are not disadvantaged in app review, distribution, discoverability, access to platform functionality or interoperability (paragraph 14 of both proposed CRs). Developers who steer should be able to offer lower prices or different terms, and a platform may not require that its own billing option always presents the best available offer, nor make use of its in-app payment system conditional on a developer foregoing steering. These provisions reflect the same insight that animates the fee framework, that a formal right to steer is worth little if the surrounding conditions quietly make it unattractive. These may feature more broadly in the interpretative notes that the CMA may publish to accompany Google and Apple’s CRs.

Two tracks on one substrate

The appearance of Kent in a DMCCA decision is a small thing on the page and a larger thing in principle. Collective proceedings and digital markets regulation have grown up as separate enforcement tracks in the UK. The first is private, damages-focused and retrospective, driven by class representatives and funders. The second is public, remedial and prospective, driven by a regulator wielding bespoke statutory powers. They share a factual substrate, since the same conduct that grounds a damages claim also draws regulatory attention. Yet, they have rarely referred to one another in their formal reasoning.

What makes the cross-reference more than a citation is the tension it sits within. The CMA is clear elsewhere in the Apple decision (Apple decision, paragraph 4.82) and Google decision (Google decision, paragraph 4.83) that it does not regard itself as bound by the ex post competition law framework, and its fair and reasonable standard is a bespoke regulatory construct rather than an application of the abuse doctrine the CAT applied in Kent. And yet, when it comes to interrogating cost and value, the CMA reaches for the CAT’s factual findings as support. The two tracks remain distinct in purpose and in legal test, but they are beginning to lend each other authority. A finding won in the CAT becoming available to the CMA, and a conduct requirement in turn furnishing material that a future claimant might build upon.

That borrowing is productive, but it is not costless. A finding reached to assess an abuse and quantify loss is not automatically transferable to a regulatory question under the DMCCA about the level of a steering fee which is decided based on a different test. The CMA’s careful framing, that its analysis appears consistent with the CAT’s findings rather than governed by them, and its repeated flagging of the pending appeal, suggest it is alert to this. The discipline will lie in continuing to respect the different purposes for which the CMA and CAT make their findings, even as each draws on the other.

Conclusion

The Steering Fee framework is the analytical core of both proposed CRs, and it is a careful piece of work. A principles-based approach anchored in cost, value and administrative simplicity is more defensible and more durable than a fixed figure, and the CMA has been candid that it expects steering fees to come in well below current App Store charges. The use of Kent v Apple gives that framework an external anchor, and marks the first point at which a collective proceedings judgment has informed a DMCCA decision. Both consultations close on 28 July 2026. How the CMA handles the Kent material in its final decisions, and what the Court of Appeal makes of the judgment in the meantime, will be worth watching closely.

AI Disclosure: An AI tool assisted with research and language. The analysis and argumentation are the author’s own.

Dr Anush Ganesh, Lecturer, School of Law – University of Leeds; Affiliate, Shaping Competition in the Digital Age (SCiDA) Project