How platform fee changes flow into bids

Developers are winning fee concessions faster than at any point since the app stores opened. Apple swapped its EU per-install fee for a 5 percent Core Technology Commission in January. Google's proposed Play settlement, which this site covered in March under "Google Play Fees Cut to 20 Percent: The Epic Deal Lands", promised lower commissions if Judge Donato approves it, which as of this month he has not. The Ninth Circuit sent the Epic v Apple link-out fee question back to the district court in April. Each concession is celebrated as money returned to developers.

Most of it will not stay there. In a market where the majority of performance spend runs through automated ROAS bidding, a fee cut is a bid increase with a delay measured in weeks. The developer gets a better margin for one reporting cycle, and then the ad auction takes its share.

The mechanism is arithmetic, not strategy

A target-ROAS bidder works from a simple identity. It predicts revenue per install, applies the advertiser's target, and derives the most it can pay. The revenue it predicts is net revenue, because that is what the advertiser's MMP or SDK reports back after the store has taken its cut.

Change the store fee and the net revenue per install changes in the reported data. The bidder's model updates, its predicted value per user rises, and its bids rise with it. No one in the studio makes a decision. No one in the network changes a setting. The cut appears in postbacks and the auction absorbs it.

As an illustrative example only: a game earning a hundred units of gross revenue per hundred installs at a 30 percent commission reports seventy units net. Cut the commission to 20 percent and it reports eighty. That is a 14 percent rise in net value per install. A bidder holding a fixed ROAS target now has roughly 14 percent more headroom on every bid. If the studio's competitors in the same auction enjoy the same fee cut, all of their bidders gain the same headroom at the same time, and the clearing price in a competitive auction moves toward the new ceiling.

Who ends up holding the money

The flow is worth tracing to the end because the winners are not the obvious ones.

The store loses fee revenue, by design. The developer's margin improves for as long as the auction takes to re-clear, which in a genre dominated by automated bidding is a few weeks of learning and a few more of competitors' models catching up. The ad platform gains, because higher bids at the same supply mean higher CPMs. AppLovin's first quarter results on May 6 showed revenue up close to 59 percent year on year; whatever else drives that, an industry handing its bidders more headroom does not hurt.

The speed depends on how the fee change reaches the data. A commission change applied at the store flows into every SDK-reported purchase from day one, so bidders see it immediately. A change that only applies to a web shop or a link-out path, as the Epic v Apple remand may eventually produce for US link-outs, reaches the bidder only through the share of revenue routed that way, and the re-clearing is slower and partial. Buyers should expect the two cases to behave differently and not assume one fee headline means one market response.

There is a distributional effect within the developer side too. Studios that bid to a CPI target rather than a ROAS target do not raise their bids when fees fall. In an auction where everyone else did, they lose share. Studios with a large organic base convert the fee saving into pure margin because they buy less of their volume. Studios that are mostly paid see it competed away almost entirely.

The second-order effect on measurement

A fee change also corrupts the comparison a buyer most relies on. Cohort ROAS before and after the change is not like for like: the same players, buying the same items, now produce a different net revenue figure. A team that does not adjust its historical baseline will read the post-change period as an improvement in cohort quality, and may credit a creative or a targeting change that did nothing.

The fix is dull and necessary. Restate historical net revenue at the new fee rate when comparing cohorts across the change date, or compare on gross revenue and apply the fee as a separate line. Web shop revenue, which carries its own payment costs rather than a store commission, needs the same treatment or the blended figure drifts.

A rule for the week a fee changes

The decision rule is to revisit bidding targets before the algorithm does. Specifically:

• On the day a fee change takes effect in reporting, decide deliberately whether the extra net margin should become bid headroom, be retained as margin, or be split.
• If retaining margin, raise the ROAS target in proportion to the fee change so the bidder's headroom is unchanged.
• If chasing share, hold the target and accept that the gain will be temporary as competitors follow.
• Tell the creative team, because a bidder with more headroom will spend faster against the current winning concepts and bring fatigue forward.
• Set a review at six weeks to check whether CPMs in your core auctions have moved by roughly the fee delta. If they have, the market has re-cleared and the question becomes whether your share is where you want it.

The fee negotiations working through the courts are real money. The question of who keeps it is settled not in San Francisco or Brussels but in the bid landscape of your top three networks, by software that never reads the judgment.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Alternative app stores will not lower your CPI
• Ad Platform Contract Renewal: 2026 Terms to Check First