Alternative app stores will not lower your CPI

A studio head asked us last month whether a move to alternative Android stores would bring CPI down. The answer is no. The reasoning matters more than the answer does, because the same misunderstanding is about to drive a lot of planning decisions as the Epic v Google injunction keeps reshaping Play.

CPI is a price for attention, set in an ad auction. The commission a store takes is a price for distribution, set by the store. Two different markets. An alternative store changes the second one and leaves the first alone, and in the near term it pushes effective CPI up.

Where the price of an install is actually set

When a network shows a rewarded video for a puzzle game, the winning bid reflects what the network predicts that user is worth to the highest bidder, less the network's cut. Nothing in that calculation references which store the install button opens. Meta, Google's ad side, AppLovin, Moloco: every one of them prices the user, not the checkout.

One indirect route exists. A store choice can touch the auction through the conversion signal, and if the alternative path reports fewer attributed installs or reports them later, the bidder sees a weaker campaign and pays less per impression while also delivering fewer installs. On a dashboard that reads as a lower CPM and a higher CPI at the same time, which is the opposite of the outcome the plan assumed.

So a store with a lower commission does not make the impression cheaper. What it does is raise the studio's net revenue per install. The store fee is a downstream margin event, and confusing the two leads teams to put "alt store CPI savings" into a plan where the real line should read "alt store net revenue uplift, minus friction".

Friction is a CPI increase in disguise

Every additional step between an ad click and a first session is a leak in the funnel, and every leak raises the cost of the installs that survive it. Third-party stores on Android still require the user to install the store and grant permissions, and in some flows to confirm a developer identity. Each of those is a place where a share of users abandon.

Buyers already know what this does to the arithmetic. Take an illustrative example: a campaign delivers a click to a Play listing at a given cost and converts a third of those clicks into installs, while the alternative-store path converts a quarter. That shows an effective CPI a third higher on identical media. The auction did not change; the store did.

Attribution adds a second leak. This site laid out the setup in "Android Alternative App Stores and Attribution Setup" last summer, and the practical position has not moved much: MMP referrer support for non-Play installs is uneven, and automated bidders read an unattributed install as no install at all. A bidder that cannot see conversions bids less, and less bidding means less scale at the same price rather than a lower price.

Why the saving gets competed away

Now the effect that most of the coverage misses. Suppose somebody solves the friction problem and a meaningful share of a genre's installs move to lower-commission stores. Every studio in that genre now has more net revenue per install, and most of them run ROAS-targeted bidding. A tROAS bidder that sees higher net revenue per install raises its bids mechanically, without a human deciding anything.

Once enough of the category's spend sits on such bidders, the commission saving turns into bid headroom, and the bid headroom turns into higher CPMs. The store gave up margin, the ad platform picked it up, and the studio's blended economics settle somewhere between where they started and where the optimistic model said they would land. Fee cuts on Play run on the same dynamic, if Judge Donato ever approves a version of the settlement he doubted in April: a windfall to ad networks as much as to developers.

The studios that keep the saving are the ones whose competitors cannot follow. Strong organic share does it. So does sitting in a genre where a few very large spenders hold the top of the auction, spenders who already run direct web stores and price their bids off different economics.

A rule for deciding whether it is worth it

Model the move as three separate lines, with explicit assumptions on each.

Net revenue uplift: the commission delta applied to the share of revenue you expect through the alternative path, not to all revenue.

Friction haircut: the expected drop in click-to-install and in attributed installs, expressed as an effective CPI increase.

Re-pricing haircut: the share of the net uplift you expect the auction to absorb over two to three quarters as competitors' bidders adjust, higher in crowded genres.

If the first line still exceeds the sum of the other two, pilot it, and judge the pilot on net revenue per paid install rather than on CPI. If a vendor or an internal deck shows CPI falling because of a store move, something else changed: the creative, the geo mix, the attribution window. Ask for the same period on the Play path before you accept the comparison.

Where alternative stores genuinely help acquisition is reach, not price. Epic's store on Android, Samsung's store, the rivals Google has had to surface under the 2024 injunction: each carries a small audience that is unusually engaged with games and unusually likely to have a payment method on file. That is a segment argument, and segment arguments belong in the ROAS column of the plan, which is where this whole conversation should have started.

Related archive reading

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

• If app stores become optional, what happens to UA
• The auction, not the audience, sets your CPI