The Mobile Publishing Machine: why the studio builds the game and the publisher keeps the money

I have spent my time in mobile growth, ran mobile user acquisition for 5 games at time. One person, every channel, every dashboard. So when one of my friend ask me why studios sign their game away to a publisher for half the revenue or less, I do not give the polite answer. I give the real one.

Conceptual illustration of a mobile game publishing machine turning app installs into money, showing the user acquisition arbitrage model.

The game is not the product. The arbitrage is.

Start from first principles

A publisher like Voodoo, Homa or Supersonic is not in the business of making good games. They run a user acquisition arbitrage engine. They buy an install for (X) from Meta, TikTok or an ad network, and squeeze (Y) in lifetime value out of that user through ads and in app purchases. If (Y) beats (X) reliably and at scale, they win. The game is just the funnel that makes the trade work.

Once you see this, everything about the revenue split stops feeling unfair and starts feeling like exactly what it is: a financing structure.

How we got here

2017 to 2020 was the hypercasual gold rush. Dead simple games, near zero retention, monetized almost entirely on ads. The whole model ran on one thing: cheap, granular, deterministic targeting on Facebook. You could find a buyer for any user because the platform knew everything.

Then April 2021. Apple shipped App Tracking Transparency on iOS 14.5 and the deterministic identifier went dark for anyone who tapped "Ask App Not To Track." Overnight, the targeting that hypercasual was built on stopped working. CPIs climbed. Pure ad monetization stopped clearing the math.

Two things happened in response, and they define the industry you see today.

Thing one: SKAN, and why measurement got brutal

Data points going dark into fog, illustrating Apple ATT and SKAdNetwork privacy changes that broke mobile user acquisition targeting.

Apple replaced user level attribution with SKAdNetwork, a privacy first framework that tells you a campaign worked without telling you who it worked on. SKAN 4 gives you up to three time staggered postbacks across a 35 day window: days 0 to 2, 3 to 7, 8 to 35. You get coarse conversion values, delayed timers, and privacy thresholds that null out your data when the cohort is too small to anonymize.

Read that again. You optimize on aggregated, delayed, partially censored signal. As of early 2025 SKAN drove over 40 percent of all iOS attribution. Apple is now moving everyone toward AdAttributionKit, introduced in iOS 18.4, which finally adds re engagement windows and configurable attribution intervals that SKAN never had. But the core reality holds: on iOS you are flying with instruments, not eyes.

This is the part outsiders miss. Privacy did not just hurt targeting. It raised the skill floor so high that a solo developer cannot model LTV well enough to bid confidently. Publishers can, because they see thousands of cohorts at once. Measurement scale became a moat.

Thing two: creative became the algorithm

Hundreds of ad creatives flowing into a machine with a few winners, showing why creative volume drives mobile UA on Meta and TikTok.

When platforms lost your targeting signal, they automated everything and handed it to machine learning. Advantage+ on Meta, Smart+ on TikTok, App Campaigns on Google. You no longer set the audience or the bid. The platform throws your ad at a broad pool and lets the creative do the sorting. The creative is the targeting now.

Which leaves exactly one lever in your hands: how many winning creatives you can produce per week, per channel.

Here is the number that ends the debate. Same campaign, same audience, same bid. Creative A lands 15 installs per 1000 impressions, Creative B lands 4. That is a CPI of roughly 0.80 versus 3.00. No bid tweak closes a 4x gap. Only creative does. Meta even rebuilt its retrieval engine, Andromeda, to handle tens of thousands of ads at once, because feeding the machine more variants is the optimization.

So the publisher edge is not "they know more channels." The edge is throughput. A studio shop pushing hundreds of fresh creatives a week into a system that burns through winners in days. A solo cannot match that volume across Meta, TikTok and a dozen networks at once. The ceiling is what makes a game studio go bankrupt.

The economics, in actual numbers

Coins cascading through filters into a small share at the bottom, illustrating the mobile game publishing revenue split and recoupment waterfall.

Hypercasual today: ARPDAU around 0.03 to 0.08, D7 retention 6-9%. At US CPIs of 2.00 to 3.50, that math no longer clears.

So the survivors became hybrid casual. Same simple hook, but bolted onto a meta layer, an in app purchase economy and a season pass. ARPDAU jumps to 0.15 to 0.50, four to seven times higher. D7 retention 18 to 22 percent, the threshold that makes paid UA mathematically survivable again. Hybrid casual in app revenue roughly doubled year over year into 2025. Roughly a third of studios are mid transition from hyper to hybrid right now. The ones still clinging to pure ad hypercasual are dying quietly.

The split, and the trap inside it

Here is the part every studio should tattoo on their hand. The headline percentage is the least important number in the deal.

The waterfall is what matters:

Gross revenue comes in. Apple or Google take 30% off the top first, before anyone else touches it. Ad network and mediation fees come out. What remains is "net." Then, in many deals, the publisher recoups its UA spend, the ad money it burned, before a single dollar is split. Only the profit after all of that gets divided.

So when a publisher advertises a 50% revenue share, ask the only question that counts: 50% of what. Fifty percent of net after the publisher recoups all UA can be worth less than 25% of net with no UA recoupment, because if the title has not cleared its ad spend, half of zero is zero. Homa has run 50/50 in its game jams. Voodoo runs a share that shrinks as they invest more, with studios on long term royalties. Nobody publishes the standard terms, because the terms are the leverage.

The studio is not selling a game. The studio is selling its upside in exchange for someone else carrying the risk of bankruptcy. For a team with no UA capital and no measurement infrastructure, that trade can be rational. For a team that can run its own UA, self publishing keeps the 70 percent after the store cut, and you eat the lottery risk yourself.

Why the house wins by design

Three structural reasons, and none of them are about making better games.

Lottery economics. Under 1% of hypercasual prototypes scale. A studio making five games and missing goes broke. A publisher tests five hundred at once and needs two hits to carry the slate. They are not selling games, they are selling risk management at scale.

UA capital plus data. They have the marketing budget the studio ran out of, plus the cohort data to bid into censored SKAN signal with confidence.

The recycling network. This is the quiet weapon. Casual users get satiated, and once satiated, buying fresh users from Meta stops being profitable. So publishers cross promote: they buy unprofitable games for their user base and funnel those players into profitable titles they own, then repeat. A standalone studio has no network to recycle into, so its LTV is structurally lower than the same user inside a portfolio.

The 2026 reality

The smartest read on this industry is what AppLovin did. In 2025 it sold its entire game studio business to Tripledot for 400 million in cash plus a 20 percent stake, and the CEO said plainly they were never game developers at heart. They kept the ad tech and the data. The richest player in the room sold the studios to go all in on the layer that actually compounds.

That is the whole lesson. The value was never in making the content. It was in owning the distribution, the UA capital and the data.

So if you are about to enter this, the question is not "is my game good." The question is which side of the arbitrage you are on. The side that owns the machine, or the side that supplies it raw material. That answer decides who eats.

Sources: SKAN and AdAttributionKit 2026, RocketShip HQ, Hybrid casual analysis Dec 2025, NextBigGames, Voodoo Secret Sauce, Deconstructor of Fun, Tripledot acquires AppLovin studios, MobileGamer.biz.