Ad tech integration hurts mid-sized studios most

Vertical integration in mobile ad tech is usually debated as a question about the biggest players: whether AppLovin's Axon and Unity's Vector, each sitting on top of a mediation layer plus an SDK footprint and a large publisher base, have too much of the market. That debate points the camera at the wrong end of the table. The studio that pays the most for integration is not the one with a billion-dollar franchise, nor the two-person team shipping its first hybrid-casual title. It is the studio in between, with somewhere between thirty and three hundred people, a few live games, and a UA budget large enough to matter and too small to command terms.

The reason is structural, and sharper than "the middle has no bargaining power". The mid-sized studio's post-install outcomes are exactly the data an integrated model needs to learn what a paying user in its genre looks like. The studio supplies that data free through an SDK it cannot remove, and the model uses what it learned to price and sell the same users to a better-funded competitor in the same genre. The mid-sized studio is not merely a customer of the stack. It is an unpaid supplier to its own competition.

What integration actually sells

Strip the marketing off and an integrated ad platform bundles four things: demand (its own network), supply access (its SDK in thousands of games), allocation (mediation that decides which bidder wins in those games), and a view of outcomes (post-install events flowing back through the same SDK). Together they give the operator a near-complete picture of user value that no individual buyer or seller on the platform has.

The platform then prices against that picture. On the buy side, the automated bidder pays what it predicts a user is worth minus its margin, and the advertiser cannot audit the prediction. On the sell side, mediation allocates impressions under rules the publisher cannot fully observe. Unity's February disclosure that Vector had reached 56 percent of Grow revenue, followed by the decision to close the ironSource network on April 30 and put everything behind the model, said where Unity thinks the value sits: in the model that sees the most, not in the marketplace anyone can inspect.

Why the middle pays most

The very large studio has first-party data at scale, direct web stores, an in-house data science team, and the ability to walk away from any single network. It negotiates data-sharing terms and runs incrementality tests rigorous enough to call a vendor's bluff. Integration is an inconvenience, not a tax.

The very small studio takes every default. It has no bargaining position and nothing to lose by giving it up. The stack's automation replaces a UA team it could not afford, and its data is too thin to teach the model much, so it isn't really subsidising anyone.

The mid-sized studio is the awkward case. It spends enough that a two-point swing in effective margin is material, and generates enough post-install data across a few titles to teach the model a great deal about its genre. But it cannot self-build measurement, cannot fund incrementality on every channel, and cannot credibly threaten to leave the mediation layer that also monetises its games. It pays the margin on the buy side and feeds the model on the outcome side, and the model does not distinguish between advertisers when it decides what a match-3 payer looks like.

How to tell it is happening to you

The leakage is invisible in any single dashboard, but it leaves a signature across two or three.

The first sign is CPM inflation on your best-performing genre segments shortly after a well-funded competitor launches or scales in the same category. Your target has not changed and your creative has not fatigued, but the clearing price in auctions you used to win rises, because another bidder on the same platform now values those users highly, and its valuation drew partly on your outcomes.

The second sign is that the rise is uneven. Segments where your game is unusual, and the model had only you to learn from, hold steady. Segments where you share an audience with the new entrant move first and most. Cut CPM trends by country and by the placements that produced your top payers and the pattern is usually visible within weeks of the scale-up.

The third sign is a widening gap between your MMP-reported ROAS and the network's modelled ROAS on the same spend, in the network's favour. The model is more confident about your users than you are, and it has no reason to be unless it learned from a broader pool than your campaigns.

None of these proves the mechanism alone. Together they are a reasonable basis for a hard conversation at renewal.

The second-order effect is consolidation

The trade-off most coverage misses is what mid-sized studios do about this. Many sell. PocketGamer.biz and Aream counted Q1 2026 games M&A at $7.7 billion across 52 deals, more than triple a year earlier, and the pattern in the mid-market deals is a small profitable team joining a group that already has negotiating scale with the platforms. Scopely's majority stake in Loom Games in February fits that shape.

That is rational for each seller and bad for the market. Every mid-sized studio absorbed into a larger group removes one independent buyer from the auction and adds its data to a pool that negotiates as a bloc. Fewer and larger advertisers face fewer and larger platforms, and the room to grow from small to large on a studio's own UA economics narrows. Liftoff's refiled S-1 in April and the reported Moloco IPO preparations are, read this way, the independent DSP tier raising capital because a two-platform market leaves room for a third only if it can afford its own model.

Three tests, and what to do with the result

Three questions place a studio.

Can you run a clean geo holdout on your largest integrated channel and act on the result? If not, you cannot audit the prediction you are paying for.

Does your mediation provider also run a network that wins a large share of your impressions? If so, the same party sets your sell-side allocation and your buy-side price.

Would any one platform change its terms if you paused spend for a month? If not, you've no standing to price your data.

Two or three unfavourable answers put a studio in the middle. The response is not to leave the integrated stacks, which is neither possible nor sensible, but to stop feeding them for free. Negotiate data-sharing scope explicitly at renewal rather than accepting SDK defaults, and put the CPM signature above on the table. Route a fixed share of spend, as an illustrative example a fifth, to channels whose model does not also monetise your games, so there is always an external price to compare against. Review the mediation decision annually as a strategic choice, not a technical one set at launch. And keep an MMP view of outcomes independent of any network's SDK, even where it costs more, because it is the only outcome record the stack does not also own.

The studios that survive as independents through this cycle will be the ones that behave, at their size, as the large ones do on terms: a counterparty with something to withhold, not a data source that happens to also buy media.

Related archive reading

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

• Negotiating with ad networks and vendors
• The consolidation endgame for mobile ad tech