First-party data is overrated for most studios

For a studio with one or two live games, no login relationship with its players and no channel to reach them outside the app, first-party data is a cost centre with a strategy attached. The advice to "own your data" has become the default recommendation of every vendor deck since 2021, and for most of the studios receiving it the honest response is that they have nothing to do with the data once they own it.

Measurement leads will bristle at this. The direction of travel is unmistakable: less signal from the platforms, more value in what you can observe yourself, and yesterday's reports in the business press that AppsFlyer has raised over a billion dollars at a 2.7 billion valuation, with four strategic minority holders in Google, Meta, Unity, Moloco, say the market believes first-party pipes are worth a great deal. It does. To the companies that aggregate across thousands of apps. The question is whether the same asset is worth much to a studio that aggregates across two.

Data is worth what you can do with it

The value of first-party data is not intrinsic. It is entirely a function of the actions it enables, and there are only four that matter for growth: reach a player again, model who your best players are, feed that model to a buying platform, and move players between your own titles. Each requires a precondition most studios don't meet.

Reaching a player again requires an identifier you control and a channel that delivers to it. For a mobile game without an account system that is, at best, a push token with an opt-in rate that is falling and, on iOS, a retargeting pool constrained by ATT. Roughly six in ten iOS users still decline tracking, on Adjust's first-quarter figures; the reachable share of your iOS base is the minority that consented, and the platforms already know who they are.

Modelling your best players requires enough of them to model. A studio with a few thousand monthly payers is not training anything on its own data that a network's cross-app model has not already inferred more precisely from a hundred million devices. You can describe your payers. You can't out-predict a platform that watched them pay in twelve other games first.

Feeding a model to a buying platform requires the platform to accept and act on it. Self-attributing networks accept seed lists and conversion signals, and they use them, but the advantage is theirs: your signal refines their model, and their model is the asset.

Moving players between titles requires titles. This is the one case where first-party data unambiguously pays, and it belongs to portfolio publishers. A studio with a single hit has no destination for a cross-promotion.

The costs that do not appear in the pitch

The data warehouse recurs. So does the pipeline maintenance, the analyst who keeps the schema honest, and the privacy review every time a new field lands in the schema. An illustrative budget for a mid-sized studio running a serious first-party stack, including tooling and cloud costs plus a fraction of an engineer, is comfortably into six figures a year before anyone has run a single campaign against it. That is a reason to know what it is buying.

The larger cost hits the funnel. Studios that commit to first-party strategy tend to reach for the mechanisms that generate first-party data: an early account creation prompt, an email capture, a web shop with a login. Each of these adds friction between install and first session, and each is a small tax on the day-one retention that every network you buy from is optimising towards. A studio that adds a login wall to collect emails it cannot yet use has traded a certain cost for an uncertain future benefit and, in the process, made its paid installs slightly more expensive to convert.

There is a quieter second-order effect. The vendors selling first-party strategy are, in many cases, the same vendors who benefit from the data flowing through their pipes. That doesn't make the advice wrong. It does mean the advice isn't neutral, and a studio should weigh it as it would weigh any recommendation from a party with a stake in the answer.

Three tests before you build

Run the studio against each of the following; a pass on all three justifies a first-party programme, and a pass on one or none doesn't.

Reach. Can you contact a meaningful share of your players outside the app, through a channel you control, at a cost below what a network would charge to re-acquire them? Push alone rarely qualifies. An account system with email or a web shop with repeat visits might.

Repeat. Do players move between your titles, or return to a title after a long absence in numbers large enough to matter? If you can't name the destination for a re-engaged player, you don't need a first-party pool to find them.

Rate. Do you generate enough qualified events per week, purchases or high-value retention milestones, to train or even meaningfully segment anything? As an illustrative threshold, a few hundred payers a week is a segmentation exercise; a few thousand starts to be a modelling one.

An illustrative example. A puzzle studio with one live title and a second in soft launch has push as its only channel and no account system, with a low four-figure count of weekly payers. It fails reach and repeat and marginally passes rate, and the first-party programme quoted to it would consume most of a year's measurement budget.

The decision rule says no, or at least not yet.

What to build instead

Failing the tests doesn't mean doing nothing. It means spending the same budget where a small studio has actual pull.

Event design. The single highest-return measurement investment for most studios is a clean, stable, well-documented set of in-game events that map to real player value. This costs a few weeks of engineering and analyst time and it improves every downstream system: your SKAN and AdAttributionKit conversion value schema, your network optimisation events, your own retention reporting.

Conversion value schema. On iOS the postback is the signal the platforms actually receive. A schema that encodes early value well is worth more to your buying than any warehouse, because it is the one first-party asset the platforms must act on.

Platform models. Accept that the networks' cross-app models are better than yours at prediction and feed them the cleanest possible signal. The advantage is in the quality of what you send, not in retaining a copy.

Experiment infrastructure. A studio that can run a clean geo or cohort holdout learns more about incrementality in a month than a first-party pool will tell it in a year.

When the answer changes

The tests are not permanent. A second title that shares an audience, a web shop that players actually return to, a payer base that grows past the modelling threshold: any of these flips a test from fail to pass, and the case for first-party investment strengthens accordingly. Revisit the three tests every six months and build when they pass, not when a vendor says the industry has moved.

Until then, the studio that spent its measurement budget on event design and a working holdout will know more about its players than the one that spent it on a warehouse to store what it could not act on.

Related archive reading

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

• The agency incentive problem nobody fixes
• Why players lie in surveys, and what their session data says instead