Scenario planning for a world with no device IDs

The no-identifier future that measurement teams keep planning for is not the one arriving. Nobody is switching the device ID off. Instead its coverage of acquisition is shrinking, door by door: third-party stores on Google Play in the US, web shops that bill outside the app, cross-developer bundles on iOS. The identifier survives on every device and attaches to a smaller share of the players who matter. That is already the live case, and it is the case most measurement roadmaps do not have a line for.

The argument here is that there are three plausible identifier scenarios over the next two to three years, that the third is the one already under way, and that a small set of moves pays off whichever one dominates. Fund those no-regret moves now and hold the rest as options with written triggers. Practitioners who read the Android identifier as safe for the rest of the decade, and their measurement as therefore settled, will disagree. The doors are the problem, not the ID.

Why identifiers exist at all

An identifier is a commercial permission, not a technical fact. Apple withdrew the default availability of its identifier in 2021 because it judged the trade between advertising revenue and platform positioning differently from Google. Google kept its identifier because its business collects a large share of the advertising revenue that flows through it. Google's Privacy Sandbox for Chrome was largely retired around the turn of this year, which tells you that even the platform holder found replacing identifiers harder than keeping them.

That framing explains why the future is uncertain. Each platform's identifier policy will move when its commercial calculus moves, and regulators can move that calculus from outside. Adjust's March report on gaming app sessions put ATT opt-in at 39 percent in the first quarter, barely changed on a year earlier. Apple's WWDC in June left AdAttributionKit and SKAdNetwork untouched, and that stability is a data point, not a guarantee.

The three scenarios

Scenario one: the present persists. iOS stays coarse and aggregated. Android keeps a resettable identifier with consent prompts that most users accept by default. This is the base case most teams are already built for.

Scenario two: Android converges on iOS. Google restricts or degrades the advertising ID, and Android attribution moves to an aggregated, platform-mediated framework. The industry has rehearsed this since Google announced Privacy Sandbox on Android years ago, and it hasn't happened, which is why many treat it as a dead scenario. The commercial logic has not gone away.

Scenario three: identifiers stay but the doors multiply. Third-party stores on Google Play in the US, web shops that bill outside the app, plus cross-developer subscription bundles on iOS: each creates a path into your game where the identifier either does not travel or arrives without the context that made it useful. The identifier survives while its coverage of your actual acquisition shrinks. This is the scenario least discussed and, as argued above, the one already under way.

The second-order effect

The obvious consequence of losing identifiers is worse attribution. The less obvious one: first-party identity gets much more valuable, and that changes who inside a studio owns measurement.

A logged-in player with a first-party account is measurable across every door in scenario three and across both platforms in scenario two. That makes a login flow and an account system, along with a web shop identity, into measurement infrastructure rather than monetisation features alone. Studios have mostly justified web shops on margin. In two of three scenarios the stronger justification is that they are the only place a studio sees a player with an identity it owns.

There is a trade-off. Pushing login earlier costs early retention, and the studios best placed to measure will be the ones willing to pay some day-one friction for it. That decision belongs to product, not to the UA team, which is exactly why it rarely gets made.

The MMP's role changes too. Under scenarios two and three attribution becomes modelling, and the model's assumptions matter more than the SDK does. AppsFlyer's Series E in June, as reported by Axios and Calcalist, brought four names onto its investor list as minority strategic holders: Google, Meta, Unity, Moloco. When the intermediary that models your attribution is partly owned by the parties it measures, you want your own first-party truth set to check the model against. That is a reason to build one, not a reason to leave the MMP.

A framework: no-regret moves and options

Sort every candidate investment by how many scenarios it pays off in.

No-regret moves, worth funding regardless:

• A first-party analytics layer that defines cohorts identically for every entry path, with a studio-owned qualifying event.
• An account system and login incentive that a majority of payers adopt within the first sessions.
• A conversion value schema and its Android equivalent that you actually iterate on quarterly, rather than set once.
• An incrementality testing cadence, because every scenario reduces deterministic attribution and geo or holdout tests are the fallback in all of them.

Then the options, funded only when a trigger fires. Deep investment in Android-specific probabilistic modelling waits until Google publishes a deprecation timeline. Per-store attribution integrations wait until any third-party store passes a set share of your Android installs; five percent is an illustrative threshold. Server-side web shop attribution waits until web shop revenue passes a set share of IAP, illustratively ten percent.

Write the triggers down with the plan, so you exercise the option on evidence rather than on the mood of a quarterly review.

What to do this quarter

This site wrote in January, under the title SKAN 4 and AdAttributionKit, a year in, that buyers get back less than they hoped. That is the state of the most mature no-identifier system in the market, and it is what Android looks like if scenario two arrives. The teams in the best shape then will not be the ones with the cleverest fingerprinting fallback. They will be the ones who can answer, from data they own, what share of last month's payers arrived through which door, whether an identifier came along or not.

Related archive reading

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

• The consolidation endgame for mobile ad tech
• Could ATT Be Rolled Back? A Scenario Worth Planning For