Android Alternative App Stores and Attribution Setup

Reports this week have the Ninth Circuit denying Google's appeal against the Epic v Google injunction, which removes the last near-term obstacle to opening Android in the US to rival app stores and alternative billing. For a UA team this isn't primarily a legal update. It changes where installs on Android can originate from, and nobody designed a single-path Play Store attribution setup for what is coming.

What the injunction actually requires

The injunction that followed the December 2023 jury verdict bars Google from blocking rival app stores on Android and from requiring Google Play Billing as the only payment option, alongside other anti-steering restrictions. Google has been taking phased compliance steps through the year while its appeal was pending, and this week's Ninth Circuit ruling removes the argument that those steps were provisional. Android in the US is heading toward more than one legitimate install and payment path for a given app, on a timeline a UA team can no longer treat as hypothetical.

Where install attribution gets harder

Most Android attribution today leans on the Play Install Referrer API to pass a click identifier through to a fresh install, and that API covers only installs originating from Google Play. An install that comes from a rival store, or a sideloaded APK distributed through a web link, does not carry the referrer through the same path, which leaves an MMP falling back to a different attribution method for that traffic or losing sight of it entirely. So a campaign that assumes an MMP can attribute every Android install the way it attributes a Play Store install will start under-reporting the moment meaningful volume moves through an alternative channel.

A setup checklist for the transition

A practical starting point for a measurement team before alternative-store volume becomes material: confirm with the MMP in use what attribution method it applies to non-Play install sources today, since coverage varies significantly between providers and some are further along than others; set up a distinct source or channel tag for any traffic pointed at a rival store or a direct APK link, so performance from that path is visible separately rather than folded into a generic Android bucket; audit whether anyone built fraud or duplicate-install filtering logic that quietly assumes the Play Install Referrer as a check, since that assumption breaks for non-Play traffic; and build a reporting view that can show Play-attributed and non-Play-attributed installs side by side, because the two are likely to have different cost and quality profiles for some time.

Why this matters more for genres already testing web shops

Studios already running web shop experiments in response to the US contempt ruling in Epic v Apple are the ones most likely to feel an Android measurement gap first, since they have already built the habit of routing some purchase and install traffic outside a platform's default path. A team that has already done the work of tagging and separating web shop traffic on iOS has a template it can reuse for Android's rival-store traffic, rather than starting the tagging discipline from nothing. Teams that have not yet built that habit on either platform should treat this as the moment to build it once, for both, rather than solving the same problem twice a few months apart.

Fingerprinting is not a fallback plan

The obvious workaround for the referrer gap is probabilistic matching, inferring an install's source from device and timing signals rather than a passed identifier. This has known accuracy limits even in mature use, and regulatory scrutiny of probabilistic attribution has only increased, not decreased, over the past few years. Building an Android measurement plan whose accuracy depends on fingerprinting filling the gap left by alternative stores is trading a legal exposure problem for a new one, and it produces attribution numbers a team should trust considerably less than referrer-based data even where a vendor claims high match rates.

Budget attribution differently from install attribution

The same gap in referrer coverage affects more than just install counts. Post-install revenue and retention attributed back to a specific campaign rely on the same identifier chain that the Play Install Referrer supplies for Play-sourced installs. An install from a rival store or a sideloaded APK never carried that identifier through, so it will show weaker downstream attribution even if the underlying user behaves identically to a Play-sourced one. A team that notices apparently poor retention or revenue performance from an alternative-store cohort should rule out an attribution gap before concluding the traffic itself is lower quality, since the two look identical in a report that does not separate them by install source.

Talking to the rest of the team before the numbers move

A measurement change like this one tends to surface as a confusing dip in reported installs or an unexplained shift in blended cost per install weeks after the underlying cause, if nobody outside the measurement team was told to expect it. Flagging the change ahead of time to whoever reviews weekly UA reporting, media buyers, finance, studio leadership, with a plain explanation of why a portion of Android installs may temporarily show as unattributed or under a generic source while MMP coverage for non-Play channels catches up, avoids a scramble to explain a metric that looks broken but is not. It's a small step. It's also the one teams skip most often, because the engineering and MMP configuration work absorbs all of a measurement team's attention while the reporting-side conversation waits.

What to do before volume shows up

The volume through rival Android stores in the US is not large yet, which is exactly why now is the moment to get the plumbing right rather than after a campaign has already been running blind through a channel nobody built reporting for. As we covered in SKAN Conversion Value Mapping, Revisited for 2025, the discipline of building a measurement setup for where the platform is going rather than where it has been applies just as directly on Android as it does to Apple's SKAdNetwork successor. Android distribution is moving in the same direction Google's compliance obligations point it, and an attribution setup that already accounts for more than one install path won't need a rebuild when that volume arrives.

Related archive reading

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

• Third-party app stores are live on Google Play in the US. A day-one UA playbook
• Alternative app stores will not lower your CPI