Most conversations about why mobile growth teams struggle to work together focus on organizational structure, incentives, or team willingness. Those things matter. But in our experience, the more common blocker is technical: teams can’t share signal across paid, organic, CRM, and creative because their measurement infrastructure doesn’t support it.
Integrated growth requires three things technically: shared data across teams, a way to identify the same user across different surfaces, and a signal that moves between departments fast enough to act on. In 2026, all three of these are harder than they sound, and the reason usually comes back to attribution.
This article is a practical diagnosis of what’s actually broken in the measurement layer, and what teams can do about it. We’ll cover how ATT and AdAttributionKit changed the game, why the damage is worse for integrated teams than siloed ones, and how web-to-app and first-party data can partially fill the gaps. Lastly, we’ll also propose what a well-instrumented growth stack looks like today.
How ATT and AdAttributionKit changed mobile attribution
Before ATT, mobile attribution worked at the user level. A tracking identifier (IDFA) followed the user from ad impression to install to post-install behaviour. Every team could see the same user journey. MMPs like Adjust, AppsFlyer, and Singular stitched the picture together, and while the system wasn’t perfect, it gave teams a shared view of what was happening.
ATT changed that fundamentally. With consent rates sitting between 14% and 35%, depending on how you measure, the majority of iOS users are now invisible to deterministic tracking. Apple’s replacement framework, AdAttributionKit (which evolved from SKAdNetwork) provides privacy-preserving measurement, but with significant constraints: postbacks are delayed by 24 to 48 hours, conversion values are limited to 64 possible combinations across three measurement windows (days 0-2, 3-7, and 8-35), and the data is aggregated rather than user-level. Only campaigns hitting high crowd-anonymity thresholds receive fine-grained conversion values at all; lower-volume campaigns get coarse values (low, medium, high) or nothing.
According to AppsFlyer’s 2025 survey, only about 25% of app marketers said SKAN had a noticeably positive impact on iOS UA, while 30% still reported challenges and negative impact. As one industry analysis put it, “most mid-market app advertisers had simply stopped treating SKAN as their primary measurement source. They used it as a sanity check against modeled platform data.”
Why fragmented attribution data is worse for integrated growth teams
Here’s what most attribution articles miss: when each team works in isolation, partial data is a local problem. The paid team has incomplete attribution, so they optimise based on what they can see. Moreover, the ASO team has store-level conversion data without post-install visibility, and the CRM team sees in-app behaviour without acquisition context. Each team learns to work within its own constraints.
When you’re trying to connect these functions, partial data becomes a system-wide problem. Think of it this way: the paid team can’t tell the ASO team which campaigns are driving high-quality users because the attribution is too noisy, while the ASO team can’t tell the CRM team which store page variant a user saw because there’s no persistent identity connecting the store visit to the install. Lastly, the CRM team can’t tell the paid team which acquisition sources produce users who retain and spend because the user-level link between acquisition and retention is broken.
In a siloed setup, these gaps are invisible. Each team has “enough” data to do its own job. In an integrated setup, the gaps become the central bottleneck, because the whole point of integration is to move the signal between teams, and the signal is fragmented at the source.
How to track users across web, app store, and owned channels
Stitching together a user journey across web, app store, paid, and owned channels requires being able to identify the same person across those surfaces. In 2026, that’s one of the hardest technical challenges in mobile growth.
Before ATT, the IDFA provided a shared identifier that followed the user everywhere. Today, pre-install sessions are mostly anonymous. A user who sees an ad, visits a web landing page, and then downloads the app from the store may appear as three separate events in three separate systems with no connection between them.
Web-to-app flows partially solve this problem by moving the conversion to a surface the team controls. On the web, you have pixel-based tracking, server-to-server attribution, and UTM parameters. A user who converts on a web landing page and then installs the app can be identified across both surfaces if the team has set up shared login systems and deep linking properly. That’s a meaningful improvement over sending all traffic to the app store and hoping the SKAN postback tells you something useful 48 hours later.
But web-to-app only covers traffic you route through the web. Organic app store discovery, direct search, and browse traffic still pass through Apple’s or Google’s measurement systems, with all the limitations that entail. The result is a split: some of your user journey data is clean and connected (web-to-app), and some is fragmented and delayed (store-native). Managing both simultaneously, and knowing which data to trust when they tell different stories, is one of the operational challenges that doesn’t get talked about enough.
Why first-party data is the foundation of mobile measurement in 2026
Our CEO Andy Carvell’s integrated growth manifesto makes the case that first-party data should be treated as a strategic asset, not departmental property. When the external signal keeps shrinking, the data each team already collects (keyword rankings, creative test results, retention curves, push notification engagement, purchase patterns) becomes the most reliable foundation for decision-making.
But treating first-party data as strategic infrastructure requires real investment. Most teams collect first-party data adequately. Where they fall short is in making it accessible across departments. As we covered in our framework for cross-team data sharing, the practical requirements include shared dashboards that combine data from paid, organic, CRM, and product into a single view; consistent tagging and taxonomy across systems so that data from different tools can actually be compared; regular reporting cadences where each team shares what they’re seeing with the others; and someone who owns the process of making sure this data actually moves.
The technical infrastructure (a data warehouse, a BI tool like Looker, event tracking through something like Amplitude or Mixpanel) is table stakes. The harder part is the operational layer: deciding what gets shared, how often, and who’s responsible for making sure it happens.
How incrementality testing works when attribution falls short
When last-click attribution breaks down, incrementality testing offers a fundamentally different approach. Instead of trying to assign credit to individual touchpoints, incrementality asks a simpler question: did this campaign cause outcomes that wouldn’t have happened without it?
The most common approach is a geo holdout experiment. You split your markets into two groups: a treatment group where the campaign runs normally, and a control group where it’s paused. You compare outcomes between the two groups, and the difference is the incremental effect of the campaign. According to data from Stella’s platform, across 225 geo-based tests, the median incremental ROAS was 2.31x, and most teams discovered their incremental ROAS was 30-70% different from what the platform-reported ROAS showed.
Incrementality testing doesn’t replace attribution. It complements it by answering a question attribution can’t: whether the spend is generating net-new results or simply capturing demand that already existed. For teams spending above a certain threshold (generally $500K+ per month across channels), a regular incrementality testing cadence is becoming standard practice. For smaller teams, periodic tests on the largest campaigns can still surface insights that fundamentally change how budget is allocated.
The limitation is that incrementality testing requires sufficient scale and patience. A meaningful geo holdout needs enough markets, enough time (typically 4-8 weeks), and enough volume to reach statistical significance. It won’t tell you which creative variant is working best. For that, you still need the attribution and experimentation layer.
Do you still need an MMP in 2026?
Mobile Measurement Partners like Adjust, AppsFlyer, and Singular remain important in a post-ATT world, but their role has changed. They still provide the infrastructure for ingesting SKAN and AdAttributionKit postbacks, unifying reporting across networks, managing deep links, and providing fraud detection. For teams running paid campaigns across multiple networks, that aggregation layer is essential.
Where MMPs fall short is in providing the connected, cross-surface measurement that integrated growth requires. MMPs were built for a world where user-level tracking was possible. In a post-ATT world, they’ve adapted by layering probabilistic modeling and machine learning on top of SKAN data, but the result is modeled rather than deterministic. As the 2026 industry consensus has settled, most sophisticated teams use SKAN/AAK postbacks as a sanity check, layer MMP modeling on top, and pair everything with first-party signals and incrementality testing.
The minimum measurement stack for 2026 looks something like this: a server-side conversion API with strong match rates, platform-reported conversions treated as directional rather than definitive, periodic incrementality testing, and a first-party data foundation that connects the dots across owned surfaces. For teams with larger budgets, adding a probabilistic MMP, a dedicated incrementality testing cadence, and marketing mix modeling rounds out the picture.
How PressPlay and Catchbase fit in
Within Phiture’s stack, PressPlay and Catchbase address specific parts of the measurement challenge. PressPlay generates creative performance signals through automated store page experiments on Google Play, producing data about which visual approaches and messaging convert, that’s immediately shareable across the growth team. Because this testing happens on the store page rather than through paid campaigns, it produces a clean, first-party signal that isn’t subject to ATT or SKAN limitations.
Catchbase optimises Apple Search Ads spend, operating in the one paid channel on iOS where Apple provides deterministic attribution. The performance data from Catchbase connects naturally to ASO (which keywords are worth bidding on vs covering organically) and to CRM (which Apple Ads cohorts retain and convert).
Neither tool solves the measurement problem on its own. But both create a first-party signal that flows across teams, which is the operational foundation that integrated growth depends on.
What a well-instrumented integrated growth stack looks like in 2026
Putting it all together, a growth team that’s serious about integrated measurement in 2026 needs several layers working together:
An event tracking foundation (Amplitude, Mixpanel, or similar) that captures in-app behaviour and makes it accessible across teams through shared dashboards and a consistent taxonomy.
An MMP layer (Adjust, AppsFlyer, or Singular) that ingests SKAN/AAK postbacks, unifies reporting across ad networks, and provides probabilistic modeling where deterministic attribution isn’t available.
A first-party data sharing process that moves insights from paid, organic, CRM, and creative teams on a regular cadence, through shared dashboards and structured reporting.
Web-to-app flows for campaigns where full attribution visibility matters, giving teams ownership of the user journey from first click to install.
Incrementality testing on a periodic basis for the largest spend categories, providing ground-truth data about what’s actually driving net-new outcomes.
Automated experimentation tools like PressPlay (for store page creative) and Catchbase (for Apple Ads) that generate first-party signal at volume without relying on degraded third-party tracking.
Someone who owns the connections. A data team or growth lead who’s responsible for making sure signal moves between systems and reaches every team that needs it. Without this, even the best-instrumented stack produces data that stays in dashboards nobody outside one team ever opens.
What’s next
The measurement landscape will keep evolving. AdAttributionKit is gaining features (re-engagement measurement, configurable windows, country codes at high anonymity tiers), but it’s not becoming less privacy-restrictive. Google’s Privacy Sandbox is introducing its own constraints on Android. The teams that build their measurement foundation around first-party data, incrementality, and cross-team signal sharing will be better positioned than those still waiting for the old deterministic model to come back.
At Phiture, we help mobile teams build the measurement and growth infrastructure that integrated growth requires. Our integrated growth approach connects paid, organic, creative, and CRM into one system, and our tools generate the first-party signal that feeds it. If you’re working on connecting your measurement stack, get in touch.
FAQ
Why is attribution the real barrier to integrated growth?
Integrated growth requires signal to flow between paid, organic, CRM, and creative teams. When attribution is fragmented (delayed postbacks, aggregated data, no persistent user identity), there’s no clean data to share. Each team ends up working with a partial picture, and the connections between them break down.
What is AdAttributionKit and how does it differ from SKAN?
AdAttributionKit is Apple’s successor to SKAdNetwork. It retains SKAN’s privacy-preserving approach (aggregated postbacks, conversion values, crowd anonymity) while adding re-engagement measurement, configurable attribution windows, and support for alternative app stores. Functionally, it operates similarly to SKAN 4, and both can coexist in the same app.
What is incrementality testing?
Incrementality testing measures whether a campaign caused outcomes that wouldn’t have happened without it. The most common method is a geo holdout, where you compare markets where the campaign runs against markets where it’s paused. The difference reveals the true incremental effect, which is often significantly different from what platform-reported attribution suggests.
Do you still need an MMP in 2026?
Yes, for most teams. MMPs provide the infrastructure for ingesting SKAN/AAK postbacks, unifying reporting across networks, and detecting fraud. Their role has shifted from deterministic user-level tracking to probabilistic modeling and data aggregation, but that aggregation layer remains important for teams running multi-network paid campaigns.
How do web-to-app flows help with measurement?
Web-to-app flows move conversion moments to a surface the team controls, where pixel-based tracking, server-to-server attribution, and UTM parameters provide full visibility. This gives teams clean, connected data for the traffic they route through the web, partially compensating for the signal loss on store-native traffic.
What does a minimum measurement stack look like in 2026?
At minimum: an event tracking platform for in-app behaviour, an MMP for campaign attribution and SKAN/AAK reporting, a first-party data sharing process across teams, and periodic incrementality testing on major spend categories. Teams with larger budgets add web-to-app flows, automated experimentation tools, and marketing mix modeling.
Table of Contents
