App Attribution: Bloom’s 2026 Strategy Shift

The year 2026 presented a significant challenge for Anya Sharma, Head of Growth at “Bloom,” a burgeoning meditation app. After Apple’s iOS 14.5 privacy changes, the once-clear waters of app attribution had become murky, particularly with the limitations of SKAdNetwork. Bloom had seen consistent growth, reaching nearly 5 million monthly active users, but Anya felt they were flying blind on what truly drove their most valuable installs. “We’re spending millions on user acquisition,” she explained to her team, “and while SKAdNetwork gives us a signal, it’s a blurry snapshot, not the high-definition video we need to scale effectively.” Her problem wasn’t just about measuring installs. It was about understanding the entire user journey, from initial ad impression to a subscribed, engaged user. How could Bloom accurately measure campaign performance and allocate budget when the traditional attribution models were no longer viable?

The SKAdNetwork Conundrum: Understanding Its Limits

SKAdNetwork, or SKAN, was Apple’s privacy-centric framework designed to provide app install attribution without compromising user privacy. It aggregates conversion data, delaying reporting and limiting granularity. For Anya, this meant she couldn’t see individual user-level data, which platforms like Google Ads and Meta relied on heavily for optimization. The conversion window was limited, and the postback data, while anonymous, often lacked the rich detail needed to truly understand user quality beyond a basic install. “It tells us that an install happened and which campaign likely drove it, but not much about who that user is or what they did immediately after install,” Anya lamented. This lack of detailed post-install event data made it difficult to optimize for high-value actions like subscription sign-ups or content engagement, which were critical for Bloom’s revenue model. Many in the industry initially hoped SKAN 4.0, released in late 2022, would provide a panacea. It introduced more granular conversion values, multiple postbacks, and hierarchical source identifiers. While an improvement, it still operated within a privacy-preserving framework, meaning the data remained aggregated and delayed. According to a 2025 eMarketer report on mobile advertising trends, over 60% of app marketers still cite “attribution accuracy” as their top challenge, even with SKAN 4.0 in play. The report highlighted a growing reliance on blended data approaches rather than a single source of truth. This confirmed Anya’s growing suspicion: SKAN was part of the puzzle, but far from the whole picture.

Building a Multi-Layered Measurement Strategy

Anya knew Bloom needed to move beyond a singular reliance on SKAN. Her team started by categorizing their data sources into three main pillars: SKAN data, first-party data, and probabilistic modeling. This layered approach aimed to construct a more complete view of their users. First, they focused on maximizing the utility of SKAN. This involved carefully configuring their conversion values within SKAN 4.0 to capture the most meaningful early-lifecycle events for Bloom, such as completing the onboarding tutorial, initiating a free trial, and making a first subscription. They worked closely with their ad partners to ensure these conversion values were mapped effectively. “It’s about making the most of the limited signals you do get,” Anya noted. “We spent weeks refining our conversion value schema, prioritizing actions that directly correlated with long-term retention and revenue.” Next, Bloom significantly enhanced its first-party data collection. This meant collecting data directly from their app and website after a user installed or visited, with explicit user consent. They implemented a strong customer data platform (CDP), opting for Segment, to unify user profiles across their app, website, and CRM. This allowed them to track in-app behavior, subscription status, and engagement metrics directly. “The beauty of first-party data is its precision,” Anya explained. “When a user signs up for a premium meditation course, we know exactly who they are and what they did. The challenge is connecting that back to the initial ad touchpoint without privacy violations.” This required a careful approach to user consent, prominently featuring clear privacy policies and opt-in mechanisms within the app. The third pillar involved exploring probabilistic modeling and advanced analytics. While deterministic, user-level attribution became largely obsolete with privacy changes, probabilistic methods offered a way to infer user journeys. These models use aggregate data points like IP addresses, device types, operating systems, and time stamps to make educated guesses about which ad exposure might have led to an install or conversion. Bloom partnered with a specialized mobile measurement partner (MMP), AppsFlyer, which had invested heavily in privacy-centric probabilistic attribution solutions. AppsFlyer’s “Privacy Cloud” allowed Bloom to analyze aggregated, anonymized data sets to identify trends and patterns that suggested campaign effectiveness, even without individual user IDs. This wasn’t a perfect science, but it provided valuable directional insights.

Implementing Server-to-Server Integrations and Data Warehousing

A critical technical shift for Bloom was the move towards server-to-server (S2S) integrations for event reporting. Instead of relying solely on client-side SDKs, which could be blocked by ad blockers or impacted by network issues, Bloom configured their backend servers to directly send event data to their MMP and advertising platforms. This provided more reliable and secure data transmission. “S2S reporting is a non-negotiable now,” Anya stated. “It gives us higher fidelity data and better control over what information is shared, and when.” This was particularly important for reporting important post-install events back to ad platforms, helping their algorithms optimize for quality users without exposing raw user data. To bring all these disparate data streams together, Bloom invested in a modern data warehouse solution, opting for Snowflake. This allowed them to ingest data from SKAN postbacks, their CDP, AppsFlyer, and their internal analytics systems into a single, centralized repository. Their data engineering team then built custom dashboards and reports using business intelligence tools like Tableau. This provided a well-rounded view that Anya hadn’t had before. “Before, we were looking at fragmented reports from each ad platform, trying to stitch them together in spreadsheets,” she recalled. “Now, we have a unified source of truth, allowing us to compare performance across channels with much greater confidence.” One editorial aside here: many companies think buying a CDP or a data warehouse solves all their problems. It doesn’t. These are tools. The real work is in defining your metrics, ensuring data quality, and building the right queries and dashboards. Without a clear strategy, you just have a very expensive data junk drawer.

The Role of Incrementality Testing in a Privacy-First World

Even with a strong blended attribution model, Anya recognized the inherent limitations of any attribution system in a privacy-constrained environment. Attribution tells you which touchpoint likely contributed to a conversion. Incrementality testing, however, answers a different, more fundamental question: “Would this conversion have happened without this specific ad exposure?” Bloom began implementing rigorous incrementality tests. For instance, they would pause specific ad campaigns in certain geographic regions or for defined user segments (control groups) while continuing them in others (test groups). By comparing the performance of the test group against the control group, they could isolate the true incremental lift provided by the campaign. “This is where the rubber meets the road,” Anya explained. “If an ad campaign drives a lot of attributed installs but shows zero incremental lift in sales when we pause it, then we’re essentially paying for conversions that would have happened anyway.” They ran these experiments on their major acquisition channels, particularly for their Google App Campaigns and Meta campaigns. For example, they might run a brand awareness campaign on Meta targeting a specific lookalike audience in Atlanta, Georgia. They would then create a matched control group in a demographically similar area, say, Athens, Georgia, where the campaign was not shown. By comparing new user acquisition rates and subscription conversions between the two cities over a set period, they could quantify the true incremental value of that specific campaign. This approach, while more complex to set up and analyze, provided Anya with the conviction she needed to make significant budget allocation decisions.

The Outcome: Clearer Vision and Strategic Allocation

After nearly a year of iterating on their new measurement framework, Anya’s team at Bloom saw tangible results. They were no longer solely optimizing for raw installs reported by SKAN. Instead, they were making decisions based on a richer understanding of user quality and incremental impact. “We discovered that some campaigns that looked great on a last-touch SKAN report actually provided very little incremental value,” Anya revealed. “Conversely, some upper-funnel awareness campaigns, which traditionally looked ‘expensive’ on a direct attribution model, were driving significant incremental growth when measured through our blended approach and incrementality tests.” This allowed Bloom to reallocate marketing spend more effectively, shifting budget from channels with low incremental impact to those that truly moved the needle on long-term subscriptions. For example, they reduced spend on certain generic keyword campaigns in Google Ads that were primarily capturing users already on the verge of installing, and instead increased investment in creative-led brand campaigns on Meta that were proven to introduce Bloom to entirely new, high-value audiences. By combining SKAdNetwork data with strong first-party analytics, probabilistic modeling, and a commitment to incrementality testing, Bloom transformed its app measurement capabilities. Anya and her team could now confidently answer the question of what truly drove their most valuable installs, moving from blurry snapshots to a much clearer, actionable picture of their growth. Working through the complexities of app attribution in a privacy-first era requires a multifaceted approach, moving beyond single-source solutions to build a resilient and insightful measurement framework that combines diverse data streams and rigorous testing.