App Tracking Without Illusions: GA4, UTM, MMP, and SKAdNetwork in Mobile Campaign Attribution 📱

In digital marketing, we have grown accustomed to a relatively simple and linear user journey – like a straight flight through empty space:

ad click → landing page view → conversion

All campaign parameters travel effortlessly in the URL query string, the browser retains cookies, and the analytics script seamlessly logs the session and attributes the conversion.

However, in mobile app install campaigns, this straight path collides with the massive gravity of the app stores.

Why the Install Step is an Attribution "Black Hole"

In the mobile app universe, your app is the center of the solar system, and ad networks (like Meta or Google) are distant planets. The user journey (our space probe) looks very different:

ad click on planet "Meta" → entering the gravitational pull of the App Store → download → installation → landing (first launch) → conversion in the center (App)

When a user clicks an app install ad, they leave their familiar orbit and enter the isolated sandbox of the app store. At the time of installation, the newly downloaded app has no access to URL query parameters, browser cookies, or the user's flight history. All source data disappears beyond the event horizon.

The install step is a black hole that physically breaks the traditional tracking chain and swallows the data.

Google Analytics 4 (GA4) is exceptionally good at tracking user behavior once they are inside the app. However, it cannot automatically link that first app launch back to an ad click on an external advertising network. For that, a dedicated attribution architecture is required.

1. Two Different Questions: Product Analytics vs Attribution

When evaluating mobile apps, we must separate two entirely different analytical problems.

2. How GA4 Works in Mobile Apps

The Firebase SDK automatically tracks standard technical events and allows you to define custom product events, such as:

• first_open (the first time the app is launched – the app equivalent of an install),
• onboarding completion and account sign-ups,
• subscription trial starts and in-app purchases.

first_open vs Install

It is critical to remember that:

store download ≠ install reported by network ≠ first_open in GA4

A user can download an app from the store but never actually open it. In this case, the ad network will report an install, but GA4 will never log a first_open event.

By default, without advanced configuration, GA4 will attribute the acquisition source of most new users as (direct) / (none) or organic. The campaign details simply get lost in the transition through the app store. While Google allows you to manually populate this data by sending a campaign_details event shortly after install, doing so requires additional technical integrations.

3. UTM vs MMP – Why They Are Not the Same

A common misconception is that UTM query parameters in ad links can replace a full MMP integration.

A UTM parameter is merely a static label describing a click source. An MMP is an active attribution engine.

When a user clicks a link with UTM parameters and the app is already installed, a deep link (or App Link/Universal Link) opens it directly. The app reads the UTM parameters and passes them to GA4 – tracking works perfectly.

The issue arises with new installs. The presence of UTM parameters in the ad link does not mean they will magically survive the journey through App Store or Google Play and arrive in the app.

How MMP Attribution Works

The MMP SDK integrates directly with mobile operating systems and ad networks (Meta, Google, TikTok). When a user clicks an ad, the MMP logs the click (and the device identifier). Once the app is installed and launched, the MMP SDK gathers available device signals and queries its server. The system performs matching (either deterministic via Google Play Install Referrer, or probabilistic) to attribute the install, and instantly sends a postback to the winning ad network.

{
  "install_referrer": "utm_source=facebook&utm_medium=paid_social&utm_campaign=summer_sale&fbclid=...",
  "referrer_click_timestamp_seconds": 1783856200,
  "install_begin_timestamp_seconds": 1783856240,
  "google_play_instant_enabled": false
}

4. iOS, ATT, and SKAdNetwork / AdAttributionKit

While Android attribution is relatively straightforward, iOS became a battlefield after Apple introduced App Tracking Transparency (ATT).

Users must opt-in to tracking. If they select "Ask App not to Track" (which over 60% of users do), the app loses access to the IDFA identifier and is barred from using fingerprinting. To address this, Apple built SKAdNetwork (SKAN), now evolving into AdAttributionKit, as a privacy-preserving attribution framework.

How SKAdNetwork Operates

Apple acts as the attribution referee:

1. The user clicks an ad digitally signed by the ad network.
2. Apple records this click at the iOS system level.
3. After the app is installed and opened, iOS determines if it qualifies for attribution.
4. Instead of sharing user data, Apple sends a delayed, anonymous postback directly to the ad network or MMP.

{
  "version": "4.0",
  "ad-network-id": "v9wttpbfk9.skadnetwork",
  "transaction-id": "61a34b22-87ff-4375-9c88-12c83ff111a4",
  "campaign-id": "84",
  "source-identifier": "1084",
  "attribution-signature": "MEYCIQCc3v1qH3...",
  "redownload": false,
  "source-app-item-id": 123456789,
  "fidelity-type": 1,
  "coarse-conversion-value": "high",
  "postback-sequence-index": 0
}

The Mystery of Conversion Values

Since Apple does not report individual user events (e.g., "purchased a $19 item"), post-install activity is measured using Conversion Values. This is a encoded digit (ranging from 0 to 63 in "Fine" mode, or represented as "low", "medium", "high" in "Coarse" mode).

An example conversion value schema:

• 0 = app launched (first_open)
• 5 = user signed up
• 12 = added product to cart
• 40 = completed purchase (low value)
• 60 = premium subscription purchased

Analytics platforms (including Firebase and MMPs) allow you to map these values in their dashboard, while the SDK manages updating the value on-device. Apple transmits the raw value, and the receiving platform decodes it based on your predefined mapping schema.

5. Why Meta Ads, MMP, and GA4 Data Diverge

In a healthy app tracking setup, discrepancies across dashboards are expected. Each platform answers a different question and uses different attribution windows.

Common Sources of Discrepancies:

• Download vs. Launch: Meta counts an install when a user clicks download in the app store. GA4 only logs the install when the user physically opens the app (first_open) and the SDK initializes.
• Lack of Cross-Channel Deduplication: If a user clicks a TikTok ad, then a Meta ad, and installs the app, both Meta and TikTok will claim 100% credit (total: 2 installs). An MMP deduplicates this and assigns 1 install to the last click.
• Attribution Windows: Different systems default to different windows (e.g., 1-day, 7-day, or 30-day click-through).

6. Designing a Rational App Analytics Architecture

Attempting to find a single tool that does everything is a recipe for failure. A mature mobile marketing technology (MarTech) stack divides responsibilities:

┌────────────────────────────────────────────────────────────────────────┐
│                        MOBILE ANALYTICS ARCHITECTURE                   │
├───────────────────┬────────────────────────────────────────────────────┤
│ GA4 / Firebase    │ Product usage, user flows, retention, and LTV      │
├───────────────────┼────────────────────────────────────────────────────┤
│ MMP (AppsFlyer)   │ Paid attribution, deduplication, and fraud checks  │
├───────────────────┼────────────────────────────────────────────────────┤
│ SKAN / AAKit      │ Privacy-safe iOS tracking and campaign optimization│
├───────────────────┼────────────────────────────────────────────────────┤
│ Warehouse (BQ)    │ Raw data merging, CRM revenue reconciliation       │
└───────────────────┴────────────────────────────────────────────────────┘

Using this architecture, business reporting is clean and actionable:

1. Product Dashboard (GA4/CRM): Tracks actual product health (active users, conversion funnels, subscription churn, renewals).
2. Acquisition Dashboard (MMP): Measures paid acquisition efficiency (ROAS, blended CAC, network performance).
3. Executive Dashboard (DWH): Joins network spend with CRM subscription lifecycles to calculate accurate payback periods.

Interactive Attribution Path Map

Below is an interactive map that visually represents the entire user journey and shows what each analytics system sees (GA4, MMP, SKAdNetwork) at every stage. Click a perspective and each node to see the details.