What is Google ICM (Integrated Conversion Measurement)?

What exactly is Integrated Conversion Measurement (ICM)? How does on-device conversion measurement work — the process of tracking conversions directly on the user’s device? What role do Firebase (GA4) and the On-Device Measurement (ODM) library play in this system? And finally, what steps must advertisers take to implement ICM in their apps and connect it with Google Ads and MMPs (Mobile Measurement Partners)? You’ll find all the answers in the text below.

The world of mobile user acquisition is undergoing a fundamental transformation in how campaign effectiveness is measured. Data that for years served as the foundation of performance marketing is no longer freely available.

Increasing privacy regulations (such as GDPR and CCPA) and system-level restrictions have drastically reduced the effectiveness of traditional deterministic attribution methods. User identifiers — IDFA on iOS and AdID (GAID) on Android — still exist, but their accessibility is becoming increasingly limited.

The turning point came in 2021 with the introduction of App Tracking Transparency (ATT) in iOS 14.5. Access to IDFA now requires explicit user consent, which only a small percentage of users grant. As a result, marketers have lost the ability to precisely track iOS users. Google is following a similar path — the Privacy Sandbox for Android is gradually limiting access to AdID, signaling the end of the era of full visibility on that platform as well.

The decline of traditional attribution: how privacy is reshaping mobile user acquisition?

The major industry players have developed their own privacy-compliant measurement frameworks to adapt to the new landscape:

AEM (Aggregated Event Measurement) by Meta – AEM is a framework introduced by Meta (Facebook) to measure conversions and events in apps and websites running Meta ads, especially after Apple introduced its App Tracking Transparency (ATT) policy in iOS 14.5. AEM enables Meta to collect and report aggregated conversion data even when a user opts out of tracking. It prioritizes selected conversion events while preserving user privacy through aggregation and anonymization.

ADC (Advanced Dedicated Campaigns) – ADC is a new type of iOS campaign introduced in TikTok Ads, designed to improve iOS ad performance despite privacy limitations such as SKAdNetwork (SKAN). ADC leverages real-time signals (from SANs and in-app user behavior) to enhance ad delivery models, increase conversions, and improve ROAS. (A separate article will cover TikTok’s ADC in detail.)

A new era: aggregation, modeling, and probabilistic attribution

Mobile marketers must accept that perfect attribution is a thing of the past. Real-time optimization is giving way to trend-based strategies, testing cycles need to be longer, and creative performance has become increasingly important in a world of limited targeting.

ICM and similar solutions don’t aim to recreate the old world of tracking — instead, they’re building a new paradigm, where effective marketing and user privacy coexist.

And in 2025, when it comes to attribution on iOS, things are still… well, a bit chaotic 🙂

Airbridge MMP

In the next part, we’ll take a closer look at how ICM works, the practical benefits of implementing it, and the key preparation steps advertisers should take.

Probabilistic attribution: a new approach

The advertising industry has responded to these privacy-driven changes with the development of probabilistic attribution.

Unlike deterministic methods, probabilistic models use statistics and machine learning to estimate which touchpointsinfluenced a conversion.

The analysis takes into account data points such as:

• IP address,
• device type and version,
• operating system,
• time of interaction,
• geographic location,
• and user behavior patterns.

The model then assigns a probability score to each touchpoint, indicating how much it likely contributed to an app installation or other conversion event.

The waterfall attribution model

App Attribution Partners (AAPs) use what’s known as the waterfall model, which ranks different types of attribution based on their level of reliability:

• Click-based deterministic – based on identifiers such as IDFA, AdID, gClid, or gBraid
• Click-based probabilistic – clicks matched using indirect data (IP, device type)
• View-based deterministic – impressions linked to the same device identifier
• View-based probabilistic – impressions matched statistically
• Organic – when there isn’t enough data for attribution

This hierarchy forms the standard foundation of modern attribution models — and it’s exactly what ICM builds upon.

What is Integrated Conversion Measurement (ICM)?

Integrated Conversion Measurement (ICM) is a Google solution designed to improve the measurement of mobile app conversions in environments with limited access to user identifiers.
ICM combines both deterministic and probabilistic elements, providing a more complete view of campaign performance.

The system is particularly relevant for iOS 14+ and Android users in the EEA, where traditional user-level tracking is restricted.

The benefits of on-device conversion measurement

One of the key components of ICM is on-device conversion measurement — tracking conversions directly on the user’s device.
This approach allows advertisers to measure app installs and in-app actions such as logins, purchases, or subscriptions in a privacy-compliant way, without sending user identifiers outside the device.

Google currently supports two main variants of this approach:

• First-party data–based measurement – designed for apps where users sign in and share data such as an email address or phone number. These data points are processed locally via the Google Analytics for Firebase (GA4F SDK) and can be used to attribute conversions within a campaign.
• Event data–based measurement – a newer variant in which the system uses anonymized, temporary event signals from the device (such as event time, device type, or operating system). These data are analyzed locally and processed in a way that prevents user identification, ensuring full compliance with privacy regulations.

Through this approach, ICM combines attribution accuracy with strong data protection — without relying on traditional advertising identifiers.

How ICM works step by step?

1. Ad click, install and app open
The user clicks on a Google App Campaign ad, install and opens the mobile app. At this point, the attribution process begins within ICM, operating in a privacy-safe way without using any user identifiers.

2. Local event registration (Firebase / GA4 + ODM)
The app includes Firebase (GA4) integrated with the On-Device Measurement (ODM) library. On the device, app events such as first_open, purchase, or subscribe are logged.
All data are processed locally — no personally identifiable information (PII) leaves the device.
During this process, the ODM library generates an encrypted token (odm_info) that stores the context of the ad click (e.g., campaign, time, ad format) in an anonymous form.

3. Attribution request (MMP / AAP SDK)
When a conversion occurs in the app (for example, an install or purchase), the Mobile Measurement Partner (MMP/AAP) SDK — such as AppsFlyer, Adjust, or Airbridge — sends an attribution request to Google with the event data.
This request also includes the odm_info token previously generated by Firebase/ODM.

4. Google response (Attribution API)
Google analyzes data from ODM and from the MMP to determine whether the conversion belongs to a Google Ads campaign.
The system classifies the match type as either:

• Deterministic – when identifiers such as gClid, gBraid, or Play Referrer are available, or
• Probabilistic – when the match relies on anonymized signals such as device model, country, OS version, or timestamp.
  Google’s response includes parameters like timestamp, campaign ID, and match type.

5. Validation and reporting (MMP / Firebase)
The MMP (Appsflyer, Singular, Airbridge, Adjust) validates the data and attributes the conversion to the appropriate Google Ads campaign using last-touch logic.
Meanwhile, Firebase (GA4) reports conversions in an aggregated and privacy-safe way, and Google Ads uses this data for modeling and Smart Bidding optimization.

Practical notes

• Implementing ICM requires updating to the latest versions of your attribution partner SDKs and deploying the GA4F SDK with ODM support.
  The odm_info token must be correctly passed from the app to the partner (for example, via S2S).
• If the partner SDK is blocked for users who have not granted consent, ICM will not attribute conversions — the odm_info token will not be transmitted.
  Probabilistic attribution uses shorter conversion windows, so a minimum of 24 hours is recommended.
• For the EEA, UK, and Switzerland, a separate version of ICM (independent of ODM) is planned, scheduled for rollout in Q4 2025.
  The GA4 SDK is required for proper ODM implementation on iOS.

Signals used for verification

ICM leverages a wide range of technical and contextual signals to increase measurement accuracy, including:

• operating system version,
• event time (time of day),
• device type,
• and geographic location.

These help Google’s model determine, with high probability, which conversions can be attributed to a given campaign.

Benefits of ICM

• Better conversion representation – improved visibility in both AAP and Google Ads reports, even for users with limited tracking.
• More accurate Cross-Network Attribution (XNA) – better-quality conversion data within Google Ads.
• Event-level and quasi-real-time reporting – data is available faster than with SKAdNetwork.
• Privacy compliance – ODM ensures that data is processed locally, fully respecting ATT and DMA requirements.

How advertisers can implement ICM?

To fully benefit from Integrated Conversion Measurement (ICM) and achieve more accurate conversion reporting on iOS and Android, advertisers should follow these steps:

1. Check availability
Contact your Google Account Manager to confirm whether ICM is already available in your region.
The system is currently in beta and is being gradually rolled out to more advertisers.

2. Update Firebase (GA4 SDK)
Ensure that your iOS app uses the latest version of Firebase (GA4), which includes the GoogleAdsOnDeviceConversion (ODM) library.
This enables local conversion matching without sending user data off the device.

3. Define your data path

• First-party data path – for apps where users sign in (e.g., with an email or phone number). These data are hashed and processed locally, offering the highest match precision.
• Event-data path – for apps that don’t require login. The system uses anonymous event data (e.g., app opens, purchases) for probabilistic conversion matching.

4. Update your attribution partner SDK (AAP/MMP)
Install the latest SDK provided by your partner (e.g., AppsFlyer, Adjust, Airbridge, Singular, or Tenjin) with ICM support.

• Enable probabilistic attribution if required by the partner’s configuration.
  Adjust your attribution windows — the recommended minimum is 24 hours.

5. Connect Firebase (GA4) with Google Ads
Make sure your Firebase and Google Ads accounts are linked to allow conversion imports.
Select the key events you want to report as conversions, such as:

• first_open – first app launch,
• signup – user registration,
• purchase – in-app purchase or microtransaction,
• trial_start – trial period start.

6. Choose your integration method

• AAP SDK – standard integration using the partner SDK update and ODM module.
• Server-to-Server (S2S) – sending the odm_info parameter directly from the app to the partner via API.
• App Conversion Tracking API (without AAP) – developer-side integration using the new API v1.1 (available from June 2025).

7. Keep in mind key technical requirements

• ICM does not yet cover iOS traffic from the EEA, UK, or Switzerland – a standalone version (without ODM) is planned for Q4 2025.
• The AAP SDK cannot be blocked – parameters such as odm_info (iOS) and gBraid (Android) must be transmitted even without tracking consent.
• The GA4 SDK is required for ODM to function correctly on iOS.
• The odm_info token must be properly passed between the app and the partner (e.g., via S2S), otherwise conversions will not be attributed.
• Probabilistic attribution operates with shorter conversion windows – the recommended minimum is 24 hours.

Why ICM is a breakthrough?

ICM represents a major step toward balancing privacy and measurement efficiency. It combines deterministic data (gClid, gBraid, Play Referrer, AdID) with probabilistic modeling and on-device measurement, providing more accurate reporting while maintaining full compliance with privacy regulations.

Key advantages:

• Can be used in Google Ads iOS campaigns, for example for ROAS optimization
• Privacy first – no sharing of IDFA or AdID, full compliance with ATT and DMA
• Greater conversion visibility – recovery of some data lost under SKAdNetwork
• Faster reporting – event-level data available with minimal delay
• Flexible integration – via SDK, S2S, or API, depending on the advertiser’s setup
• Future-ready – ICM prepares the ecosystem for the post-IDFA and Privacy Sandbox era

Summary

Integrated Conversion Measurement (ICM) is Google’s new, comprehensive measurement framework for mobile apps that merges deterministic and probabilistic signals to deliver a more complete and privacy-safe view of campaign performance.

ICM enables marketers to:

• recover part of the data lost from iOS and Android,
• optimize campaigns more effectively for ROAS,
• operate in full compliance with privacy regulations.

Implementation requires updating SDKs and working closely with an MMP (AAP), but within the coming months, it is expected to become a key measurement standard in the post-IDFA world.

You can read more about ICM here.