Google ODM and ICM Explained: What App Advertisers Need to Know in 2026

在本期《 Tenjin ROI 101, Ashley Black, founder of Candid Consulting and former Google ad sales lead, explains some of the most misunderstood terminology in iOS app advertising. With nearly a decade inside Google and six years leading the app ad sales team, Ashley shares a perspective that is hard to find: she knows how these products were built, and she knows how they actually perform in the real world.

“Since leaving the company, I think I’ve actually gained even more insight into how the products really work versus how we were told they work internally,” she says. “I bring a unique perspective in that I have firsthand experience on the inside of how things work mechanically, and now how to apply that knowledge.”

The focus is on three products that every iOS app advertiser should understand: ODM, ODM 2.0, and ICM. If those acronyms mean nothing to you right now, that is exactly the problem… 

These tools are core to how Google measures, models, and reports iOS conversions, and whether or not you have them configured correctly has a direct impact on the data you are making decisions with. Ashley breaks down what each one does, who it is actually built for, and what you need to do, or not do, to get value from it.

Let’s start with the most important thing: clarifying terminology. ODM stands for On-Device Measurement. ICM stands for Integrated Conversion Measurement. Both are Google’s responses to the measurement challenges that came with iOS privacy changes.

“When you hear these acronyms, we’re really talking about Google’s response to the challenges of measuring and improving their iOS ad products for app advertisers,” Ashley explained. “They could do a much better job at naming these things, to be honest.”

While both ODM and ICM technically exist for Android as well, they are primarily relevant for iOS. Android has the Play Store and stronger measurement infrastructure, so the needle moves less there. The real story is on iOS, where signal loss has hit advertisers hard.

There are three distinct products under this umbrella:

• ODM (On-Device Measurement) using first-party data
• ODM 2.0 using event data
• ICM (Integrated Conversion Measurement) which launched roughly nine to ten months ago

Each works differently, requires different setup, and delivers different results. Next, we are going to walk through each one.

ODM: On-Device Measurement Using First-Party Data

ODM 1, as it is sometimes called, launched around 2023. At its core, it is Google’s way of privately linking ad clicks to in-app actions using your own data, typically a user’s email address or phone number.

“It does this in a privacy-safe, compliant way,” Ashley noted. But she was also clear about the limitations: “I don’t think we saw a ton of adoption of this, and for a few reasons. First, it does require dev work to be able to share this data with Google. Second, it really only benefits apps that collect a lot of emails and phone numbers.”

For gaming apps, this is rarely relevant. For subscription-based apps, it can be, since email collection is often baked into the onboarding flow.

Even if you do qualify, the performance uplift is modest: 

“You might see a modest improvement, maybe a 10 to 15% reduction in CPA, if you see a difference at all,” Ashley said. And critically, “this only affects Google’s own reporting of conversions. You’re not going to see a big difference in your MMP.”

ODM 1 also tends to have the most impact for campaigns serving on Google’s owned and operated properties like YouTube and Search. If your distribution skews toward the Display Network, it is unlikely to move the needle much.

The setup requires sharing hashed login data, like an email or phone number, back to Google.

“A lot of advertisers I’ve spoken to have shied away from it, saying it feels like a lot of work for not that much gain,” Ashley said.

What you should know about ODM 1: it’s worth exploring if you are a subscription app with a strong email collection and significant traffic on Search or YouTube. Otherwise, it may not justify the implementation effort.

ODM 2.0: Event Data & What Google Hasn’t Told You

ODM 2.0 is a different product entirely. Instead of relying on first-party login data, it uses event data and what Google describes as “temporary anonymous signals from the device itself.”

Ashley explained:

“I’ll be a little more straightforward here than I think Google has been publicly. While Google hasn’t said it explicitly, I think most people in the industry have inferred that ODM 2.0 is also using IP addresses to inform its models, which then leads people to conclude that it involves a form of fingerprinting. I’m not going to go further into that, but that’s what’s been widely inferred from the documentation.”

The good news is that ODM 2.0 does not require anything from you as the advertiser.

“You just needed to make sure you were on the right version of the Firebase SDK,” Ashley explained. “At this point, if you’re keeping your SDK up to date, you’re probably already on the right version.”

It is also universally available, unlike ODM 1.

“The good news with ODM 2.0 is that it’s good for everyone, unlike ODM 1, which had specific requirements around user data and inventory.”

Google rolled out ODM 2.0 gradually by inventory type: 

“From what I understood, by the end of last year, ODM 2.0 had rolled out on Search and AdMob. The YouTube rollout was planned for Q1 of this year,” Ashley said. So depending on where your primary iOS traffic comes from, you may already be seeing its effects, or you might start to soon.

That said, the real-world impact has been measured. “In my personal experience working with advertisers, I haven’t seen anything dramatic as a result of ODM 2.0. Some people may have, but nothing major in my experience.”

Like ODM 1, ODM 2.0 only affects Google’s front-end reporting, not what you see in your MMP.

“You wouldn’t necessarily see it in SKAN. You might, indirectly, if Google gets better signals, it can optimize for better users, which could result in better SKAN performance, but the primary place you’d see it is in Google’s front-end reporting.”

The key takeaway for ODM 2.0: No action needed beyond keeping your Firebase up to date. There’s an upside, even though it might not be so dramatic.  

ICM: Why Google Integrated Conversion Measurement Changes Your MMP Reporting

This is the one that matters most for how you interpret performance data across your stack.

ICM, or Integrated Conversion Measurement, launched in May 2025. Ashley offers a straightforward description: 

“Let’s just call it what it is: Google sharing data with MMPs. That’s what this is.”

For a long time, Google’s modeled iOS conversion data lived entirely inside Google’s own reporting. MMPs had no visibility into it. “The MMPs were essentially saying, ‘We’re not going to report on your numbers if you won’t tell us how you got there,'” Ashley explained. “Eventually, Google came to an agreement with most of the major MMPs and started sharing that information. Now the MMPs are saying, ‘Okay, we feel more comfortable including your data in our own modeling.'”

This means that if you use an MMP like Tenjin, you can now start to see Google’s modeled conversion data reflected in their single source of truth reporting. It is the first time Google’s modeled data has been visible outside of Google’s own platform.

ICM is most valuable in specific scenarios. “It’s most valuable if you use your MMP’s probabilistic reporting. If you’re strictly looking at SKAN, this won’t have any impact on you,” Ashley noted. It also has a significant impact for campaigns with heavy traffic on Search or web display, especially where the ATT prompt is not shown or where opt-in rates are low.

One important distinction: ICM does not make your Google campaigns perform better in the Google dashboard.

“Your Google campaigns will continue to operate the same way they always have. It’s just that the MMP is now reporting more Google conversions because it has more confidence in how Google modeled them.”

So what is the real value? 

“It’s really just for 归因,” Ashley confirmed. “Now, if you’re using your MMP data to feed optimizations back into Google, that could potentially improve performance as a result. But at its core, this is mostly an attribution story.”

To make sure ICM is actually working for you, check your settings.

“In almost all of them, there is some kind of toggle in your attribution settings for Google where you need to enable probabilistic modeling or reporting. So just check your settings and make sure that’s turned on, otherwise ICM won’t be reflected in your MMP.”

When it comes to ICM it is your most important action item if you use MMP probabilistic reporting. So, check your attribution settings and confirm that the toggle is enabled. 

How to Build a Web-to-App Strategy on Google in 2026

Beyond ODM and ICM, Ashley pointed to a trend that is quickly becoming one of the most talked-about strategies in app user acquisition: web-to-app.

“The big trend I’m seeing a lot of testing around, especially with subscription-based apps, is web-to-app strategy. A lot of marketers have been doing this on Meta for a while, and now they’re starting to look at whether they can replicate it on Google.”

But “web-to-app” means different things to different people. Ashley broke it down into three distinct approaches:

1. Google’s official web-to-app product 

This is really deep linking, designed for users who already have the app installed, driving them from Search back into the app. “In my opinion, it’s not what most of the industry means when they say web-to-app.”

2. Web-based ad product driving directly to the App Store 

This uses a web-based ad format, not UAC, driving users straight to the App Store via MMP tracking URLs. No landing page in between. “This is what I see a lot of marketers experimenting with right now.”

3. Web landing page to app 

This drives users to a web-based landing page, maybe a quiz or a long-form flow or a checkout, before pushing them to the App Store. It is often used by subscription apps to avoid store fees.

The second version is where the most energy is right now.

“You gain back a lot of control. You’re essentially running a Search campaign with keywords, and the measurement becomes much more straightforward, pretty much deterministic at that point. So you feel a lot more confident in your results.”

That said, the setup is not simple. “It’s a complicated setup,” Ashley acknowledged. “It’s a lot of steps and can be pretty cumbersome. But once you get it working, the payoff is real.”

On how Google might respond, Ashley was thoughtful.

“I imagine they’d prefer marketers to stay within UAC, since that’s their product for mobile. But I think marketers are somewhat desperate for a combination of performance and transparency. The iOS product on UAC has just been tough. People are trying different ways to leverage Google’s Search and YouTube traffic for iOS, and web-to-app is one avenue they’re exploring.”

Quick Reference: ODM vs ICM

 主要结论

ODM, ODM 2.0, and ICM are not very flashy features, but they are very meaningful tools for anyone running iOS app campaigns on Google. Together, these tools represent Google’s effort to rebuild measurement confidence in a post-ATT world.

As Ashley puts it:

“ODM 1 requires dev work from your team, not a good fit for everyone. ODM 2.0, nothing you need to do except make sure you’re on the right version of Firebase, impact may vary. ICM, better reporting of Google’s modeled conversions inside your MMP, check your MMP settings to make sure it’s enabled.”

Here is what to action right now:

• ODM 1: If you run a subscription app collecting that’s emails or phone numbers and running on Search or YouTube, assess whether the dev work is worth it for your setup.

• ODM 2.0: Keep your Firebase SDK updated. That is it.

• ICM: Check your MMP attribution settings for Google and confirm that probabilistic modeling is toggled on. If it is not, you are leaving reporting accuracy on the table.

• Web-to-app: If you are a subscription app looking for more transparency on Google, this is worth exploring, particularly the web-based ad format driving directly to the App Store.

If you have not looked at your attribution settings recently, now is the time. A few small checks could give you a significantly clearer picture of what your campaigns are actually doing.

营销内容经理

塔拉-迈耶