Everything You Need to Know About GA4’s Cross-Devices Tracking
Built on a completely new architecture, GA4 can gather more data and help you reach more insights than ever before.
With six in ten internet users shopping on one device but continuing or finishing on a different one, according to Google¹, being able to uncover the complete user interaction with your website or app, across multiple devices, is essential to so many stakeholders.
Not to say that Google Universal didn’ have the capability to track cross device but GA4 promises to do a better job.
Naturally, this also helps you optimize ad-spend and getter a better understanding of your users’ behaviur across devices.
Pretty exciting, right? But why is GA4 so much better than the old UA?
Keep reading because we’ll walk you through:
- Cross devices tracking: GA4 vs Universal Analytics
- How the user tracking methods work together
- Cross-device tracking methods dive-in
GA4 vs. Universal Analytics
If Universal Analytics was based on sessions and pageviews, GA4 is built around events and parameters.
The users’ journey is and never was linear, as they may use multiple touchpoints and devices before converting on the mobile app or website. A user’s first interaction with your business may come from his smartphone. The next could happen while he’s browsing on his work computer, while the purchase could very well be on his personal computer.
So what’s new in GA4?
First of all, thanks to GA4’s new measurement model, it’s now able to track web and app data together in the same place (through data streams), enabling you to have a more accurate overview of the entire user journey. Moreover, for more granular analysis, you have separate data streams such as iOS, Android, and web.
Although three out of four methods used by GA4, already existed in Universal Analytics, you have now more flexibility in:
- Choosing a reporting identity (more on that below)
- Visualizing the data in one place. In Universal Analytics you needed to create a separate User ID View [and consequently, the data associated with this implementation would have been available only in that view]. In GA4, the data starts flowing into the entire property, once the user_id parameter is sent.
Then, GA4 introduces a fourth method: Behaviour Modelling, which covers the data gaps left by the other three methods.
Now how does GA4 know how to unify this intricate adventure into a single cross-device user journey?
Well, the four methods (also called identity spaces) work together to enhance data accuracy. These four methods are:
- User ID
- Google Signals = Google’s Identity Graph of logged in users
- Device ID
- Modelling
We’ll dive into each one of those shortly after the following setup mentions.
How Cross-Device Tracking Methods Work Together
In your GA4 property, you’ll have three options to choose from to define your Reporting Identity. Notice that the first two options consist in the combination of all or some of the above identity spaces.
Option 1: Blended.
This option first looks for a User ID. If no User ID is collected, GA4 will look for Google Signals information. If there’s no data coming from Google Signals (which is not recommended at all), then Analytics moves on and will look for a Device ID. Finally, if the Device ID is also not available, then GA4 uses Modelling.
Simply put, GA4 goes through every identity space, one by one, and if it does not find any available data, it moves on to the next best method.
This option is the most powerful, taking advantage of 100% of what GA4 has to offer in terms of cross-device tracking.
Option 2: Observed.
Basically, this is the “Blended” option without the Modelling identity space :). However, it’s still a good alternative. GA4 will try to identify the user based on his user ID. If this doesn’t work, it will start looking for Google Signals data. If the user doesn’t allow Ads personalization, Analytics will then resort to Device ID.
In essence, as it was with the first option presented, GA4 takes each identity space one by one until it manages to identify one.
Option 3: Device-based.
This is the most basic of setups. When this method is chosen, the device ID is the only identifier used to generate reports. We don’t recommend it as is the least accurate way to track users.
GA4 cross-device tracking methods
Now let’s take a closer look at each identity space.
1. User-ID
The User-ID is one of the identity spaces that help you better track user behavior across devices and platforms, as well as across different sessions. How is that?
Well, you can tell Analytics how to stitch your own identifiers to individual users, and the outcome will be more accurate user counts and a much clearer roadmap of user behavior on your platform.
As opposed to the other identification methods (which take a click to be activated) this is a custom implementation and it’s one of the most powerful identification methods.
However, this identity space is only available onl websites where people fill in data (forms, logins etc)like email, name, phone number so there is a key generated.
For example,, you assign a string of characters to each user that logs in to your website. For example, you can use the email address of your user when they login to generate a unique ID that you will then reference throughout your website or app. Note that you cannot have Personal Identifiable Information collected by the User ID, but rather random character strings.
Now, what about those users who trigger events before logging in? you ask.
GA4 is smart about that too.
If an event is triggered before signing in, Analytics will use the session ID to associate that session with the user ID given when the sign-in occurs.
What about after logging out?
In the other instance, when an event is triggered after the user has signed out, Analytics will not associate any subsequent events with that user ID.
It’s important to note that the data collected and recorded before the implementation of the User ID will not be reprocessed and associated with a user ID.
If you’d like to have a more accurate view on how many people interact with the website, this customly set-up identity space works really well together with the other easy to turn on identifiers.
Remember, we don’t try to push one of these identifiers; ideally, they should work together towards more accurate tracking.
2. Google Signals
What about what happens when users don’t log in to your website and don’t fill in any identification data?
Well that’s when Google Signals might help. So, make sure you enable it. (we’ll show you below how to do it).
Google signals helps improve cross-device tracking when your users are signed in with Google in their browser, and they have Ads personalization enabled, (which, by the way, is turned on by default), so Google google allows you to use their cross-device identity graphs.
The ease of this reporting space is that you don’t need any manual tracking code implemented (such as in the case of User ID). However, not all users that navigate on your website are Google users as well, so you need to see this method as a companion to the other types of identifiers.
None of these identifiing methods are almighty independently, but they work very well together.
Now let’s take a look at how to, enable Google Signals.
- Navigate to the Admin section of your GA4 property
- In the Property column click Data Settings, then Data Collection
- Toggle the upper right button to enable Google Signals
One more cool side of enabling Google Signals is that once you do, you’ll also see coming data about the users demographics and interests.
3. Device-ID tracking method
This is regarded as the standard method for identifying users.
It only recognizes devices – not users per se.
Why is that? It is a browser-based (getting its value from the client ID, relying thus on the _ga cookie) or mobile-app-based identifier of a unique installation of the app.
So, if John interacts with your website from his smartphone, but then purchases from his laptop, each device/browser will have different cookies, meaning different IDs. So, you’ll not be able to actually correlate John’s visit on the mobile with his purchase on desktop.
When it comes to tracking users accurately, device-tracking does help but for tracking users’ journey accurately needs to be supported by some of the other identifying methods.
4. Modeling
Also known as “Behavioral modeling for consent mode”, this identity space becomes useful for those users who decline identifiers such as cookies.
More specifically, if your website or app displays a consent banner, then without this GA4 feature you are losing the users who disagree with their data being tracked.
To fill the gap of that missing data, Analytics uses information from other users who do accept cookies (also known as observed data). That data is used to create models of the behavior of users who refuse the analytics tracking, which translates into modeled data.
Google does that by using machine learning best practices, which also means there are certain requirements your GA4 property has to meet, such as:
- Minimum 1,000 events per day registered with analytics_storage=’denied’ for at least 7 days
- Minimum 1,000 daily users sending events with analytics_storage=’granted’ for at least 7 of the previous 28 days.
- Consent Mode enabled across all pages/screens of your website and app.
- Moreover, for webpages, you have to ensure that “the tags are loaded before the consent dialog appears, and Google tags load in all cases, not only if the user consents”
Your property becomes eligible for behavioral modeling once all of the above prerequisites are met, and you’ll be able to see that by clicking on the arrow next to each Report name.
Final Thoughts
There’s no debate that GA4 represents an obvious improvement over its predecessor.
By implementing all these identification methods, the users count you see in your CRM vs the users you see in GA4 will tell you the same story.
And it doesn’t matter if you’re a publisher, marketer, web analyst, or conversion rate specialist – GA4 benefits everyone.
If you’re behind your GA4 migration schedule, feel free to reach out to us so we can help smoothen the process. Book your free consultation now or write us an email and let’s see how we can step in and take some of the load off your shoulders.
1 Google/Ipsos Connect, GPS omnibus, U.S. online respondents 18+, n=2,013, Mar. 2016.
