What digital marketers should know about Google Analytics 4
Data is collected differently, stored differently, and even visualized differently. All of this change can be frustrating, but GA4 is quite a bit more advanced than the previous version.
Google Analytics 4 (or GA4) is a completely new version of Google Analytics. The goal of this post is to share the key differences between Google Analytics 4 and the previous version of Google Analytics that you are probably familiar with and highlight the key features that you should be aware of as a Digital Marketer.
What makes Google Analytics 4 unique
If you were around back in 2013 you might remember when the last version of Google Analytics was launched, called “Universal Analytics”. At that time, you needed to update your tracking code to migrate to the new version, but everything else stayed pretty much the same.
Unfortunately, this is not true about the upgrade to Google Analytics 4. Data is collected differently, stored differently, and even visualized differently. All of this change can be frustrating, but GA4 is quite a bit more advanced than the previous version. Those Digital Marketers who migrate to the new version will be rewarded with an Analytics tool that provides a better representation of user behavior, respects user privacy preferences, and allows you to spend less time collecting and aggregating data.
This is possible because of three technologies that Google has been working on for the past few years:
The first is Firebase Analytics. Firebase is a suite of products for developing mobile apps that Google acquired back in 2014. Firebase Analytics uses something called the “event-driven data model” to better describe behavior and measure user engagement. I will speak more about this in a moment, but the important thing to know about Firebase Analytics is that it is the backend for Google Analytics 4. This means that data captured across both websites and native apps now share a single format.
The second technology that GA4 is built on top of is Google Signals. You’ve probably heard about Google Signals before because this is the identity software that uses Google accounts to recognize logged-in users. It is the same method that Google uses behind the scenes to create audiences, and Google Analytics 4 can now use this feature to recognize users instead of relying on first-party cookies.
The Global Site Tag
Finally, Google Analytics 4 is also built on top of the global site tag. The benefit of this technology is that it allows you to make changes in the user interface that actually change the code that is deployed to your website. As an example, with GA4 you can flip on a feature to track when users play a YouTube video, and the code to do this will be automatically deployed to your site without a change in the tag manager.
So that’s a pretty high level overview, and there’s a lot we can talk about next. But I’d like to dig a little bit deeper into what I see as the most important of these features: the event-driven data model.
The Event-Driven Data Model
First of all, let’s remember what the home screen looks like when you log into Universal Analytics today.
As you’ll notice, sessions are undeniably the most important metrics in legacy Google Analytics reports. When someone asks the question: “How much traffic did our campaign drive?”, the answer is usually given in the number of sessions. When someone asks the question: “Are users engaged with the site?”, then the answer is usually bounce rate (a subset of sessions), session duration, or pages per session. When someone asks the question: “How well are we converting against a goal?”, the answer is usually the Ecommerce or goal conversion rate (both of which are calculated with sessions).
The problem with sessions
The problem with this is that the concept of a session can be difficult to apply to mobile and single-page apps, because the fact is that apps are more variable than traditional websites, and the assumptions that we make about how users experience the web do not always hold true for how users experience an application.
As an example, if you’re a runner you might open a mobile app to track your speed and let it run in the background for hours. How many sessions should that create? Are bounce rate and pages per session really useful measures of engagement in this situation?
How the event-driven data model solves this problem
The solution to these problems is the event-driven data model, because it eliminates the concept of a hit type (social, pageview, transaction, etc), and replaces it with three simple components: events, event parameters, and user properties.
This may seem like a small change, but it strips away all of the assumptions that we previously made about the data. When something happens, it is tracked with an event (ex. link_clicked). Parameters are just pieces of information that describe the event (link_text). And user properties are simply pieces of information that describe the user who initiated the event (current_customer). That’s it.
Google did not invent the event-driven data model (numerous products have been applying it to mobile apps for years), but with Google’s market share, GA4 will be the first time it has been applied on such a large scale. So it’s a new concept to most marketers.
The impact of these changes is that page views and sessions are no longer the fundamental building blocks that they once were. They still exist, but you are not required to use them where they don’t make sense because the focus has shifted to users and events. As you can see in the new Home screen for GA4 below, the most important metric in the Google Analytics 4 reports has changed from sessions to users.
Google has been talking about emphasizing users over sessions for years, but GA4 really forces this change.
The event-driven data model also enables a series of new dimensions and metrics that can be generated without relying on the concept of a session. There are several examples of this, but the first one I want to share is the move from “goals” to “conversion events”.
From goals to conversion events
As you may recall, a user completes a goal in Google Analytics when they take some action during their session. If the action is taken multiple times during the session, we would still only count that as a single goal completion.
GA4 has eliminated the concept of a goal, and replaced it with conversion events.
A conversion event is simply any event that you’ve marked as important to your business. So this could be an event to indicate a lead form has been submitted, a video has been completed, an element has been clicked on, or anything else. As you send data to Google Analytics 4, the “Configure > Events” report will populate with all of the event names that have been received. You can send up to 500 unique events, and you simply flip the radio button to mark any event as a conversion from here. Once you do that, you will be able to import these conversions into Google Ads just like you would import a goal.
Acquisition vs. re-engagement
When you are evaluating how well your traffic channels are driving conversions, you now have to decide if you are evaluating how well you are acquiring new customers or re-engaging existing customers. If you choose the “User acquisition” report, your conversions will use first-touch attribution. But if you choose the “Traffic acquisition” report your conversions will use last-touch attribution.
There are three important things that Paid Search Managers should know about conversion events:
- The user can complete multiple conversions within a session
- Each conversion event must have a unique name so that it can be marked as a conversion with the radio buttons I showed a moment ago
- Qualifying for an audience can trigger a conversion event
This brings us to our next topic: audiences.
You might remember that the old version of Google Analytics allowed you to create user segments (for example: all users who added an item to the shopping cart but did not make a purchase). Then, you could promote that segment to an Audience, and share it with Google Ads for remarketing and identifying look-alikeslook-a-likes.
In Google Analytics 4, the concept of a segment has been merged with the concept of an audience. Instead, you simply create audiences. Audiences can be applied to any report, and they can also be shared with Google Ads.
Another thing that is different about audiences is that once you’ve created one, it is automatically shared with everyone else who uses Google Analytics 4. So you do not need to pass links around to your coworkers so that they can download the audience you are using.
And, lastly, Google has launched a series of predictive audiences that can be automatically generated for you (which are similar to the Smart audiences you might be familiar with). These audiences use Google’s machine learning to score the probability that a user will make a purchase or churn within the next 28 days so that you can invest your remarketing budget in reaching the customers who will have the greatest impact.
Ok, so that’s audiences. Let’s talk about engagement metrics.
New Engagement Metrics
A moment ago I mentioned that all of the metrics that were previously calculated based on sessions have changed.
This is important to Digital Marketers because these include all three of the tools that we previously had for measuring the quality of a click: bounce rate, pages/session, and average session duration.
These have been replaced by a new and very important metric that is automatically recorded in GA4 called “engagement time”, which is the amount of time that the user actively viewed your content. If the user is on a mobile app, this is the time that the app was in the foreground. And on a website, this would be the time that the browser tab was active.
Google Analytics 4 then uses this metric to calculate: engaged sessions.
Engaged Sessions & Engagement Rate
An engaged session is a session with greater than 10 seconds of engagement time. You can divide the number of engaged sessions that you had during a time period by the total number of sessions to calculate another new metric “engagement rate”. This is the metric that you will use instead of bounce rate in GA4 (read more about engagement rate here).
Engagement Rate is a much more useful metric for measuring user engagement, especially with sites like blogs and news outlets where a successful session may only include a single pageview.
Now I do want to point out that engaged sessions and engagement rate are both session-based metrics. Sessions have not gone away with GA4, despite the greater emphasis on users. But, we also have a new metric called Active Users.
An active user is someone who has had at least 1 engaged session during the date range that you’ve selected.
If you pull up either of the Acquisition reports you can see how these new metrics are front and center. I expect that a lot of Paid Search Managers are going to struggle to let go of the old metrics, but I actually think that this is a big step forward, and I hope that you’ll find these tools to be useful once you become familiar with them.
Before we move on, there’s one more thing that I want to point out about these new engagement metrics.
Improved data import
None of these are impacted when you import external data. The details on this are a bit technical, but this solves a really big problem with Universal Analytics. If you ever tried to upload offline transactions, for example, you created a bunch of single hit sessions in Google Analytics, which drove up your bounce rate and reduced your pages/session and avg. session duration.
This was very frustrating for a lot of analysts, but since those events do not contribute to engagement time in GA4, they do not have any impact on your engagement metrics. This makes the integrations with Salesforce or call tracking tools much more seamless than they were before.
Okay, I have three more items to discuss, and all of them circle around User Privacy.
New privacy controls
First of all, Google Analytics 4 provides a long list of new privacy controls that marketers can use to ensure they are compliant with the latest regulations.
Disable ads personalization
The first is the option to disable Ads Personalization. This is useful for marketers who would like to use Google Analytics to understand user behavior, but who do not plan to build audiences for remarketing. In this case, a user with “Edit” permissions can completely disable audiences for remarketing so that no one in the company can flip it on.
However, Digital Marketers (such as yourself) also have the freedom to flip this on only within specific geographies. So, for example, it’s now possible to disable Ads Personalization within the EU, but continue to use this feature for all other users.
Not for personalization
But even within a geographical region where you are using Ads Personalization, you can exclude specific events that may be private in nature so that they cannot be used to generate audiences.
Websites and apps that collect medical information are a good use case for this. If you have an event that identifies that the user has generated an appointment with a doctor, you may choose to mark this event as “NPA” (not for personalization) so that no one on your team can create an audience that considers this data point.
So those are the most important new privacy controls that Digital Marketers should be aware of, but I should mention that there are also several others.
How Google Analytics 4 is embracing user privacy
I think that it is important to point out that enabling many of these privacy controls will create gaps in your data. And historically, most Analytics tools have worked very hard to eliminate data gaps like this (for example: we use to write code to detect when users are running an ad blocker, we’ve deployed tricks to help recognize users across domains, or when they log in with a different device, etc.). These new privacy features in Google Analytics 4 actually move in the other direction — they give you more controls to embrace privacy when the user requests it.
And the reason is that Google is taking the first steps to transition us into a world of incomplete data, where we do not rely so strongly on cookies.
Over the past 3 years or so, Safari and Firefox have taken large steps to limit how long a cookie can exist, and eliminate cookies that are used for tracking users across sites. Most marketers don’t realize that the impact of this is already showing up in your data.
For example, most websites are right now showing a higher number of users in Safari than they two years ago. This isn’t because you’re driving more traffic, it’s because the cookies that we use to identify a person are being deleted between sessions if those sessions are more than 7 days apart.
So, Google Analytics has to help marketers prepare for regulatory restrictions that are coming from GDPR and CCPA, but they also have the new burden of helping marketers prepare for technical restrictions that are being imposed by browsers.
In response, Google has announced two features that are coming soon to GA4: Reporting Identity and Conversion Modeling. So I’d like to wrap up with a quick overview of what we know about these features and how they will work once they are released.
Traditionally, Google Analytics has identified a user on the web by setting a cookie (called the Client ID), or by using something called the App Instance ID in a mobile app.
If you’re lucky enough to have logged-in users on your site, you have the ability to set your own unique identifier for users (called the User ID). The benefit of doing this is that you could see how frequently users log in to your site from different devices.
Right now, if you go to your property settings and click on “Reporting Identity” you’ll see two options: “By device only” (which means that you are only using the Client ID and do not have Logged-in users), or “By User-ID, Google Signals, and then device”.
As I mentioned before, this feature will be available for users who are logged in to a Google account on their device and have opted-in to ads personalization (so not everyone). If you enable this feature, GA4 will still use the user ID if it is available since it is the most accurate way to identify a user. But, if the user ID is not available and Google Signals is, then GA4 will use Google Signals to identify the user.
As a result, you will be able to identify a portion of your users across devices, even if they are not logged in. This is important because it means that you will generate very complete data for the small subset of your users who are logged in to Google, using Chrome, and have enabled ads personalization.
Having good information about this small subset of your users will help you fill the data gaps that exist in the rest of the user population. And this is called “Conversion Modeling”.
Conversion modeling is different from Attribution modeling. The idea is that Google uses machine learning to fill the gaps that we know exist in our data. So, for example, if we know that Safari is reporting 100 users on the site last month, we could estimate that you probably only had 80 because 20 of those were the same user with deleted cookies.
The downside of this approach is that we are going to become more reliant on black-box algorithms and estimated data. But the benefit is that we can respect a user’s privacy request without the concern that it will cause our data to be less useful for making marketing decisions.
How to get started with Google Analytics 4
If you’ve made it this far, then hopefully I’ve convinced you to get started with Google Analytics 4. My recommendation is to start today, but take it slow. If you’re running an old version of Google Analytics, you can add Google Analytics 4 tags to a website without impacting the existing Google Analytics implementation.
My recommendation is to dual-tag your site, so that data is sent to both versions of Google Analytics for 6 months or so. This allows you to continue using the old version of Google Analytics for your day-to-day reporting, and spend an hour or so a week looking at the new metrics and pulling reports from GA4. Plan to fully switch over to GA4 entirely in 2022 by removing those old tags.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.