The introduction of predictive metrics in Google Analytics 4 is a huge game changer for digital marketers. Instead of reports about past behavior of website visitors, Google’s Artificial Intelligence can now give you a glimpse of the future and a clear hint on where to put your marketing focus in the next week(s).
Here is another prediction for you. Read the article and you will impress your clients or boss the next time they talk to you about GA4.
What are predictive metrics in GA4?
Google Analytics 4 has 3 predictive metrics. Simply put, these are based on the data of your site visitors, structured event data that is added and Google’s machine learning algorithms. The combo generates predictions about the future (non-)purchase behavior of your users.
To help you quickly understand this hocus pocus mumbling, I have poured the predictive metrics into a matrix.
As you can see in the image above, the predictions are for the near future, based on the recent past.
What is more important is that they are all related to making money.
- You can lose it (churn)
- Or win it (purchase / revenue)
However, not all websites sell online. And not all websites that allow visitors to make a purchase will be eligible for predictive metrics.
Let’s find out if your site can be qualified at all for GA4‘s predictive analytics power.
Are GA4 predictive metrics available for all websites?
In order to profit from surfing on the waves of the future of marketing, your site needs to meet a couple of conditions.
Let me start with the most obvious one. (Did you see that coming?)
Your website needs to have GA4 installed
I sincerely hope that’s the case for your organization or clients. True, you still have time left, but there are solid reasons to not postpone migrating to GA4.
To support you with a smooth transition, we composed a free GA4 migration checklist.
PRO TIP: When your GA4 property is set up, enable benchmarking data sharing. This will feed anonymous data to the predictive learning models, and as a result, improve the quality of predictions for your account.
To do this, first click on the Admin clog in the left bottom corner of your analytics account.
In the Admin tab, click on Account Settings and select Benchmarking so it turns blue.
Finally, click on the Save button at the bottom of the page.
That was the easiest part.
Let’s move forward to a more challenging one.
You need to track purchase events in GA4
Predictive metrics are only possible when you have configured one of 3 events in your GA4 property:
- purchase: your property needs to send it and you need to collect the value and currency parameters to be eligible for both purchase and churn probability.
- ecommerce_purchase: GA4 is not a full-blown ecommerce analytics tool yet. For now, you can best use the purchase event.
- In_app_purchase: this is collected automatically, but for Android apps, you need to connect your account to your Firebase account.
To check if you have set up the required event(s), navigate to Configuration > Events. Then, look at the list of events.
PRO TIP 1: Use as many event recommendations related to user behavior as possible. This will give AI a better understanding of what visitors are doing and not doing on your site.
PRO TIP 2: Limit events that are not related to user behavior because they can confuse the predictive learning models.
You can now chill and wait…
Predictive metrics need (quite) a lot of data to work
Google Artificial Intelligence is worth nothing without data to process. Besides, the data must be of a sufficient volume in order to be statistically reliable.
You can compare this to teaching a kid the letters A,B and C and then hope (s)he can read Ulysses by James Joyce. Good luck, but it won’t happen.
“Enough” data is vague and relative: it took only 4KB of ram to make Apollo 11 land on the moon.
Luckily, Google uses different measuring units than bits and bytes.
How many users do you need for predictive metrics in GA4?
The predictive models need visits from at least 2,000 users. Half of them must have triggered a predictive condition. The other 1,000 must not have triggered a predictive condition.
This applies to all predictive metrics. Your site can be eligible for one metric, but not for others.
Sounds stressful?
Let me add some more adrenaline to the cocktail that’s now running in your veins.
Are predictive metrics a set and forget feature in GA4?
You absolutely need the minimum number of visits in respectively the last 7 (churn probability) and 28 days (purchase probability and predicted revenue).
Every. Single. Day.
The moment your site no longer meets the requirement set by Google, predictions become unreliable. In that case, you will no longer have access to (certain) predictive metrics in Google Analytics 4.
Let’s stay positive and see what you can do in the fortunate event your site is and remains eligible for predictive metrics.
What can you actually do with predictive metrics in GA4?
There are two practical applications of predictive metrics in Google Analytics 4. We have dealt with them on the Data Driven blog, so I will only briefly mention them here.
Predictive audiences
Predictive metrics make predictive audiences possible. You can target these with, for instance, re-engagement campaigns.
To check if your property is eligible for predictive audiences, you need to follow this path in your GA4 account:
Configure > Audiences > New Audience
On the next screen, you click on Predictive in the Suggested Audiences section.
The status of your audiences will be red (not eligible) or green (ready to use).
LTV exploration
Purchase probability and Churn probability are used in the Lifetime Value Exploration.
That’s the GA4 report that gives you insights into the average revenue of your users, as well as other important money related metrics.
Now you know what the GA$ predictive metrics are and can mean for you, it’s time for a final question.
Are predictive metrics a sign of the future of GA4?
Are predictive metrics a first hint that Google Analytics is shifting towards a predictive analytics tool, rather than a pure reporting tool?
The question popped up in my head one morning while I was desperately trying to wake up with coffee. (To be honest, the question sounded more like Will GA4 predict everything one day?)
My coffee grounds did not give an answer about the future of GA4 as a predictive analytics tool.
The combination of reporting and predicting is an exciting step forward. As data driven marketers, you and I can only hope that this is just the beginning of something bigger.
Key takeaways about predictive metrics
The 3 predictive metrics (Purchase Probability, Churn Probability and Predicted Revenue) in Google Analytics 4 are a big step forward for marketers.
Instead of only knowing what people did on your website, Google’s algorithms help you predict future purchase behavior of your visitors.
Predictive metrics are not available in all GA4 properties. You need to activate specific events that are related to purchasing activities on your site or app.
Another requirement is that your property needs to have at least 1,000 visitors who do and 1,000 visitors who don’t trigger the events for a respective predictive metric.
Finally, when your site is eligible for metrics, you can use them inside GA4’s Audiences and the LTV exploration.
I hope this information will help you boost your marketing efforts. I also hope to meet you here in the near future. Until then, may the force of data be with you.
